WPS6402


Policy Research Working Paper                           6402




                  Entrepreneurship Programs
                   in Developing Countries
                       A Meta Regression Analysis

                                    Yoonyoung Cho
                                   Maddalena Honorati




The World Bank
Human Development Network
Social protection and Labor Unit
April 2013
Policy Research Working Paper 6402


  Abstract
  This paper provides a synthetic and systematic review on                          impact for youth and on business knowledge and
  the effectiveness of various entrepreneurship programs                            practice, but no immediate translation into business set-
  in developing countries. It adopts a meta-regression                              up and expansion or increased income. At a disaggregate
  analysis using 37 impact evaluation studies that were                             level by outcome groups, providing a package of training
  in the public domain by March 2012, and draws out                                 and financing is more effective for labor activities. In
  several lessons on the design of the programs. The                                addition, financing support appears more effective for
  paper observes wide variation in program effectiveness                            women and business training for existing entrepreneurs
  across different interventions depending on outcomes,                             than other interventions to improve business
  types of beneficiaries, and country context. Overall,                             performance.
  entrepreneurship programs have a positive and large




  This paper is a product of the Social protection and Labor Unit, Human Development Network. It is part of a larger
  effort by the World Bank to provide open access to its research and make a contribution to development policy discussions
  around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors
  may be contacted at ycho1@worldbank.org and mhonorati@worldbank.org.




         The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
         issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
         names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
         of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                       Produced by the Research Support Team
       Entrepreneurship Programs in Developing Countries: A Meta
                                             Regression Analysis ∗

                                     Yoonyoung Cho and Maddalena Honorati †




Keywords: Meta Regression Analysis, Entrepreneurship Programs, Microenterprise Development,
Training, Financing, Counseling
JEL codes: O12, O16, J24
Sector Board: Social Protection (SOCPT)



∗
  This paper has been prepared as a part of broader “entrepreneurship project� at the Social Protection and Labor Unit, World Bank,
Washington DC. It is also a background paper for the World Bank’s 2013 World Development Report on Jobs. The study was
funded by the Governments of Austria, Germany, Norway, South Korea, and Switzerland under the auspices of the Multi Donor
Trust Fund on Labor Markets, Job Creation, and Economic Growth. The findings, interpretations, and conclusions expressed here
are personal and should not be attributed to the World Bank, its management, its Executive Board of Directors, or any of its
member countries. We thank David Robalino and David Margolis for their support and insights, Jesko Hentschel and team members
of the World Development Report on Jobs for their valuable comments, and Eshrat Waris for assistance in constructing the data.
We also thank Jamele Rigolini for a thorough review and thoughtful suggestions, and IZA/WB conference participants for their
valuable comments.
†
  Yoonyoung Cho(Corresponding): The World Bank and IZA, email: ycho1@worldbank.org; Maddalena Honorati: The World
Bank, email: mhonorati@worldbank.org
1.       Introduction

Fostering entrepreneurship is widely perceived to be a critical policy agenda to expand

employment and earning opportunities and to reduce poverty. Sound macroeconomic conditions

and business environment including infrastructure, regulation, and legal environment have

typically been emphasized to increase entrepreneurial activities and create jobs. While these

remain relevant, in developing countries, increasing attention is being paid to the role of labor

policies that aim to reduce constraints and enhance productivity among the self-employed and

small-scale entrepreneurs. 1 This is particularly pressing in developing countries where wage and

salary employment is limited and the majority of jobs are created and operated in self-employment

(Haltiwanger et al., 2010; Ayyagari et al., 2011; Gindling and Newhouse, 2012). The demographic

pressure, including youth bulge in many countries in Africa and South Asia, adds urgency to

creating more jobs and provides a justification of entrepreneurship promotion to absorb the large

inflow of workers.

         In recognition of the importance of self-employment in job creation, interventions to

promote entrepreneurial activities (hereafter “entrepreneurship program�) are increasingly being

implemented around the developing world. Such entrepreneurship promotion programs largely

vary by objectives, target groups, and implementation arrangements, and often combine different

types of interventions depending on the constraints to entrepreneurial activities that each program

aims to address. Based on the evidence that some entrepreneurial traits and skills are strongly

related to business set-up and success 2 and that access to finance is a dominating constraint to


1
  Note that the terms “self employed� and “entrepreneurs� are used interchangeably. Also, small-scale entrepreneurship is used
instead of microenterprise or subsistence business.
2
  For example, Ciavarella et al. (2004) using data from the United States find a strong relationship between some attributes of
personality (measured by the Big Five: conscientiousness, emotional stability, openness, agreeableness, and extroversion) and
business survival. Crant (1996) also points to personality as a predictor of entrepreneurial intentions.


                                                              2
entrepreneurship, programs have provided individuals in the labor market with the opportunities of

training, counseling, and access to finance. Frequently used interventions include technical

(vocational), business (managerial), and financial skills training, financing support such as

microcredit loans and grants, and counseling ranging from mentoring and advisory services to

post-program consulting. 3 Outcomes of interest range from labor market performance such as

employment, business creation, hours of work, earnings, and profits and business performance to

supply side changes such as improved technical and non-cognitive skills, business knowledge and

practice, attitudes, aspirations, and more active financial behavior (borrowing, saving). Target

groups are also very diverse with different population groups facing different barriers to

entrepreneurship and self-employment (women, youth, social assistance beneficiaries, etc.). Some

programs target potential entrepreneurs (the unemployed, school drop-outs, or graduating

individuals) to foster self-employment and new business creation; others target existing

microentrepreneurs or microfinance clients to increase their productivity. Programs can be further

modified according to the context of the policy environment, reflecting cultural factors (fear of

failure or beliefs on gender roles), infrastructure (water and electricity), and legal and regulatory

conditions (high entry barrier due to administrative hassles), among others, that can hinder

individuals from starting and growing a business. 4

          Evidence on the effectiveness of entrepreneurship promotion programs is still scarce, and

findings from existing impact evaluations are widely heterogeneous. Early evaluations from Latin

America’s Jovenes programs targeted to vulnerable youth, though not conventional

entrepreneurship programs, suggested that vocational and life skills training combined with an

3
   In addition, microfranchising, value chain inclusion, small business networks, support for technology transfer, business
incubation, and many others are being implemented.
4
  Microfinance programs, for instance, often target female entrepreneurs in order to address issues related to a cultural factor while
relieving credit constraints.


                                                                  3
internship in private firms could be potentially useful for self-employment promotion as well

(Attanasio et al., 2011; Card et al., 2011). More recent impact evaluations of skills training for self-

employment and business development targeted at vulnerable individuals in Malawi, Sierra Leone,

and Uganda, for instance, found generally positive effects on psycho-social well-being but mixed

results in labor market outcomes (Cho et al. 2012; Blattman et al., 2012; Casey et al., 2011,

respectively). The complexity increases as the training programs combine other financial and

advisory support especially for social assistance beneficiaries (Almeida and Galasso, 2009;

Carneiro et al., 2009; Macours et al., 2012). And even the seemingly similar programs have

heterogeneous results in different places (in Peru, business training programs from Karlan and

Valdiva, 2011; in Tanzania, Berge et al. 2011; in Bosnia and Herzegovina, Bruhn and Zia, 2011).

Likewise, the effects of financing through microcredit or grants vary widely across studies. A

series of studies in Sri Lanka suggest that the returns to capital were large and grants significantly

improved labor market (business) outcomes (De Mel et al., 2008a; 2008c; 2012). However,

evaluations on the effects of expanding access to credit in Mongolia, Bosnia and Herzegovina,

India, South Africa, Morocco, and Philippines (Attanasio et al., 2012; Augsburg et al., 2012;

Banerjee et al., 2009; Karlan and Zinman, 2010; Crepon et al., 2011; Karlan and Zinman, 2011,

respectively) suggest that access to credit did not automatically improve entrepreneurial activities.

       In this paper, we exploit the heterogeneity of design and implementation features to shed

light on the effectiveness of the programs. We synthesize the impacts of different entrepreneurship

programs and disentangle the effects of design factors from those of context and study

characteristics using a meta regression analysis. A meta analysis is a statistical procedure of

combining the estimated impacts of multiple studies in order to draw more insights and greater




                                                   4
explanatory power in probing differential program effects. 5 Since a meta analysis examines the

extent to which different program and study characteristics – design and implementation features,

data sets, and methods of analysis – affect estimated results, it is particularly useful to synthesize

the findings from various studies on a similar topic.

         An important contribution of this paper lies in the coverage of programs and methodology

in the meta analysis. Although many entrepreneurship programs are being implemented and

evaluated in developing countries, to our knowledge, few attempts have been made to review the

impacts of such interventions in order to synthesize emerging lessons. David Roodman’s open

blog reviews existing microfinance and saving programs, and concludes that such programs do

little to reduce poverty on average with large heterogeneity. McKenzie and Woodruff (2013)

provide narrative reviews on business training programs and find that their overall impact is

generally small and that a knowledge gap exists in understanding what programs work better and

how. Karlan et al. (2012) also review microenterprise development programs that include business

training and capital infusion, and highlight heterogeneity in results. Our paper, by cross-examining

the effectiveness of diverse entrepreneurship programs and by disentangling the contribution of

various factors in explaining success, provides a systematic review based on impact evaluation

results that are available and draws synthetic lessons. In this sense, this paper is in line with recent

studies such as Card et al. (2010) and Kluve (2010), which examine the effectiveness of various

active labor market programs in developed countries based on meta analysis. However, unlike

Card et al. (2010) and Kluve (2010), we move beyond positive significance in examining program

effects. We indeed construct the effect size for each observation based on the coefficients, standard

5
  See Stanley (2001) for discussion on the methodology of meta analysis in synthesizing multiple studies. There has been useful
synthetic research that employed this meta analysis method in the field of labor market analysis. For example, Jarrell and Stanley
(1990) and Stanley and Jarrell (1998) examined the magnitude of wage gaps between union-nonunion and male-female workers,
respectively, using multiple studies that estimated the gap.


                                                                5
errors, and sample size reported in each study. This allows us to discuss program success without

confounding it with statistical power from sample size and obtain more robust measures to assess

the effectiveness of the programs, which we will discuss below in more detail.

         We find that the impacts of differential combinations of interventions vary depending on

the outcomes of interest and target groups as well as the specific context. Overall, entrepreneurship

programs have a positive and large impact for youth and on business knowledge and practice, but

no immediate translation into business set-up and expansion or increased income. At a

disaggregate level by outcome groups, providing a package of training and financing is more

effective for labor activities. Additionally, financing support appears more effective for women

and business training for existing entrepreneurs than other interventions to improve business

performance. Our findings suggest that involving the private sector in program delivery can

enhance the effectiveness.

         The meta analysis, of course, is unable to completely resolve the technical issues that are

embedded in the original studies. 6 It is unclear whether a program’s performance is driven by its

design or implementation from an individual study. By synthesizing multiple studies, we can infer

which design and implementation features are more associated with positive impacts, on average.

However, because available information differs across studies, it is possible that the meta analysis

may omit important determinants of program performance. We undertake estimations based on

different models to address unobservable variables. Another caveat of meta analysis is that the




6
  It may be the case that if the impact evaluation was not well powered against certain outcomes due to insufficient sample size, it
will more likely yield insignificant impacts even when the true impact exists. Even if an overall impact is well examined for the
general target group, heterogeneous impacts on subgroups may suffer more from insufficient power (Card and Krueger, 1995).
Similarly, insignificant results are less likely to be written up and reported in a study. Since, in the meta regression, we use all
significant and insignificant estimates in the study that are relevant in terms of outcome of interest (we report on average 25
estimated per study), we are automatically absorbing the methodological bias originally present in the study.


                                                                 6
results may change if more impact evaluation studies are added. 7 Therefore, findings and

conclusions of this meta analysis need to be interpreted with caution, keeping these limitations in

mind.

          The next section describes the procedure for constructing data and discusses main features

of the entrepreneurship programs in our sample studies. Section 3 presents a standardization and

estimation strategy using meta regressions and discusses methodology. Section 4 then discusses

the main findings of the meta analysis. Finally, Section 5 concludes the study.



2.        Constructing a Data Set for the Meta Analysis



2.1. Selection Criteria and Search Strategy



To comprehensively collect studies that evaluate entrepreneurship programs, we apply the

following selection criteria. First, we include impact evaluations of interventions that aim at

promoting entrepreneurial activities of potential or current entrepreneurs. These are interventions

targeted to address various external and individual constraints to entrepreneurship, such as skills,




7
  There are quite a few studies in the pipeline that did not meet our March 2012 criteria, but are advanced in presenting results. Cho
et al. (2012) examined the effects of vocational and business training through apprenticeship on vulnerable youth in Malawi, and
found little impact on business set-up despite large positive impacts on intermediate outcomes such as business knowledge and
psycho-social well-being; De Mel et al. (2012) investigated the impacts of business training and grants on the set-up and growth of
female enterprises in Sri Lanka, and found that the training expedited business set-up for potential entrepreneurs and the package of
training and grants improved the performance of the existing enterprises; Karlan et al. (2012) investigated the role of business and
managerial skills improvement through business consulting in improving the performance of microenterprises in Ghana, and found
little evidence of profit increases and the entrepreneurs revert back to their old practices after about a year; Abraham et al. (2011)
investigated the access to savings on consumption smoothing and insurance against risks for micro-entrepreneurs in Chile and
found positive impacts; Bandiera et al. (2012a) examined the effectiveness of the BRAC’s ultrapoor entrepreneurship training and
coaching intervention targeted to poor women in Bangladesh and found substantial increases in assets, savings and loans, and
improved welfare. Similarly, Bandiera et al. (2012b) found that combining vocational training for business creation with
information on risky behavior and health and providing a place to socialize increase the likelihood of engaging in income
generating activities by 35 percent for adolescent girls in Uganda.


                                                                  7
credit, information, cultural norm, and regulations, 8 including training related active labor market

programs (ALMPs) designed to enhance technical skills for self-employment and small-scale

entrepreneurship as well as financial support intervention. However, pre-employment technical

and vocational education and training (TVET) and programs solely promoting wage employment

or large-scale firms are not considered here. 9 Similarly, evaluations of interventions facilitating

access to financial products including micro-insurance or saving are excluded unless they focus on

the outcomes of entrepreneurial activities. 10

         Second, only impact evaluation studies that rigorously estimate the effects using a

counterfactual based on experimental or quasi-experimental design are selected. Many programs

whose evaluation is dependent on anecdotal evidence or tracer studies without appropriate

comparison between treatment and control groups are not considered. Unfortunately, renowned

programs such as Grameen Bank’s microcredit program, large-scale programs such as Know

About Business (KAB) that are being implemented in many countries by the International Labour

Organization (ILO), and many programs by innovative non-governmental organizations (NGO),

including Accion International, Ashoka, and Youth Business International, could not be

considered.

         Third, given that the main interest of this paper is to examine the effects of

entrepreneurship interventions as a tool to reduce poverty and improve the livelihoods of

individuals in developing countries, we focus only on the studies undertaken in developing




8
  See Banerjee and Newman (1993) for occupational choice model and its constraints.
9
  Examples include De Mel et al. (2010) impact evaluation of a wage subsidy program and Bloom et al. (2012) study on the effects
of supporting a large firm in India.
10
   Among the programs to insure individuals against risks, they are included, for example, if they hedge the negative impacts of
weather on their agri-business, but are excluded if they provide access to health insurance.


                                                               8
countries over the past ten years. Some well-documented studies on developed countries are

excluded here. 11

            Finally, manuscripts are included in this analysis only when they are available in the public

domain as a working paper or published paper by the end of March 2012. Ongoing studies, where

project description, impact evaluation design, and some preliminary results are available but the

draft paper is not, are excluded for now. Adding these studies in the future could change the

overall findings from our analysis.

            Based on the above-mentioned criteria, we first collected papers from the literature review

and references in early studies. We also used web-based search functions such as Google Scholar

and Ideas to find recent working papers. In doing so, we relied on the major working paper

domains such as the National Bureau of Economic Research, World Bank Policy Research

Working Paper Series, and IZA Working Papers.



2.2. Coding and Sample Overview



Using the selected papers, we gather detailed information on outcomes of interest, and intervention

and study characteristics. Intervention characteristics include intervention types (training or

financing, for example), duration of intervention, location (country, urban/rural), and target group

(youth, women, microcredit clients). Study characteristics include methodology (experimental vs.

quasi-experimental), sample size used in the study, and publication format (peer reviewed journals

vs. working papers). Other information we extract includes whether government agencies,

international donor agencies, NGOs and community organizers, universities and researchers, or


11
     Examples include Cole and Shastry (2009) on the United States and Ooseterbeek et al. (2010) on the Netherlands.


                                                                  9
microfinance institutes (MFI) and banks deliver the interventions. When the core information

could not be obtained from the paper, we directly contacted authors to provide supplementary

information.

            And more importantly, we extract information on the effect of the program. The primary

measures of the effect that are comparable across studies include: an indicator whether the

program had a positive and significant effect and a ‘standardized effect size’ reflecting the size of

effects on the treatment compared to the control group as a proportion of its standard deviation –

whether it be probability difference, percentage growth, or changes in levels. An indicator of

positively significant effect measures whether the program works, whereas the standardized effect

size measures the extent to which the program affects the outcomes. We use both measures to

conduct our meta regression analysis, as discussed in more detail in Section 4.

            Most of the studies contribute multiple observations because they examine more than one

outcome and different beneficiary groups. For example, when the impact of a particular

intervention on business practice is examined and the business practice is reflected in two

measures – indicators of book keeping and separation of personal and business account – both

observations are counted for the outcomes of business practice. Whenever available, we record

separate estimates for subgroups such as women and youth, which multiplies the number of

observations. However, when multiple specifications are used to estimate a particular outcome, we

use a weighted average of the estimates applying the number of observations as weights instead of

counting them as separate observations.

            The final data set includes 37 impact evaluation studies and 1,116 estimates for six

different types of outcomes. 12 The number of estimates collected from each study is larger than


12
     See Table A1 in the Appendix for the complete list of studies that are used here.


                                                                    10
Card et al. (2010), given a broader set of outcomes of interest and the diversity of programs

considering the nature of labor markets in developing countries. 13 The studies are from 25

countries across all six regions – AFR, EAP, ECA, LAC, SAR, and MENA. 14 Most of the

estimates are concentrated in LAC (28 percent), SAR (19 percent), and AFR (17 percent), and two

thirds of the interventions come from low income or lower middle income countries (see Figure 1).

Out of 37 studies, 16 are published in peer reviewed journals while the remaining 21 studies are

working papers. About three quarters of the studies and 80 percent of the estimates are obtained

from experimental interventions.



2.3. Constructing Key Variables



2.3.1. Outcomes of Interest

Table 1 presents a summary of the distribution and definitions of the main outcomes of interest.

The most commonly measured outcomes are labor market income and profits (27.7 percent)

followed by labor market activities (21.7 percent). Business start-up or expansion, increased

employment and hours of work, and reduced inactivity are coded as positive outcomes for labor

market activities. With respect to income and profits, a range of variables from individual salary to

business profits and assets and to household consumption that captures broad welfare is included.

Given that most small businesses operate at the household level and that individual earnings from

self-employment are often indistinguishable from business profits, they are coded together as labor


13
   For comparison, Card et al. (2010) collected 197 estimates from 97 studies focusing only on labor market outcomes. In our study,
we broaden our estimates of interest to other outcomes in addition to labor market ones; we collect 1,116 estimates from 37 studies
out of which the number of estimates for labor market outcomes are about 530. On average, we collect 25 estimates per study.
14
   The regional category follows the classification of the World Bank. AFR presents Sub-Saharan Africa, EAP- East Asia and
Pacific, ECA- Eastern Europe and Central Asia, LAC- Latin America and the Caribbean, SAR – South Asia, and MENA- Middle
East and North Africa.


                                                               11
income. Business performance then includes measures to capture the size and revenue of business

activities such as sales, number of employed workers, and size of inventories. Business knowledge

and practice includes record keeping, registration, and separation of individual and business

accounts that could potentially affect business performance. Acquisitions of business loans, saving

accounts, and insurance plans that could affect the resource allocation of a business fall into the

category of financial behavior (saving/borrowing). Finally, attitudes toward risk, confidence and

optimism, and time preference that may be related to entrepreneurial traits are coded as attitudes.



2.3.2. Type of Intervention

The interventions analyzed in the sample of our studies can be broadly classified in the following

types: training, financing, counseling, and the combinations of them. Training is disaggregated into

three subcategories: technical and vocational, business and managerial, and financial skills

training. Technical and vocational training includes basic skills that would be essential for self-

employment in certain occupations – such as electricians, mechanics, tailors, bakers, plumbers, and

handy men. The distinction between business and financial training is not always clear. Business

training teaches general practice and business applications including, for example, book keeping,

calculating profits, separating personal and business accounts, and managing inventory; financial

training is usually more specifically tailored to managing profits, making inter-temporal decisions

on investment and saving, and accounting.

          With respect to financing, it is disaggregated into (micro)credit for business or consumer

loans, 15 cash and in-kind grants, and access to financial products such as saving accounts and


15
  We code “microcredit� also as those interventions that test specific design features of a microcredit program. For example, when
the “treatment� under evaluation is a change in the rule or structure of the loan repayment, bigger size loans or group liability versus
individual liability, we consider them “microcredit� treatment for “microcredit clients.�


                                                                  12
micro-insurance. Counseling is seldom used as a stand-alone program; rather it usually is added to

the main intervention. About 41 percent of estimates include training, 67 percent financing

support, and 21 percent combine counseling (See Table 2). Among the different types of

interventions, microcredit programs are by far the most common followed by business training

(See Figure 2). Depending on the outcomes of interest, certain interventions are more frequently

used than others: financial training seems more relevant for business outcomes, whereas technical

and vocational training can be useful for labor market outcomes. (See Table A3 in the Appendix

for the distribution of type of intervention by outcome groups.)



2.3.3. Beneficiaries

Programs are often targeted to women, youth, microcredit clients, social assistance beneficiaries,

and existing business owners. (See Table A1 in the Appendix for the target group of each paper

used here.) In cases where programs are not targeted to a certain group, studies report outcomes by

gender, region, or education, and it is not uncommon that the effectiveness of programs is

heterogeneous across different groups. For example, studies that examine the returns to financing

support for microenterprise owners in Sri Lanka (De Mel et al., 2008a; 2008b) found that the

impacts were not as strong for women. Such gender disparity in program effects is also found in

training programs (Cho et al., 2012; Berge et al., 2011). In order to capture the heterogeneous

effectiveness by types of beneficiaries, we construct mutually non-exclusive indicators for female,

youth, microcredit clients, social assistance beneficiaries, business owners, individuals with

primary or higher education, and individuals in urban areas, using information on target population

and the subgroup of estimation. Therefore, the dummy variable “female� is equal to one either




                                                 13
when the outcome has been estimated for the subsample of female or, by definition, when the

program is targeted to women.



2.3.4. Service Providers and Delivery Mode

Whether different delivery modes of service make a difference is of interest, as implementation

quality is deemed as important as the design of programs. Bali Swain and Varghese (2011)

examine the effects of business training for microcredit clients and compare the effectiveness by

delivery modes. Their findings suggest that an NGO linkage model seems to yield better results

than programs organized and delivered by the banks. In line with this, we group service providers

into government agencies, international donor agencies, NGOs and community organizers,

universities and researchers, and microfinance institutes (MFI) and banks. It is noted that programs

often rely on multiple service providers, and the indicators of service providers are not mutually

exclusive (See Table 3). 16 In addition, we construct an indicator to identify whether the main

program is delivered by the private sector.



3.       Standardization and Estimation Strategy


3.1 Standardization



The effects of the particular interventions that we measure differ across indicators and studies, and

need to be standardized for comparability. Given that the sign and significance are neutral to the

unit of measurement yet indicate whether the program worked, one simple way of standardization

16
  When multiple agencies are involved, it is usually the case where government agencies are working with NGOs or researchers are
collaborating with government agencies or NGOs.


                                                              14
is to focus on the sign and significance of the outcomes. As used in Card et al. (2010) and Kluve

(2010), ordinal indicators of positively significant, insignificant, and negatively significant effects

can be compared across different variables and studies. Noting that relatively few observations

have negatively significant effects (about 4 percent of the entire sample), we focus on the indicator

of positively significant outcomes versus non-positive outcomes.

       The second measure to synthesize the findings across studies is to use a standardized effect

size, thereby allowing diverse studies and outcomes to be directly comparable on the same

dimensionless scale. While the statistical significance provides an answer for the “whether the

program worked or not� question, the effect size provides a measure of “how well did it work�

without confounding the effectiveness of a program with the sample size used in a study. Given

that there is a large variation in the sample size of studies used here, the standardized effect sizes

would provide a more reliable, comparable measure of program success.

       The true effect size ( θ ) is the mean difference between the treatment (𝜇 𝑇 ) and control

groups (𝜇𝐶 ) as a proportion of the standard deviation(𝜎):
                                                    µT − µ C
(1)                                            θ=
                                                       σ


The simplest and most intuitive form of its measurement would be based on mean differences in

data called Cohen’s g (Cohen, 1988), defined as:
                                                    YT −Y C
(2)                                            g=
                                                       sp


where YT is the mean of the experimental group, YC is the mean of the control group, and s P is

the pooled sample standard deviation using each group’s number of observations ( nT for treatment

and nC for control group) and standard deviations ( sT for treatment and sC for control group):

                                                    15
                                                      (nT − 1) sT + (nC − 1) sC
                                                                   2                 2

                                             sp =
(3)                                                      (nT − 1) + (nC − 1)



            The effect size then captures the magnitude of an impact of treatment compared to the

reference group regardless of the unit of measurement. Though intuitively simple, Cohen’s g is a

biased estimator, because the pooled standard deviations of the two groups may not capture the

true population value and may produce estimates that are too large especially when the sample size

is small. We use the following Hedge’s d to correct the upward bias: 17



                                                                       3
(4)                                                 d = g (1 −                   )
                                                                 4(nT + nC ) − 9

Note that d is the “effect size� of the intervention that we use throughout the paper.



3.2 Distribution of Program Effectiveness



As mentioned above, the positive significance and effect size of interventions are the two statistics

that we focus. Table 4 presents summary of the estimated impacts by outcome groups. 18 A few

stylized facts are observed. First, the proportion of insignificant estimates is quite high. In

particular, about 28 percent of the estimates are positively significant while 68 percent

insignificant, 19 and the average effect size for positively significant outcomes is 0.183. Looking

only at estimates on the outcome of labor market activities, the average effect size for positively

17
     See Cooper and Hedges (1994) for detailed discussion and derivation of Hedge’s d.
18
  See Table A2 in Appendix for average effect sizes by intervention types, population groups, providers, regions and
country income levels.
19
  These numbers are calculated at 10 percent statistical significance level, which will be our measure of statistical significance from
here.


                                                                 16
significant impacts is 0.192. Compared to the effects of ALMPs on labor market activities in

OECD countries summarized in Card et al. (2010), where 39 percent are positively significant, 36

percent are insignificant, and the average effect size for positively significant outcomes is about

0.21, the estimates in our study show greater prevalence of insignificant outcomes and slightly

lower effect size among the programs with better results. While highlighting the overall difficulty

of promoting entrepreneurship in developing countries, this reality again emphasizes the

importance of learning what types of programs work for whom and in which context.

         Second, the success rates (proportion of estimates that is positively significant and whose

effect size is greater than 0.1) as well as the dominant type of intervention largely vary with the

outcomes of interest. We plot our two measures of program success – significance and effect size –

by types of interventions and outcomes of interest (See Figure 3). The type of intervention, based

on its main component, is broadly categorized as training, financing, and combinations of the two.

We observe that training is more common than financing in promoting good business practice,

which tends to show more successful results, but less common in changing attitudes, which shows

a lower success rate. Obviously, these are stylized facts from which it is unclear whether the

differences in performance are due to the different type of intervention or the nature of outcomes.

         Third, stand-alone financing programs are associated with poorer results compared to the

programs with a training component. However, this may mask heterogeneity among the large

number of microcredit interventions. Some programs merely estimate the effects of change in the

rule of the loan, including the repayment structure and unit of clients, rather than that of the loan

provision. 20 In such cases, the effects can be smaller than other programs comparing the outcomes


20
   Such studies include those investigating group versus individual loans or ordinary versus commitment group without a control
group that does not have any access to loans (Gine and Karlan, 2009; Brune et al., 2011), and those evaluating the changes of the
rule of the loan (Field et al., 2010b; Gine and Yang, 2009).


                                                              17
with and without the loan, and therefore, whether the program is provided to already existing

microcredit clients needs to be considered.

       Finally, the effect size and t-statistics do not always yield the same conclusion about the

effectiveness of a program as mentioned above, although they are highly correlated in all the six

plots. We observe some estimates whose effect sizes are large but not significant probably due to

low statistical power and vice versa.



3.3 Estimation Strategy



We now turn to a regression analysis to examine the relationship between program effectiveness –

measured by the significance and effect size – and characteristics. The observable characteristics

of programs are broadly categorized as outcome groups such as labor market activities and

financial behavior; design factors such as the types of interventions, beneficiaries, and service

providers; and study characteristics such as an indicator of journal publication. The outcome

groups and design factors vary across and within studies whereas study characteristics remain

constant within studies. Our main interest would be to infer the impact of design factors on

program effectiveness.

       Assuming that the program effectiveness (𝐸 ) is, on average, explained by exogenous,

observable characteristics, the measure 𝐸 of an individual outcome 𝑖 of a study 𝑗 is written as a

linear form as:

                                               𝑜            𝑑
(5)                           𝐸𝑖𝑗 = 𝛼 + 𝑋𝑖𝑗 𝛽𝑜 + 𝑋𝑖𝑗 𝛽𝑑 + 𝑋𝑗𝑠 𝛽𝑠 + 𝜀𝑖𝑗




                                                           18
where 𝑋 𝑜 , 𝑋 𝑑 , and 𝑋 𝑠 contain observable characteristics of outcome groups, design features, and

study characteristics, respectively; and 𝜀𝑖𝑗 denotes an error term that is assumed to be independent

from the vector 𝑋 = {1, 𝑋 𝑜 , 𝑋 𝑑 , 𝑋 𝑠 }. We use the standard probit and Ordinary Least Squares (OLS)

model to estimate the positive significance and effect size based on equation (5).

        However, assuming that some study/program heterogeneity cannot readily be modelled by

the observable variables, we also incorporate the unobservables into our estimation. Unobservable

characteristics such as implementation quality (θj ) – how well the service is delivered to the

beneficiaries – may affect program effectiveness even after controlling for the observable duration

and service providers of training, for instance. Then equation (5) is modified as the following:



                                                 𝑜            𝑑
(6)                             𝐸𝑖𝑗 = 𝛼 + 𝑋𝑖𝑗 𝛽𝑜 + 𝑋𝑖𝑗 𝛽𝑑 + 𝑋𝑗𝑠 𝛽𝑠 + θj + 𝑢𝑖𝑗




where 𝑢𝑖𝑗 is a white noise independent of 𝑋 and θj .

        If unobserved characteristics are deemed to be invariant within each study regardless of the

outcomes, a study fixed effect model can be used to estimate the equation (6). For example, if the

implementation quality affects all beneficiaries in the same way within the program, the study

fixed effect model would remove the effects of unobservable implementation quality, capturing

only the differential effects of observable characteristics. Meanwhile, if θj varies even within a

study as well as across studies and is assumed to be taken by a random draw, the effect of θj also

needs to be measured and a random effect model can be used. Whether the unobserved part of the

study is fixed or variant is not straightforward a priori. We examine the robustness of the results

based on the different models described here.

                                                            19
4.     Results of the Meta Regression Analysis

As discussed in the previous section there are considerable variations in effect sizes across

outcome indicators, types of programs, and beneficiaries. Now we move on to a meta-regression

framework to analyze how differences in the magnitude and significance of estimated impacts are

associated with differences in outcomes variables, intervention design and implementation

features, country and study characteristics. The richness of our database allows us to include in the

model specification many potential determinants of program success apart from outcome category.

At the program level, covariates include the key variables discussed above such as types of

interventions and beneficiaries, and service providers. At the study level, we look at variables

including the impact evaluation design, publication format, study sample size, and the time interval

between program completion and end-line data collection. We begin with a pooled regression that

allows statistical power with a large number of observations, and move on to a more disaggregated

analysis that provides more detailed information with reduced statistical power.



4.1 Pooled Regressions



We analyze how the likelihood of yielding positive and significant effects is associated with the

potential determinants of program success. The probability of observing significant positive

outcomes can be described by a probit model for the event of a positive significant impact. Table

5 presents a series of probit models results to fit the likelihood of a significantly positive program

estimate. We examine the main dimensions of program heterogeneity separately (first 4 columns)




                                                 20
and simultaneously (5th column) controlling for region, country income, and study characteristics

throughout all specifications. 21

         We find that business practice and labor market activity outcomes are more associated with

positively significant impacts than labor income outcomes (omitted category) by 46.7 and 35.4

percentage points respectively (column 1). On the other hand, the likelihood of yielding a positive

and significant impact is not statistically different among labor income, business performance,

financial behavior, and attitudes outcomes, which is consistent with Figure 3 shown above. This

suggests that changing business knowledge and practice may be relatively easier than changing

behavior and increasing income, at least for the short term. 22 This finding may be different if the

long term impacts are estimated as knowledge fades away and impacts on labor market outcomes

are materialized.

         The model in column 2 examines the probability of program success by intervention

types. 23 We classify programs into ‘Training Only’, ‘Training + Counseling’, ‘Financing Only’,

‘Financing + Counseling’ and ‘Financing + Training� (omitted group). 24 Results show that

differences across interventions in the chances of success are not significant on average, although

the sign and magnitude of the coefficients suggest that training combined with counseling is more

promising than the others.

         A clear pattern emerges when comparing program estimates by population groups (column

3). Program impacts estimated for youth and the urban population are more likely to be positive

and significant than estimates for the general population. To the contrary, programs for


21
   Given the large variation in the number of estimates coded per study (from 2 to 70), we weigh regressions by the inverse of the
number of observations/estimates per study in all models to increase the relative weight of under-represented studies.
22
   End-line surveys for our sample of impact evaluations take place, on average, 18 months after the completion of the program, and
about three quarters of estimates are measured within two years.
23
   Note that we are using “program success� for having positively significant impacts at 10 percent level.
24
   See Appendix Table A3 for the distribution of estimates by types of intervention and outcomes of interest.


                                                               21
microfinance clients are less likely to yield positive impacts. Differences in program results by

gender, education, enterprise ownership, and social assistance dependency are statistically

insignificant. It is worth clarifying that the model includes dummy variables for the population

group for which the effect has been estimated; it does not necessarily capture whether the program

is targeted to that particular group.

            Compared to the case of having multiple agencies involved in program delivery (omitted

category), the programs delivered solely by banks or MFIs are less likely to be associated with

program success (column 4). NGOs are associated, though weakly, with better performance. This

finding suggests that programs could work better when delivered by providers that have strong

connections with the beneficiaries and are familiar with local contexts.

            Column 5 presents a model including all four dimensions of the covariates analyzed in

columns 1-4. That the intermediate outcomes such as business practice are more associated with

program success than the final outcome is actually reinforced in this full model. Consistent with

the simpler models, youth, the higher education group, and urban beneficiaries seem to benefit

most from programs supporting entrepreneurship, while microcredit clients experience smaller

impacts compared to the general population. In addition, when controlling for all other

characteristics, private sector delivery appears to make a difference in improving programs. Again,

the differences in the likelihood of positive significance across interventions are not significant on

average. This may be because heterogeneous programs are lumped into broad categories. When we

regroup intervention type in a more disaggregated manner and repeat the analysis, financial

training seems to perform poorly compared to other types of programs. 25 We will discuss

heterogeneous program effects in more detail in subsequent sections.


25
     We disaggregate programs as in Table A3 in the Appendix.


                                                                22
       Although not presented in the table, basic characteristics, including region and study

features, are also associated with program success. In fact, experimental results are generally more

robust than quasi-experimental ones. Interestingly, whether the study has been published in a

journal is statistically insignificant in explaining program success, suggesting that little publication

bias is observed in our sample of studies. Programs seem to work better in the longer term: The

period between the completion of the program and the end-line survey is positively related to

finding positively significant impacts. In general, the effectiveness of programs across country

income groups, when controlling for the region, reveals no statistical differences.

       Next, we examine how well the program works by also looking at the magnitude of

estimated program impacts and investigate whether the results from effect size are consistent with

those of significance (See Table 6). We use linear regression models to fit standardized effect

sizes: the OLS, random effects (RE), and study fixed effects (FE) model. The first and second

columns replicate column 5 in Table 5 at the10 and 5 percent significance level, respectively.

Columns 3 through 5 present the results from the OLS, RE model, and study FE model.

       The results across different models are quite consistent. Coefficients from the probit

models seem proportional to those from the linear RE model although significance slightly differs

across specifications (See Figure 4). In terms of FE and RE models, it is useful to recall the

assumptions based on which each model is identified. Given that the study FE model drops

observations that do not vary within studies and relies on within-study variations for identification,

it would be indeed useful in identifying the impacts of design features without compounding

factors if one program has many subcomponents with variations in design features. However, if

design features do not greatly vary within studies, caution is needed in interpreting the FE model

because the results may be drawn from very few estimates. Studies tend to measure the effects of



                                                  23
an intervention on different subpopulations regarding a series of outcomes. Therefore, the types of

outcomes and beneficiaries have sufficient within-study variations for identification, and the

results are encouragingly consistent with other models. However, the type of intervention or

service provider tends to be constant within a program except when the study examines the

impacts of differential intervention type or mode of delivery. For example, the FE result on service

delivered by “government� solely relies on the within-study variation of Bali Swain and Varghesse

(2011), which evaluates different modes of service delivery. With this in mind, we discuss the RE

linear regression model in further detail using the effect size as the dependent variable in the

following sections.



4.2 Delving Deeper into Training and Financing Programs



In order to better understand which program or program component is associated to greater

success, we disaggregate programs into training and financing. Table 7 presents results from an RE

model for effect sizes, restricting the sample to programs with at least one training component. The

three main training components include vocational, business, and financial training. 26 Among

these, business training is most common and often is provided with vocational or financial

training. Counseling services are often added for further guidance and may vary, ranging from job

search or business set-up assistance to business consulting, and to psycho-social support. Of

course, financing support is frequently combined with training and provided as a package.


26
   Different types of training and combinations of interventions are used for different outcomes of interest (See Appendix Table
A3). Vocational training such as in Jovenes programs and Uganda NUSAF are used almost exclusively to improve labor market
activities and incomes unless combined with financing. Business and financial training tend to aim to improve business practices
and knowledge as well as business performance. When business training is combined with counseling, it addresses labor market
activities to set up a new business; however financial training combined with counseling tends to focus more on business practice
and knowledge, and performance.


                                                              24
            The duration of training varies widely. In general, vocational training programs have a

longer span (about 5-6 months) as they cover skills training for certain occupations, while business

training shows the shortest duration. Although duration is an important dimension of

characterizing the training program, it is not always comparable across studies. 27 Some training

lasts for a longer period with shorter sessions (22 weeks of training for microcredit clients meeting

one time per week for one hour, Karlan and Valdiva, 2010); others last similar lengths with greater

intensity (6-8 hours of training at a weekly meeting for six months, Attanasio et al. 2011 and

Almeida and Galasso, 2009); and still others consist of a very short session for a shorter period

(two four-hour meetings per week for six weeks, Calderon et al., 2011).

            The first column of Table 7 includes dummies for the different training components and

their combinations with counseling and financial services; the second and third add the outcomes

of interest and types of beneficiaries, respectively; and the fourth model includes all covariates. As

in the previous specifications, we include study characteristics, region, and country income

classification dummies in all specifications. The combination of business and financial training is

used for the omitted category. First, results consistently show that vocational training has the best

chance of program success, especially when it is complemented with either counseling or

financing services. This is consistent with evidence that “comprehensive� models work better

(Fares and Puerto, 2009). Second, general business training, provided as a stand-alone intervention

or in combination with mentoring/counseling, seems more effective than financial training. 28

However, adding financing support to business training seems to do little.




27
     Not all the studies provide information on the duration of training. For such cases, we contacted authors for data.
28
   This implication is consistent with the finding from Drexler et al. (2011) that a simplified version of training works better than
full technical training.


                                                                    25
         Duration of training program, when considered with the quadratic form, shows that the

relationship between the likelihood of success and duration of training is a flat U-shape. This

suggests that either intensive, short training or substantially extended training would be

appropriate, although the optimal length of training may vary by the outcomes of interest and goal

of the programs. It is worth noting that regardless of specifications, private sector delivery seems

highly correlated with program success. Finally, the effects of training generally tend to fade as the

period between completion of intervention and end-line survey becomes longer.

         Now, we turn to programs with a financing component (See Table 8). We disaggregate

estimates from financing programs into microcredit and grants as stand-alone programs, grants

combined with training, microcredit combined with training, and financing (either microcredit or

grants) combined with counseling. 29 The impacts of financing programs seem less heterogeneous

than those of training intervention. Cash and in-kind transfers combined with training seem to

yield larger impacts than microcredit.                  No statistical differences are observed comparing

microcredit programs with microcredit combined with training or counseling (omitted category).

Like the case of training, private sector delivery has a strong positive association with program

success. The interval between the intervention and the end-line survey is much longer for financing

than training, and unlike training, a longer interval is more associated with higher chances of

success, suggesting that it takes time for the use of a loan or grant to emerge as changed outcomes.



4.3 Regressions on Subsample by Outcome Groups




29
  Given that only four observations from one study combine microcredit with counseling (information session) without training
(De Mel et al., 2011), we merge microcredit and grants when combined with counseling.


                                                             26
Pooling all estimates, though providing sufficient power for identification, does not allow us to

examine the determinants of program success for each particular outcome of interest, if the

determinants differ by the outcomes. A particular intervention may be more frequently used and

relevant for one outcome than the other, and its effectiveness can also vary by the outcome

measures. 30 Moreover, the effects of covariates on the effect size may be specific to the outcome of

interest. For instance, youth may benefit more from interventions in improving labor market

outcomes than changing saving behavior. Examining the sample according to the six outcome

groups separately would reduce heterogeneity across outcome types and provide specific

information on the outcome of interest. With respect to the type of intervention, however, we focus

on broader grouping to accommodate the reduced number of observations.

         The overall findings are as follows (See Table A4-A6 in the Appendix). For nearly all

outcomes, particularly labor market activities and business performance, youth is highly associated

with program success. This is largely driven by youth-targeted programs that present strong

impacts, such as Uganda’s vocational training program (Blattman et al., 2012) and Tunisia’s

business training for college graduates (Premand et al., 2011). To the contrary, women are not

associated with any large and significant impacts other than the outcome of attitudes, indicating

that entrepreneurship programs seem useful for female empowerment but may not be sufficient to

address various barriers faced by women. Existing entrepreneurs also seem to benefit from the

programs in improving their business knowledge and practice, but this is indeed negatively

associated with business performance. This is in line with the finding from a recent study, Karlan



30
   Recall the distribution of intervention type by outcome classification (See Appendix Table A3). Vocational training is commonly
used to improve labor market activities and income but rarely for business outcomes unless combined with financing. Business and
financial training tend to focus more on financial behavior and business outcomes but often are addressed to labor market activities
combined with counseling. While microfinance is widely used to improve all the outcomes considered here, it is more directed to
labor market outcomes when combined with counseling.


                                                                27
et al. (2012), that microenterprise development intervention acted as a catalyzer to adopt good

business practice in the short run but led to little changes in performance.

       When it comes to intervention type, as discussed above, a package providing both training

and financing seems to perform better in promoting labor market activities. However, training

alone can be quite useful to improve business knowledge and practice, and financing alone does

well in enhancing business performance by releasing credit constraints. This suggests that more

customized interventions can enhance cost effectiveness depending on the outcomes of interest and

the constraints. In the next two sections, we investigate the heterogeneous effects of each

intervention by target group for different outcomes, focusing on labor and business outcomes.



4.3.1 Labor Market Outcomes

Table 9 presents the results from random effects models of effect sizes estimated on the labor

market activities and income in Panel A and B, respectively. In each panel, the first row shows the

overall effects of each intervention that is relevant for labor market outcomes: training,

combination of training and financing, and financing. The subsequent rows present the

heterogeneous impacts of each intervention along the types of beneficiaries obtained from

regressions that include interactions between intervention and each variable in the first column.

       As mentioned, a package approach is found to be more effective to improve labor market

activities on average, but the extent varies by the type of beneficiary. This finding is strongest for

youth and social assistance beneficiaries but does not hold among women. For women, the impacts

from microcredit interventions, such as the expansion of access to loans for rural households in

Mongolia (Attanasio et al., 2012), seem higher than training programs. Women are generally more

severely credit constrained, and this in turn can hamper their potential gains from skills training as



                                                  28
observed in Malawi’s apprenticeship training program. 31 Financing support indeed performs better

for women throughout all outcomes. Meanwhile, for business owners, gaining access to finance

does little to increase their activities than receiving business training.

         Panel B presents the results of heterogeneity analysis on labor market income. Unlike labor

market activities, with respect to improving labor earnings and profits, there are no significant

differences across intervention types. Training seems to have greater impacts, especially for the

higher education group in increasing their income, although this was less evident in the case for

labor market activities. Examples of business training interventions that work well for educated

group are in Tunisia and Bosnia and Herzegovina (Premand et al., 2011; Bruhn and Zia, 2011). A

package approach to provide both training and financing seems to be a promising graduation

strategy for social assistance beneficiaries as it increases both labor market activities and income.

         Finally, we now consider both significance and effect size to learn from the characteristics

of successful and unsuccessful programs. As seen in Figure 3, we focus on successful programs

that yield outcomes with positive significance at 10 percent statistical level and effect size greater

than 0.1, and unsuccessful ones with t-statistics and effect size smaller than 0. It is noticed that

vocational training in Colombia, Nicaragua, and Uganda consistently yields more successful

outcomes than unsuccessful ones. All three of these programs target disadvantaged populations to

promote employment through skills development. To the contrary, the outcomes from microcredit

are more heterogeneous. Expansion of access to credit in Bosnia and Herzegovina and the

Philippines presents modest performance with more of their estimates positioning in the successful

category, while experiences in India and Mongolia do not appear to be as strong. Programs

targeting social assistance beneficiaries from Argentina’s Jefes and Chile’s Solidario contribute

31
   See Cho et al. (2012) that find a large gender disparity in participation, experiences, and outcomes from an apprenticeship
training program.


                                                             29
more to unsuccessful outcomes than successful ones, suggesting that helping the poor exit from

social assistance programs by promoting self-employment is a challenging task.



4.3.2 Business and Behavioral Outcomes

We now move onto business outcomes such as business knowledge and practice and business

performance, which are important intermediate outcomes that may lead to successful business and

increased income (See Table 9). 32 We again look into training, financing, and the combination of

the two. 33 Panel A suggests that business training alone can be quite effective in improving

business knowledge and practice. This is particularly true for those who already have their own

businesses. For women, however, training does little but financing matters in changing their

business practice, suggesting that women may not be the decision maker in business practice

unless they have capital under their control.

            With respect to business performance (Panel B), financing seems to be the most relevant

and effective intervention. Recall that business performance is captured by measures such as sales,

inventories, number of paid employees, and business expenses. Thus access to credit probably

plays a greater role in improving business performance than training. That being said, training

alone is strongly associated with business performance of youth and higher education individuals,

especially in the studies where business training was provided for these entrepreneurs (Bruhn and

Zia, 2011 for business training targeted to young entrepreneurs in Bosnia and Herzegovina;

Drexler et al., 2011 for business training for microenterprise owners in Dominican Republic),




32
     Full results of regression are presented in Table A5 in the Appendix.
33
   Unlike the above case of labor outcomes, when it comes to business outcomes, most of training observations are
from business training intervention.


                                                                   30
suggesting that a more efficient use of resources can achieve outcomes as effectively when

targeted to these groups of entrepreneurs.

       In the same manner as above, we consider which programs yield successful and

unsuccessful outcomes. As discussed above, many training programs result in improved business

knowledge and practice but are unsuccessful when it comes to business performance. Such

examples include business training with the lottery of winning larger amount of loans provided to

microcredit clients in Pakistan (Gine and Mansuri, 2011) and training for microenterprise owners

in the Dominican Republic (Drexler et al. 2011).



5.     Conclusion and Discussion


Entrepreneurship programs will continue to constitute an important policy tool in the developing

world as long as self-employment exists as a critical alternative for rationed wage employment. A

fundamental question is then which interventions and combinations of programs are more effective

in enabling the poor to start up and grow their own business. As this will depend on the types of

skills (business, technical, “soft skills�) and capital (cash, in-kind, credit) constraints, which vary

by individuals, we began this study by asking “which type of intervention is more effective for

whom and for which outcomes?� In order to answer this, we collected information on program

effects from rigorously evaluated studies around developing countries and compiled a large and

rich data set with program details. We examined the impact of interventions promoting

entrepreneurial activities on a variety of outcomes, such as labor market activities and income, as

well as on business practice and performance. We also examined attitudes and financial behavior

outcomes. Given the specificity of each program, we considered the design and implementation



                                                  31
features of each program, the context and policy environment of each country, and finally, the

study characteristics potentially affecting the estimates of outcomes.

       Our meta analysis suggests a number of important implications. Combinations of different

intervention types matter for different beneficiaries under different contexts. With respect to

training programs, it seems that vocational and business training work better than financial

training, and can be further improved by combining financing support or counseling. Business

training for entrepreneurs, in particular, appears to be a relatively cost-effective way of promoting

business performance and growth with a short intervention period, although improved knowledge

and practice through training does not always materialize as increased income. In terms of

financing, there are little variations in effectiveness of cash, in-kind grants, and microcredit.

       Investigating the effects of programs separately by outcome group suggests that programs

need to be customized for each outcome of interest addressing the specific constraints relevant for

the outcome. We find that a package promoting skills with financing support seems to have larger

impacts on labor market activity. However, training alone can be quite useful to improve business

knowledge and practice, and financing alone does well in enhancing business performance by

releasing credit constraints. For women, the largest effects come from providing access to credit,

suggesting that access to credit may have been the largest constraint to women in their earning

opportunities. Overall, involving the private sector for the delivery of programs and evaluating the

program in the longer term appear to be more closely associated with improved effects of

programs.

       Our results have important policy implications. First, programs promoting self-employment

opportunities and small-scale entrepreneurship can lead to increases in labor market outcomes with

important welfare gains. Second, providing relevant combinations of skills, capital, and counseling



                                                   32
support based on the target group’s main constraints is important to achieve better results. Third,

among widely heterogeneous effects, it is noteworthy that the impacts on both labor market and

business outcomes are significantly higher for youth. This is especially relevant in many parts of

the developing world that are facing the ‘youth bulge’ and aspiring to provide meaningful

opportunities to their young populations.




                                                33
References

Abraham R., F. Kast, and D. Pomeranz (2011). “Insurance through Savings Accounts Evidence

       from a Randomized Field Experiment among Low-Income Micro-Entrepreneurs in Chile,�

       mimeo. Harvard University.

Almeida, R. and E. Galasso (2009). “Jump-starting Self-employment? Evidence for Welfare

       Participants in Argentina,� World Development 38(5).

Attanasio, O., B. Augsburg, R. de Haas, E. Fitzsimons, and H. Harmgart (2012). "Group Lending

       or Individual Lending: Evidence from a Randomized Field Experiment in Mongolia,"

       mimeo. London: European Bank for Reconstruction and Development.

Attanasio, O., A. Kugler, and C. Meghir (2011). "Subsidizing Vocational Training for

       Disadvantaged Youth in Developing Countries: Evidence from a Randomized Trial,"

       American Economic Journal: Applied Economics 3:188-220.

Augsburg, B., R. de Haas, H. Harmgart, and C. Meghir (2012). "Microfinance at the Margin:

       Experimental Evidence from Bosnia and Herzegovina," mimeo. London: European Bank

       for Reconstruction and Development.

Ayyagari, M., A. Demirgüç-Kunt and V. Maksimovic (2011). “Young vs. Small Firms across the

       World: Contribution to Employment, Job Creation, and Growth.� World Bank Policy

       Research Working Paper No. 5631. Washington DC: World Bank.

Bali Swain, R. and A. Varghese (2011). Delivery “Mechanisms and Impact of Microfinance

       Training in Indian Self Help Group,� Journal of International Development.

Bandiera, O., R. Burgess, S. Gulesci, I. Rasul, and M. Sulaiman (2012a). “Can Entrepreneurship

       Programs Transform the Economic Lives of the Poor?� mimeo. London: London School of

       Economics.
Bandiera, O., R. Burgess, S. Gulesci, I. Rasul, M. Goldstein, and M. Sulaiman (2012b).

       “Participating in Adolescent Training Programs: New Evidence from Uganda,� mimeo.

       London: London School of Economics.

Banerjee, A., E. Duflo, R. Glennerster, and C. Kinnan (2009). “The Miracle of Microfiance? An

       Evidence from Randomized Evaluation,� mimeo. Cambridge, Mass.: Massachusetts

       Institute of Technology.

Banerjee, A. and E. Duflo (2008). "Do Firms Want to Borrow More? Testing Credit Constraints

       Using a Directed Lending Program," mimeo. Cambridge, Mass.: Massachusetts Institute of

       Technology.

Banerjee, A. and A. Newman (1993). “Occupational Choice and the Process of Development,�

       Journal of Political Economy, 101(2):274-98.

Berge, L., I. Oppedal, K. Bjorvatn, and B. Tungodden (2011). “Human and Financial Capital for

       Microenterprise Development: Evidence from a Field and Lab Experiment," mimeo.

       Continuous Mortality InvestigationWorking Paper. Bergen: The Actuaries.

Bjorvatn, K. and B. Tungodden (2010). “Teaching Business in Tanzania: Evaluating Participation

       and Performance,� Journal of the European Economic Association 8 (2-3).

Blattman, C., N. Fiala, and S. Martinez (2012). "Can Employment Programs Reduce Poverty and

       Social Stability? Experimental Evidence from Uganda Aid Program," mimeo. Washington

       DC: World Bank.

Bloom, N., B. Eifert, A. Mahajan, D. McKenzie, and J. Roberts (2012). “Does Management

       Matter? Evidence from India,� presented at the NBER Summer Institute, Cambridge,

       Massachusetts.
Bruhn, M. and B. Zia (2011). “Stimulating Managerial Capital in Emerging Markets: The Impact

       of Business and Financial Literacy for Young Entrepreneurs,� World Bank Policy Research

       Working Paper No. 5642. Washington DC.

Brune, L., X. Gine, J. Goldberg, and D. Yang (YEAR). “Commitments to Save: A Field

       Experiment in Rural Malawi.� World Bank Policy Research Working Paper No. 5748.

       Washington DC.

Calderon, G., J. Cunha, and G. de Giorgi, (2011). “Business Literacy and Development: Evidence

       from a Randomized Trial in Rural Mexico,� mimeo. Palo Alto, Calif.: Stanford University.

Card, D., P. Ibarraran, F. Regalia, D. Rosas-Shady, and Y. Soares (2011). “The Labor Market

       Impacts of Youth Training in the Dominican Republic,� Journal of Labor Economics

       29(2).

Card, D., J. Kluve, and A. Weber (2010). “Active Labour Market Policy Evaluations: A Meta-

       Analysis,� Economic Journal 120:F452-F477.

Card, D. and A. Kruger (1995). “Time Series Minimum-Wage Studies: A Meta-analysis,�

       American Economic Review 85(2):238-43.

Carneiro, P., E. Galasso, and R. Ginja (2009). “The Impact of Providing Psycho-Social Support to

       Indigenous Families and Increasing their Access to Social Services: Evaluating Chile

       Solidario, Mimeo. Washington DC: World Bank.

Casey, C., R. Glennerster, and E. Miguel (2011). Reshaping Institutions: Evidence on External Aid

       and Local Collective Action. NBER Working Paper No. 17012. Cambridge, Mass.:

       National Bureau of Economic Research.
Cho, Y., D. Kolomba, M. Mobarak, and V. Orzco (2012). “Differential Effects of Vocational

       Training on Men and Women and the Bias from Program Drop-outs,� mimeo. Washington

       DC: World Bank.

Ciavarella, M., A. Buchholtz, C. Riordan, R. Gatewood, and G. Stokes (2004). “The Big Five and

       Venture Survival: Is there a linkage?� Journal of Business Venturing 19:465-83.

Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ:
       Erlbaum.

Cole, S., and G. Kartini Shastry (2009). “Smart Money: The Effects of Education, Cognitive

       Ability and Financial Literacy on Financial Market Participation.� Harvard Business

       School Working Paper No. 09-071. Cambridge, Mass.: Harvard University.

Cole, S., A. Sampson, and B. Zia. 2010. Prices or Knowledge? What Drives Demand for Financial

       Services in Emerging Markets? Harvard Business School Finance Working Paper Series,

       October. Cambridge, Mass.: Harvard University.

Cooper, H. and L. Hedges (1994). The Handbook of Research Synthesis. New York: Russell Sage

       Foundation.

Crant, J. (1996). “The Proactive Personality Scale as a Predictor of Entrepreneurial Intentions,�

       Journal of Small Business Management 34(3):42-49.

Crepon, B., F. Devoto, E. Duflo, and W. Pariente (2011). “Impact of Microcredit in Rural Areas of

       Morocco: Evidence from a Randomized Evaluation,� mimeo. Paris: Abdul Latif Jameel

       Poverty Action Lab (J-PAL) Europe.

De Mel, S., D. McKenzie, and C. Woodruff (2012). Business Training and Female Enterprise

       Start-up, Growth, and Dynamics: Experimental Evidence from Sri Lanka, mimeo.

       Washington DC: World Bank.
De Mel, S., D. McKenzie, and C. Woodruff (2010). “Wage Subsidies for Microenterprises,�

       American Economic Review: Papers and Proceedings 100 (2):614-18.

De Mel, S., D. McKenzie, and C. Woodruff (2008a). “Are Women More Credit Constrained?

       Experimental Evidence on Gender and Microenterprise Returns,� American Economic

       Journal: Applied Economics 1(3):1-32.

De Mel, S., D. McKenzie, and C. Woodruff (2008b). "Who are the Microenterprise Owners:

       Evidence from Sri Lanka on Tokman v. de Soto." Bureau for Research in Economic

       Analysis of Development (BREAD) Working Paper No. 174. Durham, North Carolina:

       Duke University.

De Mel, S., D. McKenzie, and C. Woodruff (2008c). “Returns to Capital: Results from a

       Randomized Experiment,� Quarterly Journal of Economics 123(4):1329-72.

Drexler, A., G. Fischer, and A. Schoar (2011). “Keeping it Simple: Financial Literacy and Rules of

       Thumb,� mimeo. Cambridge, Mass.: Abdul Latif Jameel Poverty Action Lab (J-PAL) at

       MIT.

Fafchamps, M., D. McKenzie, S. Quinn, and C. Woodruff (2011). “When Is Capital Enough to Get

       Female Microenterprises Growing? Evidence from a Randomized Experiment in Ghana,�

       NBER Working Paper No. 17207. Cambridge, Mass.: National Bureau of Economic

       Research.

Fares, J. and O. Puerto (2009). “Toward Comprehensive Training,� Social Protection Discussion

       Paper No. 0924. Washington DC: World Bank.

Field, E., S. Jayachandran, and R. Pande (2010a). “Do Traditional Institutions Constrain Female

       Entrepreneurship? A Field Experiment on Business Training in India,� American Economic

       Review: Papers & Proceedings 100(2):125-29.
Field, E., R. Pande, J. Papp, and N. Rigol (2010b). “Term Structure of Debt and Entrepreneurial

       Behavior: Experimental Evidence from Microfinance,� mimeo.

Gine, X. and G. Mansuri (2011). "Money or Ideas? A Field Experiment on Constraints to

       Entrepreneurship in Rural Pakistan", mimeo. Washington DC: World Bank.

Gine, X. and D. Karlan (2009). “Group versus Individual Liability: Long Term Evidence from

       Philippine Microcredit Lending Group,� mimeo. New Haven, Conn.: Yale University.

Gine, X. and D. Yang (2009). “Insurance, Credit, and Technology Adoption: Field Experimental

       Evidence from Malawi,� Journal of Development Economics 89:1-11.

Gindling, T. H. and D.L. Newhouse (2012). “Self-Employment in the Developing World� World

       Bank Policy Research Working Paper No. 6201.

Haltiwanger, J., R. Jarmin, and J. Miranda (2010). “Who Creates Jobs? Small vs. Large vs.

       Young,� NBER Working Paper No. 16300. Cambridge, Mass.: National Bureau of

       Economic Research.

Jarrell, S. and T. Stanley (1990). “A Meta-Analysis of the Union-Nonunion Wage Gap,� Industrial

       and Labor Relations Review 44(1):54-67.

Karlan, D., R. Knight, and C. Udry (2012). “Hoping to Win, Expected to Lose: Theory and

       Lessons on Micro Enterprise Development�, mimeo. New Haven, Conn.: Yale University.

Karlan, D. and M. Valdivia (2011). “Teaching Entrepreneurship: Impact of Business Training on

       Microfinance Clients and Institutions,� The Review of Economics and Statistics 93(2):510-

       527.

Karlan, D. and J. Zinman (2011). "Microcredit in Theory and Practice: Using Randomized Credit

       Scoring for Impact Evaluation," Science 332(6035): 1278-1284.
Karlan, D. and J. Zinman (2010). “Expanding Credit Access: Using Randomized Supply Decisions

      to Estimate the Impacts,� Review of Financial Studies 23(1).

Klinger, B. and M. Schundlen (2007). "Can Entrepreneurial Activity Be Taught? Quasi

      Experimental Evidence from Central America," mimeo. Cambridge, Mass.: Harvard

      University.

Kluve, J. (2010). "The Effectiveness of European Active Labor Market Programs," Labour

      Economics 17:904-18.

McKenzie, D. and C. Woodruff (2013). "What Are We Learning from Business Training and

      Entrepreneurship Evaluations around the Developing World," mimeo. Coventry, England:

      University of Warwick.

Macours, K., P. Premand, and R. Vakis (2012). "Transfers, Diversification and Household Risk

      Strategies: Experimental Evidence with Lessons for Climate Change Adaptation," Policy

      Research Working Paper Series No. 6053. Washington DC: World Bank.

Mano, Y., A. Iddrisu, Y. Yoshino, and T. Sonobe (2011). "How Can Micro and Small Enterprises

      in Sub-Saharan Africa Become More Productive? The Impacts of Experimental Basic

      Managerial Training." Washington DC: World Bank.

Ooseterbeek, H., M. van Praag, and A. Ijsselstein (2010). “The Impact of Entrepreneurship

      Education on Entrepreneurship Skills and Motivation,� European Economic Review 54 (3).

Pitt, M., S. Khandker, and J. Cartwright (2003). “Does Micro-Credit Empower Women? Evidence

      from Bangladesh� Policy Research Working Paper Series No. 2998. Washington DC:

      World Bank.

Premand, P., S. Brodmann, R. Almeida, R. Grun, and M. Barouni (2011). “Entrepreneurship

      Training and Self Employment among University Graduates: Evidence from a Randomized
      Trial in Tunisia,� Policy Research Working Paper Series No. 6285. Washington DC: World

      Bank.

Stanley, T. (2001). "Wheat from Chaff: Meta-Analysis as Quantitative Literature Review,"

      Journal of Economic Perspectives 15(3):131-150.

Stanley, T. and S. Jarrell (1998). "Wage Discrimination Bias? A Meta-Regression Analysis,"

      Journal of Human Resources 33(4):947-73.
                Table 1. Distribution and Definitions of the Outcomes of Interest
Outcomes of interest                    Definitions                                                   Frequency
1. Labor market activities                                                                            242 [21.7%]
    Business set up and expansion       Binary indicator of business setup or exansion
    Employment (self employment)        Binary indicator of employment status or employment rates
    Hours of work                       Hours of work in labor market or business
    Unemployment                        Binary indicator of unemployment status
    Business closing                    Binary indicator of business closing

2. Labor market income                                                                                309 [27.7%]
    Household income                    Household income from various sources
    Household assets                    Household assets (durable and nondurable)
    Profits (from household business)   Profits from business
    Earnings                            Salary and payment for labor
    Consumptions                        Household expenditure/consumption on durable/non-
                                        durable goods

3. Financial behavior                                                                                 126 [11.3%]
    Having a loan (formal, informal)    Binary indicator of having a (formal/informal) loan
    Having an insurance or savings      Binary indicator of having insurance scheme or saving
    Amount of loan/saving changed       Amount of loan/savings

4. Business knowledge and practice                                                                    155 [13.9%]
    Business knowledge                  General business knowledge including the abilities to
                                        calculate profits, manage stocks, and make investment
   Innovation                           Binary indicator of adopting new technology or developing
                                        new product
   Access to network                    Binary indicator of having an access to the network of
                                        individuals with businesses or market
   Accounting practice                  Binary indicator of book and record keeping, and separation
                                        of individual and business accounts.

5. Business performance                                                                               184 [16.5%]
    Business expenses                   Amount of business expenses (inventory, salary, etc)
    Sales from the business             Size of the sales
    Number of employees                 The number of (paid/unpaid) employees
    Capital and investment              Size of stocks, investment, and inventories

6. Attitudes                                                                                      100 [9.0%]
   Attitude torward business            Attitude toward entrepreneurial activities and traits
   Confidence and optimism              Changes in confidence or positivity toward labor market
                                        prospect and future
   Risk and time preference             Risk taking attitude and discount rate of future benefits
   Decision making and reservation Changes in decision making process and reservation wage
   wage
Notes: The frequency indicates the number of observations of each outcome category. The proportions of each
category are specified in the brackets.
             Table 2. Distribution and Definitions of the Interventions of Interest
Outcomes of Interventions             Definitions                                                    Frequency
1. Training                                                                                          458 [41.0%]
    Vocational training               (In-class or apprenticeship) training on various professions
    Lifeskills training               Usually in-class training for problem solving and critical
                                      thinking
   Business training                  General knowledge on business management including
                                      customer relations, inventory and financial management,
                                      and marketing
   Financial training                 Specific knowledge on accounting and inter-temporal
                                      decision making on investment

2. Financing                                                                                         742 [66.5%]
    Cash grant                        Cash transfer for business
    In-kind grant                     Transfer in the form of tools, goods, and equipment
    Microcredit                       Loan for business for future repayment
    Savings                           Access to saving arrangement

3. Counseling                                                                                          238 [21.3%]
    Mentoring in business              Follow up advice in the process of business operation
    Arrangements for on-the-job        Guidance provided on-the-job
    advice
Notes: Counseling is added to either training or financing, it is never a stand-alone intervention. In many cases,
training and financing are provided in combination. The "frequency" column specifies the number and proportion
of observations that estimate the effect of each intervention. Due to the combinations of interventions, the
proportions do not sum up to 100%.
                                Table 3. Sample Characteristics by Region
                                    All sample      AFR          EAP         ECA          LAC        MENA          SAR
Total number of estimates             1,116         185          119         180          318         96           218
Type of intervention (% )
Training only                         22.3%        21.6%        6.7%        21.1%        37.7%          -         19.7%
Training+Counseling                   11.2%           -           -            -         20.8%        61.5%          -
Financing only                        50.1%        58.4%        93.3%       78.9%         9.1%        38.5%       60.6%
Financing+Counseling                   8.9%           -           -            -         29.9%          -          1.8%
Training+Financing                     6.3%        16.8%          -            -            -           -         17.9%


Outcome Groups (% )
LM Activity                           21.7%        10.3%        20.2%       36.7%        29.2%        31.3%        4.6%
LM Income                             27.7%        22.7%        35.3%       21.7%        31.8%        19.8%       30.3%
Financial Behavior                    11.5%        20.5%        29.4%       10.0%         2.8%        4.2%        11.0%
Business Knowledge and Practice       13.9%        10.8%          -         11.1%        16.4%        16.7%       21.6%
Business Performance                  16.5%        20.0%        10.1%       16.7%        16.7%        11.5%       18.8%
Attitudes                              9.0%        15.7%        6.7%         3.9%         3.1%        16.7%       13.8%


Population Groups (% )
Female                                26.3%        27.0%        74.8%        6.7%        15.7%        11.5%       37.2%
Youth                                 14.6%        11.4%          -         27.8%        10.4%        61.5%          -
High Education                        16.8%         8.6%        21.8%       20.0%        12.9%        61.5%        4.1%
SA beneficiaries                      14.3%        11.4%          -            -         43.7%          -            -
Entrepreneurs                         34.6%        50.3%        26.1%       26.7%        45.3%        8.3%        28.4%
Microcredit clients                   27.3%        24.3%        3.4%        21.1%        34.9%          -         49.1%
Urban                                 43.9%        76.8%        17.6%       21.1%        52.5%        61.5%       28.9%


Providers (% )
Government agencies                   21.4%        11.9%           -           -         49.1%        61.5%        0.9%
NGOs and community organizers         35.2%        15.7%        79.0%          -         56.9%           -        40.8%
Universities/researchers              20.7%        38.9%        17.6%          -          2.2%        61.5%       33.0%
MFIs or banks                         56.5%        49.7%        100.0%      100.0%       34.9%         38.5%       41.7%
Notes: Type of interventions and outcomes include mutually exclusive categories while the type of beneficiaries and
providers of the programs are not defined in a mutually exclusive way. The categories of beneficiaries can overalp and the
programs are often delivered by multiple agencies.
                          Table 4. Summary of Estimated Impacts by Outcome

                                                     Financial   Business     Business
Outcomes                     LM Activity LM Income                                         Attitudes   Total
                                                     Behavior     Practice   Performance

Significance at 10%
negative                        2.5%       3.2%        9.5%        5.2%         4.3%         3.0%      4.2%
insignificant                  68.2%       72.8%      67.5%       54.8%        71.2%        63.0%      67.6%
positive                       29.3%       23.9%      23.0%       40.0%        24.5%        34.0%      28.2%
Significance at 5%
negative                        1.7%       2.3%        7.9%        1.9%         2.7%         2.0%      2.8%
insignificant                  76.0%       78.0%      74.6%       66.5%        80.4%        73.0%      75.5%
positive                       22.3%       19.7%      17.5%       31.6%        16.8%        25.0%      21.7%


Effect Size
Overall average                 0.065      0.036       0.034       0.106        0.044        0.090     0.058
Average among positvely
                                0.181      0.136       0.204       0.254        0.154        0.180     0.183
signicant at 10%
Average among positvely
                                0.192      0.147       0.224       0.283        0.173        0.200     0.200
signicant at 5%
            Table 5. Probit Model Regressions for Positively Significant Impacts

                                       Dependent Variable=Indicator of Positively Significant Impact at 10%
                                           (1)          (2)            (3)            (4)                  (5)
                                        ME/SE          ME/SE          ME/SE          ME/SE             ME/SE
LM activities                           0.354**                                                          0.228
                                         (0.171)                                                        (0.173)
Business practice                       0.467**                                                       0.756***
                                         (0.185)                                                        (0.217)
Business performance                     -0.024                                                           0.143
                                         (0.169)                                                        (0.192)
Financial behavior                        0.009                                                           0.302
                                         (0.165)                                                        (0.185)
Attitudes                                 0.111                                                           0.283
                                         (0.160)                                                        (0.197)
Training only                                          -0.042                                            0.343
                                                       (0.298)                                          (0.367)
Training+counseling                                     0.336                                            -0.419
                                                       (0.451)                                          (0.437)
Financing only                                         -0.094                                            0.182
                                                       (0.283)                                          (0.400)
Financing+counseling                                   -0.357                                            0.329
                                                       (0.366)                                          (0.348)
Female                                                                 -0.102                            -0.176
                                                                      (0.199)                           (0.239)
Youth                                                                0.567***                         0.676***
                                                                      (0.126)                           (0.202)
High education                                                          0.177                          0.270**
                                                                      (0.115)                           (0.130)
Microenterprise owners                                                 -0.044                            -0.123
                                                                      (0.106)                           (0.155)
Social assistance beneficiaries                                         0.210                             0.253
                                                                      (0.222)                           (0.384)
Microfinance clients                                                -0.507***                        -1.161***
                                                                      (0.136)                           (0.306)
Urban                                                                 0.294**                           0.392*
                                                                      (0.119)                           (0.221)
Government only                                                                      -0.256             -0.092
                                                                                     (0.359)            (0.224)
NGOs only                                                                             0.057             -0.158
                                                                                     (0.300)            (0.279)
Universities only                                                                    -0.439             -0.146
                                                                                     (0.286)            (0.269)
MFI or banks only                                                                   -0.425**           -0.372*
                                                                                     (0.213)            (0.225)
Private sector delivery                                                              -0.054             0.467*
                                                                                     (0.241)            (0.283)
Number of observations                   1,097          1,097          1,097          1,097              1,097

Adjusted R2                              0.099          0.091          0.124          0.097            0.155

note: *** p<0.01, ** p<0.05, * p<0.1
Marginal effects (ME) are reported. Standard errors (SE) are clustered by study id and reported in parenthesis.
Regressions are weighted by the inverse of the number of estimates per study in the database.
All specifications include study characteristics (period between the completion of intervention and estimation,
journal publication, experimental design, and the square root of number of observations), region, and income
dummies which are not reported here. Column (1) specifies only with outcomes of interest, (2) with the types
of intervention, (3) with the type of beneficiaries, (4) with the type of delivering agencieis, and (5) with all of
above. Omitte categories include labor income (type of outcomes), Training+Financing (type of intervention),
and multiple providers (program providers).
                                  Table 6. Comparisons of Different Models
                                                                                                  Random          Fixed
                                                  Probit (10% ) Probit (5% )         OLS
                                                                                                  Effects        Effects
                                                       (1)             (2)            (3)            (4)            (5)
                                                      me/se          me/se         coef/se        coef/se        coef/se
LM activities                                         0.228           0.190         -0.007          0.004         -0.013
                                                     (0.173)         (0.156)        (0.021)        (0.020)       (0.022)
Business practice                                   0.756***         0.556*        0.074**        0.082**        0.070*
                                                     (0.217)         (0.284)        (0.037)        (0.040)       (0.039)
Business performance                                  0.143          -0.012          0.002          0.012         0.009
                                                     (0.192)         (0.222)        (0.019)        (0.022)       (0.019)
Financial behavior                                    0.302           0.146         -0.007         -0.002         0.017
                                                     (0.185)         (0.200)        (0.019)        (0.020)       (0.024)
Attitude and traits                                   0.283           0.073          0.020          0.035         0.032
                                                     (0.197)         (0.228)        (0.022)        (0.024)       (0.023)

Training only                                          0.343          0.523          0.033          0.029        -0.004
                                                      (0.367)        (0.380)        (0.028)        (0.022)       (0.021)
Training+counseling                                   -0.419         -0.095          0.053          0.016       -0.045**
                                                      (0.437)        (0.445)        (0.043)        (0.033)       (0.021)
Financing only                                         0.182          0.474          0.025          0.037         0.028
                                                      (0.400)        (0.453)        (0.033)        (0.026)       (0.023)
Financing+counseling                                   0.329         -0.007          0.003          0.021           -
                                                      (0.348)        (0.390)        (0.034)        (0.036)
Female                                                 -0.176        -0.251          0.002          -0.001        0.003
                                                       (0.239)       (0.279)        (0.019)         (0.017)      (0.031)
Youth                                                0.676***      0.721***        0.051**        0.070***      0.060**
                                                       (0.202)       (0.234)        (0.022)         (0.017)      (0.026)
High education                                        0.270**         0.218       0.039***         0.028**      0.025**
                                                       (0.130)       (0.235)        (0.013)         (0.012)      (0.011)
Microenterprise owners                                 -0.123        -0.090         -0.000          -0.007       -0.002
                                                       (0.155)       (0.149)        (0.014)         (0.012)      (0.011)
Social assistance beneficiaries                         0.253         0.514          0.022          -0.047          -
                                                       (0.384)       (0.474)        (0.043)         (0.042)
Urban                                               -1.161***      -0.707**       -0.041**         -0.042*       -0.004
                                                       (0.306)       (0.328)        (0.021)         (0.022)      (0.006)
Microfinance clients                                   0.392*       0.598**         -0.019          -0.023          -
                                                       (0.221)       (0.234)        (0.022)         (0.019)

Government only                                       -0.092       -1.047***        -0.019         -0.000      -0.096***
                                                     (0.224)         (0.208)        (0.030)       (0.029)        (0.009)
NGOs only                                             -0.158          0.184          -0.016       -0.034*      -0.046***
                                                     (0.279)         (0.316)        (0.021)        (0.019)       (0.000)
Universities only                                     -0.146          -0.191         -0.006         -0.049          -
                                                     (0.269)         (0.279)        (0.035)        (0.033)
MFI or banks only                                    -0.372*          -0.117         -0.004         -0.023     -0.155***
                                                      (0.225)        (0.212)        (0.033)        (0.029)       (0.019)
Private sector delivery                               0.467*          -0.106          0.017         -0.005       0.018*
                                                      (0.283)        (0.309)        (0.028)        (0.027)       (0.009)
Number of observations                                1,097          1,097           1,097          1,097         1,097
Adjusted R2                                          0.155            0.166          0.155          0.175        0.029
note: *** p<0.01, ** p<0.05, * p<0.1
Dependent variable in columns 1(2) is a dummy for positive significance of estimates at 10% (5%). Dependent variable in
column 3-5 is the estimated effect size.
Standard errors are clustered by study id and reported in parenthesis.
All specifications include study characteristics (period between the completion of intervention and estimation, journal
publication, experimental design, and the square root of number of observations), region and income dummies, which are
not reported here.
        Table 7. Random Effect Models for Effect Sizes of Estimates: Training

                                                        (1)          (2)           (3)            (4)
                                                      coef/se      coef/se      coef/se         coe/se
Business training only                               0.077**       0.061**      0.028***       0.025***
                                                      (0.031)      (0.025)       (0.008)        (0.008)
Business training+Counseling                         0.214**      0.214***      0.228***       0.236***
                                                      (0.096)      (0.080)       (0.076)        (0.075)
Business training+Financing                           0.073*        0.035        -0.010          -0.027
                                                      (0.040)      (0.048)       (0.025)        (0.035)
Vocational training only                              0.217*       0.198**      0.254***       0.244***
                                                      (0.124)      (0.098)       (0.073)        (0.059)
Vocational training+Counseling                       0.202***     0.251***      0.509***       0.517***
                                                      (0.057)      (0.062)       (0.141)        (0.120)
Vocational training+Financing                          0.125        0.125*      0.501***       0.438***
                                                      (0.077)      (0.071)       (0.076)        (0.086)
Financial training only                               -0.003        -0.012     -0.051***       -0.053***
                                                      (0.035)      (0.029)       (0.009)        (0.008)
Financial training+Counseling                         -0.033        -0.038     -0.051***       -0.050***
                                                      (0.033)      (0.029)       (0.016)        (0.017)
Duration of training                                   0.007        0.007      -0.036***       -0.027***
                                                      (0.013)      (0.011)       (0.007)        (0.008)
Duration squared                                      -0.001*      -0.001*      0.000***         0.000
                                                      (0.000)      (0.000)       (0.000)        (0.000)
Private sector delivery                              0.150**       0.129**      0.081***       0.077***
                                                      (0.071)      (0.054)       (0.019)        (0.019)
Months since completion/100                            0.204        0.014        -0.089          -0.211
                                                      (0.221)      (0.383)       (0.298)        (0.390)
Outcomes of interest                                    No           Yes           No             Yes
Type of beneficiaries                                   No           No           Yes             Yes
Number of observations                                  439          439           439            439
note: *** p<0.01, ** p<0.05, * p<0.1
Standard errors are clustered by study id and reported in parenthesis. Dependent variable is the effect size.
The omitted category of training is "Business + financial training combined".
All specifications include dummies for study characteristics (journal publication, experimental design, and
the square root of study sample), region and country income dummies which are not reported here.
 Table 8: Random Effect Models for Effect Sizes of Estimates: Financing

                                                    (1)         (2)         (3)          (4)
                                                  coef/se     coef/se     coef/se     coef/se
Grant only                                         0.015       -0.009      0.014       -0.010
                                                  (0.027)     (0.030)      (0.023)     (0.028)
Grant + Training                                   0.068       0.053       0.063*      0.053
                                                  (0.055)     (0.052)      (0.036)     (0.034)
Microcredit only                                   -0.001      -0.015      0.013       0.001
                                                  (0.021)     (0.020)      (0.016)     (0.016)
Microcredit + Training                             -0.004      -0.017      -0.003      -0.020
                                                  (0.022)     (0.021)      (0.019)     (0.019)
Private sector delivery                          0.058***    0.053***    0.071***     0.064***
                                                  (0.017)     (0.020)      (0.017)     (0.019)
Months since completion/100                      0.227***    0.214***    0.170***     0.148***
                                                  (0.039)     (0.040)      (0.033)     (0.041)
Outcomes of interest                                No          Yes         No          Yes
Type of beneficiaries                               No          No          Yes         Yes
Number of observations                              734         734         734         734
note: *** p<0.01, ** p<0.05, * p<0.1
Standard errors are clustered by study id and reported in parenthesis. Dependent variable is the
effect size. The omitted category of financing is "Microcredit/grants +Counseling ".
All specifications include dummies for study characteristics (journal publication, experimental
design, and the square root of study sample), region and country income dummies which are
not reported here.
Table 9. Random Effect Regression Model for Labor Market Outcomes

                                                Type of Beneficiaries Interacted with
 A. LM Activity                         Training Only    Training+Financing Financing only
 1. Overall                               -0.050***           0.184***      Omitted Variable
                                            (0.002)             (0.025)
 2. Type of beneficiaries
      Female                                -0.018            -0.106***            0.023**
                                           (0.048)              (0.000)             (0.011)
      Youth                                 0.006             0.208***            0.066***
                                           (0.004)              (0.024)             (0.014)
      High Education                        0.028                  -                -0.002
                                           (0.080)                                  (0.007)
      SA beneficiaries                      -0.043            0.169***              -0.013
                                           (0.041)             (0.060)              (0.046)
      Entrepreneurs                         -0.002                 -              -0.016***
                                           (0.034)                                  (0.004)
      Microcredit clients                 -0.074**                -                 -0.001
                                           (0.033)                                  (0.085)
      Urban                                 -0.014                -               0.010***
                                           (0.062)                                  (0.004)
      Num. of observation
                                                 Type of Beneficiaries Interacted with
 B. LM Income                           Training Only Training+Financing Financing only
 1. Overall                                 -0.031               -0.021         Omitted Variable
                                           (0.031)              (0.018)
 2. Type of beneficiaries
      Female                                 -0.007           -0.114***              -0.019
                                            (0.038)             (0.021)             (0.032)
      Youth                                  0.047             0.108**              -0.009
                                            (0.043)             (0.050)             (0.012)
      High Education                      0.269***                 -               0.029**
                                            (0.061)                                 (0.013)
      SA beneficiaries                       -0.073           0.078***               0.027
                                            (0.047)            (0.028)              (0.046)
      Entrepreneurs                          -0.002             -0.049               -0.018
                                            (0.074)            (0.049)              (0.024)
      Microcredit clients                    0.038              -0.052               0.010
                                            (0.040)            (0.040)              (0.035)
      Urban                                0.086**              -0.016               0.031
                                            (0.035)            (0.028)              (0.021)
      Num. of observation                                         309
 note: *** p<0.01, ** p<0.05, * p<0.1
 Standard errors are clustered by study id.
 All specifications include study characteristics (period between the completion of intervention
 and estimation, journal publication, experimental design, and the square root of number of
 observations), region and income dummies, which are not reported here.
Table 10. Random Effect Regression Model for Business Outcomes
                                                       Interacted with
 A. Business practice              Training only     Training+Financing      Financing only
 1. Overall                           0.075*                 0.012              Omitted
                                      (0.041)               (0.012)             category
 2. Type of beneficiaries
      Female                          -0.112**            -0.257***             0.335***
                                       (0.057)              (0.000)              (0.000)
      Youth                             0.467                0.104                  -
                                       (0.298)              (0.084)
      High Education                    0.011                  -                    -
                                       (0.019)
      SA beneficiaries                    -                 -0.363                  -
                                                            (0.226)
      Entrepreneurs                   0.084**              -1.069**                 -
                                       (0.039)              (0.525)
      Microcredit clients               0.037                0.400              -0.739**
                                       (0.093)              (0.318)              (0.288)
      Urban                            1.037*                -0.115                 -
                                       (0.550)              (0.110)
      Num. of observation                                     155
                                                       Interacted with
 B. Business performance           Training only     Training+Financing      Financing only
 1. Overall                          -0.075***            -0.142***             Omitted
                                       (0.024)              (0.032)             category
 2. Type of beneficiaries
      Female                           -0.001             -0.140***               -0.042
                                      (0.022)               (0.000)              (0.049)
      Youth                          0.155***                0.044                   -
                                      (0.027)               (0.037)
      High Education                 0.049***                  -                 -0.046*
                                      (0.005)                                    (0.026)
      SA beneficiaries                    -                -0.111**                  -
                                                            (0.056)
      Entrepreneurs                    0.081*                0.024               -0.031*
                                       (0.047)              (0.064)              (0.017)
      Microcredit clients             -0.079**              0.100*                0.013
                                       (0.032)              (0.055)              (0.055)
      Urban                            -0.058                0.007             -0.092***
                                       (0.037)              (0.036)              (0.027)
      Num. of observation                                     184
 note: *** p<0.01, ** p<0.05, * p<0.1
 Standard errors are clustered by study id.
 All specifications include study characteristics (period between the completion of
 intervention and estimation, journal publication, experimental design, and the square root of
 number of observations), region and income dummies, which are not reported here.
Figure 1. Distribution of Estimates: Region and Income
Figure 2. Proportion of Estimates by Intervention Type
                                   Labor Activities                                                           Labor Income




                                                                               12
    12




                                                                                   8
        8




                                                                           t-value
t-value




                                                                            4
 4




                                                                               0
    0




                                                                               -4
    -4




            -.6   -.4        -.2       0         .2   .4      .6      .8               -.6   -.4        -.2     0         .2   .4      .6      .8
                                       effect size                                                              effect size


                               Business Practice                                                        Business Performance




                                                                               12
    12




                                                                                   8
        8




                                                                           t-value
t-value




                                                                            4
 4




                                                                               0
    0




                                                                               -4
    -4




            -.6   -.4        -.2       0         .2   .4      .6      .8               -.6   -.4        -.2     0         .2   .4      .6      .8
                                       effect size                                                              effect size


                               Financial Behavior                                                              Attitudes
                                                                               12
    12




                                                                                   8
        8




                                                                           t-value
t-value




                                                                             4
  4




                                                                               0
    0




                                                                               -4
    -4




            -.6   -.4        -.2       0         .2   .4      .6      .8               -.6   -.4        -.2     0         .2   .4      .6      .8
                                       effect size                                                              effect size

                  training              financing          combined                          training            financing          combined




  Figure 3. Significance and Effect Size by Types of Intervention and Outcome.
Coefficients from Probit Model of Effect Significance
-1.5       -1       -.5       0         .5         1




                                                          -.15             -.1          -.05          0             .05      .1
                                                                     Coefficients from Random Effects Model of Effect Size
                                                        coef = 5.5857511, se = .92974551, t = 6.01




                                                        Figure 4. Comparison of Probit and Random Effects Models.
Appendix
Table A1. Studies used for meta analysis and main results
Study id             Group     country     Income   region   Year   Interval Main intervention                    Outcomes
                                                                    (months)
Almeida and          SA        Argentina Upper      LAC      2005   13       in-kind grants and technical         LM activities: -
Galasso. (2009)      beneficia           middle                              assistance for graduation of Jefes   LM income: 0
                     ries                income                              program
Attanasio et al.     Youth     Colombia Upper       LAC      2006   14       3 month vocational training, 3       Female LM
(2011)                                   middle                              months on the job training           outcomes: +
                                         income                                                                   Male LM outcomes:
                                                                                                                  0
Banerjee and         SMEs      India       Lower    South    2002   24        Policy change that increases an     Profit, sales:0
Duflo. (2008)                              middle   Asia                      influx of credit on SME's
                                           income
Banerjee et al.      Women India           Lower    South    2008   22        Expansion of microfinance           Business setup,
(2010)                                     middle   Asia                      institute                           profits: +
                                           income
Bali Swain and       Microfina India       Lower    South    2003   36        Comparing different delivery        Income: 0
Varghese. (2011)     nce                   middle   Asia                      modes of training for               Assets: +
                     clients               income                             microfinance clients: Model 1       NGO linkage model
                                                                              (bank-formed and financed           +
                                                                              group), Model 2(NGO formed but
                                                                              bank financed), Model 3(NGO
                                                                              formed and financed).
Berge et al. (2011) Microfina Tanzania     Low      Africa   2009   7         Impacts of business training and    Knowledge:+
                    nce                    income                             grants (cash) among PRIDE           Sales:+
                    clients                                                   (microfinance) clients.             Profits: + (male)
Bjorvatn and        Microfina Tanzania     Low      Africa   2009   7         Business training among             Knowledge:+
Tungodden. (2010) nce                      income                             microcredit clients,
                    clients                                                   microenterprise owners.
Blattman et al.     Youth     Uganda       Low      Africa   2011   13        Vocational training and             LM activities :+
(2011)                                     income                             tools/materials for self            Profits: +
                                                                              emploment.
Brune et al. (2011) Small      Malawi      Low      Africa   2010   17        Access to bank account: ordinary    Profit: 0
                    holder                 income                             vs. commitment group                Sales, income: +
                    farmers                                                                                       (only for
                                                                                                                  commitment group)

Bruhn et al. (2011) Young    Bosnia and Upper       ECA      2010   7         Business and financial training     Profit, sales, new
                    entrepre Herzegovi middle                                                                     business practice: 0
                    neurs    na         income

Carneiro et al.      SA        Chile       Upper    LAC      2007   48        Chile Solidario, providing psycho- LM activities,
(2009)               beneficia             middle                             social support as well as transfers. income: 0
                     ries                  income                                                                  Optimism: +
Calderon et al.      Female Mexico         Upper    LAC      2010   10        Basic business training              Profits, revenues,
(2011)               entrepre              middle                                                                  number of clients
                     neurs                 income                                                                  served: +
Card et al. (2011)   Youth     Dominican   Upper    LAC      2005   11        Vocational training+internship       LM activities: 0
                               Republic    middle                                                                  Earnings when
                                           income                                                                  working: +
Table A1. (Continued)
Study id             Group      country     Income   region   Year   Interval Main intervention                     Outcomes
                                                                     (months)
Cole et al. (2010)   Unbanke Indonesia Lower         EAP      2010   26       Financial literacy training,          Bank account:+
                     d                 middle                                 subsidies (small, medium, high)       savings: 0
                     househol          income                                 contingent upon bank account          (higher effects of
                     ds                                                                                             incentives than
                                                                                                                    training)
Crepon et al.        Rural      Morocco     Lower    MENA     2009   22       Expansion of MFI (and access to       Business setup: 0
(2011)               househol               middle                            credit)                               Income, sales: +
                     ds                     income
De Mel et al.        Microent   Sri Lanka   Lower    South    2008   39       Cash/in-kind transfer                 Profits: + (male)
(2008a)              erprise                middle   Asia
                     owners                 income
De Mel et al.        Microent   Sri Lanka   Lower    South    2007   26       Small or large cash, or in-kind       Profits, capital
(2008b)              erprise                middle   Asia                     equipment                             stock: + (male)
                     owners                 income
De Mel et al. (2011) Microent   Sri Lanka   Lower    South    2010   73       Provision of information on loans,    Take up of loan: +,
                     erprise                middle   Asia                     prodedures of getting the loan        Profits: 0
                     owners                 income                            (applications, requirements of
                                                                              guarantors,etc).
Drexler et al.       Microent Dominican Upper        LAC      2008   13       Compare the impacts of full           Business practice : +
(2011)               erprise Republic middle                                  version of financial training,        (simplified
                     owners             income                                simplified version of training, and   version), income: 0
                                                                              on-site visits for counseling.
Dupas and            Market Kenya           Low      Africa   2009   42       Access to non-interest bearing        saving, investment,
Robinson. (2009)     vendors                income                            bank account. Significant effects     expenditure: +
                     (women)                                                  on women (vendors) but no             revenue, hours
                     and                                                      impacts on men (bike-taxi drivers)    worked: 0
                     bicycle-
                     taxi
                     drivers
                     (men)
Fafchamps et al.     Microent Ghana         Lower    Africa   2010   13       Cash grant or in-kind subsidies       Profits: +
(2011)               erprise                middle                                                                  cash<in-kind
                     owners                 income                                                                  (especially women)

Field et al. (2010a) Women      India       Lower    South    2007   4        Financial and business skills         Income, loan: +
                                            middle   Asia                     training                              Business plan:0
                                            income                                                                  Results varying by
                                                                                                                    social status
Field et al. (2010b) Women      India       Lower    South    2010   37       Changes in term structure from        Profit, business
                                            middle   Asia                     short to longer term                  setup, income: +
                                            income                            repayment.There is a significantly
                                                                              positive impacts of giving grace
                                                                              period for repayment on those
                                                                              business owners.

Gine and Yang        Farmers Malawi         Low      Africa   2006   1        Provision of credit for technology Take up of loan:+
(2009)                                      income                            adoption with or without weather (only without
                                                                              insurance.                         insurance purchase)

Gine and Karlan      Microfina Philippine Lower      EAP      2006   19       Changes from group to individual Default: 0
(2010)               nce       s          middle                              liability of repayment.
                     clients              income
Table A1. (Continued)
Study id           Group     country      Income   region   Year   Interval Main intervention                   Outcomes
                                                                   (months)
Gine and Mansuri   Microfina Pakistan     Lower    South    2008   18       Business training+lottery of    Knowledge,
(2011)             nce                    middle   Asia                     winning larger amount of loans. business practice,
                   clients                income                                                            psychological
                                                                                                            wellbeing:+
                                                                                                            Profits: 0
Karlan and Zinman Marginall South         Upper    Africa   2006   27       Expansion of access to consumer LM activities,
(2010)              y         Africa      middle                            credits                         income,
                    rejected              income                                                            consumption,
                    loan                                                                                    wellbeing:+
                    applicant
                    s
Karlan and Zinman Marginall Philippine    Lower    EAP      2007   13.51233 Expansion of access to consumer     Number of business
(2011)              y         s           middle                            credits                             activities or
                    rejected              income                                                                employees: -
                    loan
                    applicant
                    s
Karlan and Valdivia Female Peru           Upper    LAC      2005   20       Business training added to          knowledge:+
(2011)              microfina             middle                            microcredit                         profits, revenue,
                    nce                   income                                                                employment: 0
                    clients
Klinger and         (Potentia G&N         Lower    LAC      2005            Business training and monetary      Business
Schündeln (2007) l)                       middle                            prize                               expansion: +
                    Business              income
                    owners
Mano et al. (2011) Microent Ghana         Lower    Africa   2008   13       Business training                   Profit, practice,
                    erprise               middle                                                                revenue: +
                    owners                income
Macours et al.      SA        Nicaragua   Lower    LAC      2009   32       Vocational training, grants         LM activities,
(2011)              beneficia             middle                                                                income, profits: +
                    ries                  income                                                                (grants)
McKenzie and        Small     Mexico      Upper    LAC      2006   3        Cash or in-kind transfers on retail Profits: +
Woodruff (2008)     retail                middle                            firms.                              (financially
                    firms                 income                                                                constrained)
Pitt et al. (2006)  Microfina Banglades   Low      South    1999   96       Microcredit                         Women's
                    nce       h           income   Asia                                                         empowerment: +
                    clients
Premand et al.      Youth     TunisiaUpper         MENA     2011   3        Entrepreneurship education          Knowledge, self
(2011)                               middle                                 (proposal) for college graduates    employment: +
                                     income                                                                     LM income: 0
Attanasio et al.   Rural    Mongolia Lower         EAP      2009   2        Microcredit (group vs. individual   Profit: + (group
(2012)             househol          middle                                 lending vs. no lending)             lending)
                   ds                income                                                                     Business setup: +
                                                                                                                LM income: 0
Augsburg et al.    Marginall Bosnia and Upper      ECA      2010   15       Expansion of access to loan         LM activities: +
(2012)             y         Herzegovi middle                                                                   Consumption: 0
                   rejected na          income                                                                  saving:-
                   loan
                   applicant
                   s
Table A2: Summary of effect size by program characteristics and estimates significance

                                                                     Insignificant or     Significantly
Effect Size                                              Overall
                                                                     negative at 10%     positive at 10%
Proportion                                               100%            71.80%              28.20%
Average                                                  0.058            0.009               0.183

Intervention Types
    Training only                                        0.057             0.002              0.212
    Training+Counseling                                  0.095             0.019              0.232
    Financing only                                       0.052             0.007              0.164
    Financing+Counseling                                 0.011            -0.002              0.103
    Training+Financing                                   0.104             0.052              0.181
Outcomes of interest
    LM Activity                                          0.065             0.017              0.181
    LM Income                                            0.036             0.005              0.136
    Financial Behavior                                   0.034            -0.017              0.204
    Business Practice                                    0.106             0.007              0.254
    Business Performance                                 0.044             0.009              0.154
    Attitudes                                            0.090             0.044              0.180
Beneficiaries
    Female                                               0.048            0.011               0.158
    Youth                                                0.102            0.010               0.226
    High Education                                       0.082            0.016               0.223
    SA beneficiaries                                     0.046            0.001               0.150
    Entrepreneurs                                        0.063            0.012               0.201
    Urban                                                0.036            0.013               0.104
Providers
    Government only                                      0.036            0.013               0.104
    NGO only                                             0.081            0.031               0.164
    University only                                      0.067            0.011               0.161
    MFI and banks only                                   0.055            0.008               0.195
    Mutilple providers                                   0.058            0.001               0.213
Regions
    AFR                                                  0.122            0.033               0.258
    EAP                                                  0.003            -0.009              0.138
    ECA                                                  0.039            0.005               0.198
    LAC                                                  0.053            0.017               0.168
    MENA                                                 0.084            -0.003              0.206
    SAR                                                  0.046            -0.002              0.121
Income level
    Low income                                           0.130            0.019               0.273
    Lower middle income                                  0.043            0.005               0.127
    Upper middle income                                  0.057            0.010               0.210
                  Table A3. Distribution of Type of Intervention by Outcomes of Interest
                                  Disaggregated                                   Business     Business     Financial
 Type of intervention                                   LM Activity   LM Income                                         Attitudes
                                   intervention                                    Practice   Performance   Behavior
                         (1) vocational                        6         16           .            .            .           .
     Training only       (2) business+financial                1         13          60           44           30
                         (3) business                           .        11          33           17           6           12
                         (4) vocational+counseling             18        17           .            .            .           .
 Training+counseling     (5) business+counseling               38         3          16            .            .          13
                         (6) financial+counseling               .                    10           10            .           .
                         (7) vocational+financing              9         18           3            3            .           .
  Training+financing
                         (8) business+financing                8         11          16            9           3            4
                         (9) grant (cash, in-kind)             10        36           .           12            .           .
    Financing only
                         (10) microcredit                      109       116         17           81           83          61
Financing+counseling (11) financing+counseling                 43        68           .            8           4           10
                             Total                             242       309        155           184         126          100
Note: the number of observations are specified in the table.
  Table A4. Random Effect Regression Model : Labor Outcomes
                                               LM Activity           LM Income
                                                     (1)                  (2)
                                                  coef/se             coef/se
Training                                        -0.050***              -0.031
                                                   (0.002)             (0.031)
Training+Financing                               0.184***               -0.021
                                                   (0.025)             (0.018)
Female                                             -0.009              -0.021
                                                   (0.027)             (0.024)
Youth                                            0.060***               0.028
                                                   (0.015)             (0.023)
High education                                     -0.002              0.049*
                                                   (0.007)             (0.025)
Microenterprise owners                          -0.015***              -0.016
                                                   (0.003)             (0.024)
Social assistance beneficiaries                    -0.027               0.017
                                                   (0.017)             (0.037)
Microfinance clients                             -0.054**               0.019
                                                   (0.022)             (0.034)
Urban                                               0.008             0.040**
                                                   (0.005)             (0.020)
Private sector delivery                             0.018               0.031
                                                   (0.015)             (0.024)
Months since completion/100                       0.109**             0.175**
                                                   (0.054)             (0.085)
Number of observations                               223                 309
Adjusted R2                                         0.373               0.179
note: *** p<0.01, ** p<0.05, * p<0.1
Standard errors are clustered by study id and reported in parenthesis. Dependent
variable is the estimated effect size.
The omitted category is Financing only. All specifications include study
characteristics (journal publication, experimental design, and the square root of
number of observations), region and income dummies, which are not reported here.
  Table A5. Random Effect Regression Model: Business Outcomes
                                      Business Practice         Business Performance
                                               (1)                         (2)
                                           coef/se                      coef/se
Training                                    0.075*                    -0.075***
                                            (0.041)                     (0.024)
Training+Financing                            0.012                   -0.142***
                                            (0.012)                     (0.032)
Female                                       -0.064                      -0.032
                                            (0.076)                     (0.028)
Youth                                         0.467                    0.155***
                                            (0.298)                     (0.027)
High education                                0.011                       0.014
                                            (0.019)                     (0.033)
Microenterprise owners                     0.078**                     -0.037**
                                            (0.037)                     (0.017)
Social assistance beneficiaries              -0.363                    -0.111**
                                            (0.226)                     (0.056)
Microfinance clients                          0.037                      -0.034
                                            (0.093)                     (0.037)
Urban                                        -0.044                   -0.077***
                                            (0.116)                     (0.023)
Private sector delivery                    -0.313*                        0.021
                                            (0.189)                     (0.031)
Months since completion/100                  -0.509                    0.347***
                                            (0.353)                     (0.134)
Number of observations                         155                         184
Adjusted R2                                   0.328                       0.445
note: *** p<0.01, ** p<0.05, * p<0.1
Standard errors are clustered by study id and reported in parenthesis. Dependent
variable is the estimated effect size.
The omitted category is training combined with financing. All specifications include
study characteristics (journal publication, experimental design, and the square root of
number of observations), region and income dummies, which are not reported here.
 Table A6. Random Effect Regression Model: Behavioral Outcomes

                                       Financial Behavior               Attitudes
                                                (1)                         (2)
                                             coef/se                     coef/se
Financing (Microcredits)                      -0.019                      -0.003
                                             (0.039)                      (0.035)
Female                                         0.017                     0.133**
                                             (0.029)                      (0.064)
Youth                                       0.237***                      -0.010
                                             (0.074)                      (0.037)
High education                             -0.109***                      0.305*
                                             (0.013)                      (0.171)
Microenterprise owners                         0.045                       0.044
                                             (0.065)                      (0.030)
Social assistance beneficiaries            (dropped)                    (dropped)

Microfinance clients                          -0.018                      0.028
                                              (0.048)                    (0.049)
Urban                                          0.026                     0.016*
                                              (0.068)                    (0.010)
Private sector delivery                        0.036                     0.129*
                                              (0.036)                    (0.072)
Months since completion/100                    0.322                      0.115
                                              (0.292)                    (0.322)
Number of observations                          126                        100
Adjusted R2                                    0.287                      0.458
note: *** p<0.01, ** p<0.05, * p<0.1
Standard errors are clustered by study id and reported in parenthesis. Dependent
variable is the estimated effect size.
The omitted category is business training. All specifications include study
characteristics (journal publication, experimental design, and the square root of number
of observations), region and income dummies, which are not reported here.