WPS6019


Policy Research Working Paper                   6019




                 When Do Donors Trust
                Recipient Country Systems?
                                Stephen Knack




The World Bank
Development Research Group
Human Development and Public Services Team
April 2012
Policy Research Working Paper 6019


  Abstract
  The 2005 Paris Declaration on Aid Effectiveness sets                              Tests show that a donor’s use of the recipient country’s
  targets for increased use by donors of recipient country                          systems is positively related to: (1) the donor’s share of
  systems for managing aid. The target is premised on a                             aid provided to the recipient (a proxy for the donor’s
  view that country systems are strengthened when donors                            reputational stake in the country’s development); (2)
  trust recipients to manage aid funds, but undermined                              perceptions of corruption in the recipient country (a
  when donors manage aid through their own separate                                 proxy for the trustworthiness or quality of the country’s
  parallel systems. This study provides an analytical                               systems); and (3) public support for aid in the donor
  framework for understanding donors’ decisions to trust                            country (a proxy for the donor’s risk tolerance). Findings
  or bypass country systems. Empirical tests are conducted                          are robust to corrections for potential sample selection,
  using data from three OECD-DAC surveys designed                                   omitted variables or endogeneity bias.
  to monitor progress toward Paris Declaration goals.




  This paper is a product of the Human Development and Public Services Team, Development Research Group. 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 author may be contacted at sknack@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
                             When Do Donors Trust
                           Recipient Country Systems?




                                       Stephen Knack*




JEL: F35, H61, O10, O19

Keywords: Aid effectiveness, Paris Declaration, corruption, use of country systems

*Lead Economist, DECHD, World Bank (Sector Board affiliation: PREM Public Sector
Governance; e-mail: sknack@worldbank.org). This research was supported by the World
Bank‟s Knowledge for Change Program (KCP). The conclusions of this paper are not intended
to represent the views of the World Bank, its Executive Directors, or the countries they represent.
Valuable comments, suggestions or other assistance were provided by Nicholas Eubank,
Yoichiro Ishihara, Daniel Lederman, Jed Friedman, John Giles, Colin Xu, Francesco Grigoli, and
two anonymous referees. The author assumes full responsibility for any remaining errors.
1. Introduction

       The literature on growth and development now recognizes a key role for the quality of

public institutions. Yet, little is known about how to apply this insight constructively in the

design of development assistance. “Good governance� in the form of capable and accountable

state institutions emerged only over the course of centuries in the West, and it may be impossible

for today‟s developing countries to follow similar paths in a dramatically compressed time

frame. Historically unprecedented progress on health outcomes and education enrollment may

contribute to unrealistic expectations regarding donors‟ role in accelerating institutional

development in poor countries. Even in relatively technocratic areas (e.g. civil service and

legal/judicial reform) of public sector governance, success rates for donor projects are low and

the evidence base underpinning the design of interventions is thin or absent (IEG, 2008).

       Moreover, there is growing recognition that international aid – and particular donor

practices - can unintentionally weaken governance in recipient countries. Aid – like natural

resource revenues - can encourage rent seeking and undermine government‟s accountability to its

own citizens, by reducing its dependence on domestic taxpayers for revenues (Moore, 1998;

Collier, 2006; Knack, 2009). Donors can also undermine government capacity, when they

fragment their aid among too many projects, sectors and countries, and insist on delivering aid

using their own reporting, procurement and other procedures (Brautigam and Knack, 2004). As

the World Bank‟s Assessing Aid report acknowledges: “At times, donors have hindered the

creation of effective public sectors because they saw end runs around local institutions as the

easiest way to achieve project success� (World Bank, 1998: 84).

       These arguments have attained the status of conventional wisdom within the international

aid community. The importance of delivering aid in ways consistent with long-run institutional


                                                 2
strengthening is a major theme in the Paris Declaration agenda for reforming aid practices. The

Paris Declaration (PD) created a set of numerical indicators on improved “alignment� of aid

activities with country systems and inter-donor coordination, to be monitored through surveys of

donors and recipient countries, with defined targets for the year 2010.

         Although the PD principles and associated indicators reflect a broad consensus within the

donor community (most notably in the OECD Development Assistance Committee), the

empirical basis for this new aid effectiveness agenda is thin (Easterly, 2007). Advocacy for

reform of donor practices is based on theory, intuition and scattered anecdotal evidence.1 This

paper does not attempt to add to the thin evidence base underpinning the Paris Declaration

principles and indicators. Rather, it is premised on the fact that they have been endorsed by the

DAC bilateral donors and multilateral agencies providing the bulk of ODA, and by most

recipient countries. Donor behavior is assessed relative to donors‟ own assertions about what

constitutes more effective aid practices.2

         Knack and Eubank (2009) provide a theoretical framework for understanding the

incentive problems among donors that can produce suboptimal levels of harmonization and

alignment in their aid activities. This framework and supporting empirical evidence are relevant

so long as most of the donor community believes that use of country systems is currently below

the optimal level. The caveat is that we cannot offer confident conclusions about which donors



1
  See World Bank (2003: ch. 11) for representative anecdotes. Azam et al. (1999) construct a formal model with
learning-by-doing externalities in tax administration, in which aid dependence with low institutional capacity can be
an equilibrium outcome. What little systematic empirical evidence exists in support of the Paris Declaration agenda
bears mostly on its harmonization provisions, not the alignment provisions that cover (inter alia) use of country
systems. Anderson (2011), Djankov et al. (2009), Knack and Rahman, 2007) and Knack and Smets (2012) all
provide empirical evidence on the adverse effects of fragmenting a country‟s aid among a larger number of donors.
2
  Here we follow Easterly (2007) and Easterly and Pfutze (2008: 3-4): “The academic aid policy literature and the
aid agencies themselves agree on many elements of „best practice‟ . . . By taking this consensus as our standard, we
are asking in effect if aid agencies operate the way they themselves say they should operate.�



                                                          3
are really providing “better� aid, in the absence of more evidence on how using country systems

affects either the quality of systems or development outcomes.

       The current study builds on the preliminary empirical tests reported in Knack and Eubank

(2008), using a much-expanded dataset. Unlike the earlier study, it addresses potential sample

selection and endogeneity problems in detail. There is a time dimension to the dataset used in

this study, allowing us to rule out omitted variable or simultaneity bias in certain key results.

While the earlier study corrected standard errors for clustering only by donors or by recipients,

this update reports more conservative standard errors, corrected for two-way clustering for

donors and recipients. In addressing endogeneity and other potential problems more thoroughly,

we have chosen to report results only for one dependent variable: use of country public financial

management (PFM) systems. Results in general hold for the additional dependent variables in

Knack and Eubank (2008), namely use of direct budget support and use of program-based

approaches (which include but are not limited to direct budget support). Finally, in the attempt

to minimize sources of endogeneity and other bias, the current study measures quality of country

systems using widely-known governance indicators that mostly reflect views of experts outside

aid agencies; Knack and Eubank (2009) relied heavily on indicators reflecting the judgments of

World Bank staff.

       Before turning to the empirical analysis, the remainder of this section presents the

theoretical arguments from Knack and Eubank (2009) in condensed form. To preview, a donor‟s

trust in a recipient‟s aid management systems is determined in our framework by three broad

considerations:

    1. Confidence it will reap sufficient benefits from investing in recipient country systems;

    2. Trustworthiness of those systems, as measured for example by corruption ratings;




                                                  4
    3. Trust in aid‟s effectiveness in general, in donor countries.

       The Paris Declaration calls for increased use of recipient country systems in managing

aid, but it explicitly acknowledges that the quality of country systems is a key determinant of

their use by donors. Recipients, with technical assistance from donors, are therefore urged to

strengthen their public financial management (PFM) systems. In the meantime, using those

systems, despite their flaws, is believed to strengthen them in the medium or long run: “Donors

can help build capacity and trust by using country systems to the fullest extent possible, while

accepting and managing the risks involved…� (OECD, 2009a: 27). On the other hand,

bypassing country systems undermines them, by diffusing accountability and fragmenting policy

and planning processes (OECD, 2009b; Mokoro Ltd. 2008a). Moreover, donors often staff their

own parallel aid management systems by “poaching� the most talented government officials.

       Donors‟ incentives to trust country systems (or alternatively to micro-manage aid using

their own parallel systems) depend in part on their perceived trustworthiness. Where recipient

aid management systems are stronger, corruption scandals tarnishing the donor agency‟s

reputation are less likely to occur, and aid-funded programs are more likely to be selected and

implemented more efficiently. The Paris Declaration recognizes that weaknesses in country

systems sometimes justify donors‟ decisions to bypass them. The developmentally-optimal level

of trust by donors in country systems varies positively with the quality of those systems. In

certain cases, trust may even be inefficiently high (Jansen, 2009).

       For a given level of trustworthiness of country systems, donors‟ trust in them can vary

substantially because they have different mandates and face varying degrees of political pressure

from their taxpayers and elected overseers. Multilateral donors can differ from bilateral donors,

and some bilateral donors may be constrained more than others by limited support for foreign aid




                                                 5
on the part of politicians and the public. Pressures to demonstrate visible achievements (that can

be attributed plausibly to a donor agency‟s own aid funds) to skeptical taxpayers or elected

officials will increase further the tendency to free ride on other donors‟ contributions to

strengthening country systems (Williamson and Agha, 2008: 34). When bilateral donors use aid

to advance diplomatic or commercial objectives, incentives to rely on their own parallel systems

for aid delivery will be further aggravated. For example, using their own procurement rules will

likely advantage donor-country contractors.

       If using country systems for managing aid does strengthen them, then donors‟ use of

country systems (in the absence of coordinated action) is likely to be sub-optimal. The benefits

of using country systems are mostly external (benefiting other donors) and realized only over the

long term, while costs are short term and fully internalized by the donor. When donor agency i

undertakes any action to strengthen recipient country systems, it is in effect providing a public

good for other donors. Stronger aid management systems reduce reputational and fiduciary risks

and increase the developmental impact of aid funds not only for donor i ‟s future aid, but also for

other donors. Meanwhile, donor i incurs the full costs, in exposing its current aid funds to

higher risks than if it bypassed recipient country systems. The benefits are not wholly external,

however: when a donor has a larger share of the aid “market� in a country, it will internalize

more of the benefits from any investments in strengthening country systems. Donor i ‟s use of

country systems is then expected to be positively related to its share of total aid that is received

by a given country, controlling for quality of recipient systems.

       The next section describes how the model‟s concepts can be operationalized and its

predictions tested using a panel dataset based on the Paris Declaration‟s three monitoring

surveys, with donor-recipient pairs as the main unit of analysis. Empirical results reported in




                                                  6
section 3 are largely consistent with the predictions of the theoretical framework. A variety of

robustness tests in this section address potential concerns regarding sample selection, omitted

variables and endogeneity bias. The final section summarizes the findings and discusses their

implications.



2. Data and Empirical Strategy

                                    Measuring trust in country systems

                               Ri
        We operationalize         , the share of recipient-managed aid, using data from the OECD
                               Ai

DAC‟s Paris Declaration Monitoring Survey (PDMS). The PDMS was designed to assist in

measuring progress toward the Paris Declaration‟s objectives between 2005 and 2010. Among

the 12 PD indicators established with goals set for 2010, 9 of them are measured using data from

the PDMS and 3 from other sources. Most of these indicators are beyond the scope of the

present study. We measure trust in country systems with PD Indicator 5a: use of country public

financial management (PFM) systems as a percentage of aid for the government sector.3 This

indicator is constructed as a simple average of three sub-indicators from the survey: use of

national (i) budget execution procedures, (ii) financial reporting procedures, and (iii) auditing

procedures, each as a percentage of aid for the government sector. Detailed criteria for these

three dimensions of use-of-PFM-systems are provided in Appendix 1. Correlations among these

sub-indicators average .73 (ranging from .54 to .77). Findings presented below change very little

if any one of its three components is analyzed instead of Indicator 5a. Mean use of PFM systems



3
  Aid for the government sector accounts for about 83% of total aid disbursements reported in the PDMS, and
excludes aid disbursed to NGOs, parastatals or private companies unless it is provided “in the context of an
agreement with officials authorized to act on behalf of central government� (OECD, 2008a). Empirical results
reported below are unaffected if government sector aid is replaced with total aid in the denominator.


                                                        7
among the 2,213 observations appearing in one or more tests reported below is 33%. (Appendix

Table A1 provides summary statistics for all variables in the analysis.)

         The PDMS data constitute an unbalanced three-dimensional panel: surveys were

conducted in 2006, 2008 and 2011, and designed to obtain information on aid flows for the

previous calendar year (2005, 2007 and 2010 respectively) for each donor-recipient pair. Each

successive survey round improved the quality and comprehensiveness of data. Guidance on

definitions was strengthened and comparability of data reported by donors and by recipients was

enhanced in later rounds (OECD, 2011; OECD 2008). The number of recipient countries

choosing to participate in the survey increased from 34 in 2006, to 54 in 2008, and 78 in 2011.

These 78 recipients account for about three-fourths of total aid worldwide (OECD, 2011).

         Coverage of aid within these recipients is comprehensive over donors. Most of the

“emerging� bilateral donors and a few of the smaller multilateral donors and vertical funds are

dropped from the analysis, due to missing data on other donor variables from the DAC database

or public opinion surveys. Dropped donors collectively account for a total of only 1.4% of aid

reported in the survey.4 The 34 donors remaining include the 22 traditional DAC donors, Korea

(a new DAC member in 2011), Turkey, 7 multilateral organizations, and 3 vertical funds.5

         The panel is unbalanced not only with respect to time, but also with respect to donor-

recipient pairs: not all donors provide aid to all recipients in each survey year. These features of

the data create potential selection problems, considered in more detail below.



4
  Their share of all aid worldwide is undoubtedly greater, although much of what is popularly termed “aid� from
China and some other new donors would not meet the DAC‟s definition of official development assistance (ODA).
5
  Bilateral donors are listed in Appendix Table A2. Multilaterals in the sample include the African, Asian and Inter-
American Development Banks, the World Bank, IMF, EU and UN. Vertical funds include the GAVI Alliance,
IFAD, and the Global Fund to Fight AIDS, Tuberculosis, and Malaria. The sample includes all donors with
sufficient data for listing in the appendix of the OECD (2011) survey report, and all donors covered by Birdsall and
Kharas (2010) in their rankings of donor performance. (The OECD tables exclude Greece, while Birdsall and
Kharas exclude Turkey, the IMF and the GAVI Alliance.)


                                                          8
                                            Independent variables

         Donor‟s ability to benefit from investments in country systems ( Ai in the model) is

proxied by donor aid share, defined as the percentage of total aid (inclusive of aid not to the

government sector) to recipient j accounted for by donor i . Higher donor aid shares are

predicted to increase use of country systems. The sample mean for donor aid share is 7%, with a

minimum value of 0.0014%6 (Switzerland in Egypt in 2010) and a maximum of 83.6% (World

Bank in St. Vincent in 2010).7

        Trustworthiness of recipient country systems ( ï?§ in the model) is measured primarily by

the widely used Control of Corruption indicator, from the Worldwide Governance Indicators

project.8 Weaker country systems, as measured by lower scores on Control of Corruption reflect

increased risks to donors of corruption-related scandals, diversion of funds to lower priority uses,

or inability to account for how funds were spent. The Paris Declaration explicitly asserts that

corruption “inhibits donors from relying on partner country systems.� Use of country PFM

systems is thus expected to be positively related to corruption scores.

        The Freedom House civil liberties indicator is used as a supplementary indicator of

trustworthiness of country systems. Civil liberties scores are higher in countries with a free and

independent media, freedom of assembly, freedom for NGOs, and greater personal autonomy

and individual freedoms.9 The use of donor and other public funds is likely to be more


6
  Values of 0 are not present in the sample because the dependent variables are all undefined for donor-recipient
pairs with no aid transfers.
7
  Larger donors at the global level do not always have higher average aid shares at the recipient level, as some
donors concentrate their aid in fewer countries. Portugal‟s average aid share (23.5%) exceeds that of the U.S.
(12.4%), although the U.S. provided more than 100 times as much total aid to countries in the sample as Portugal.
8
  This indicator is a composite of numerous corruption-related indicators collected from various sources including
surveys and risk assessment firms. It is similar in many respects to Transparency International‟s even more widely
known Corruption Perceptions Index, but includes more data sources and covers more countries.
9
  An index scored from 1 to 7 is available going back to 1972, but we use a more finely-grained version scored from
0-100 that is available for 2003 onward. The political freedoms indicator produced by Freedom House includes


                                                         9
transparent in a more open society. Moreover, the principle of “country ownership� may have a

stronger appeal to donors operating in countries with stronger mechanisms for public voice and

government accountability to broader constituencies. In short, donors will tend to trust country

systems more when aiding recipients that look more like most donor countries, i.e. those that are

relatively open and democratic. Where there is reason to think local elites are not committed to

development objectives and will divert aid to maintain their political and economic power

(Bjornskov, 2010; Angeles and Neandis, 2009), donors are less likely to provide budget support

and to trust country PFM and procurement systems. Civil Liberties – and to some degree

Control of Corruption – can measure the potential for elite capture of aid flows.

         Other recipient characteristics included as control variables are log of per capita income,

recent growth in per capita income (average annual increase in the two preceding years), log of

population, and log of per capita aid. These variables can capture the effects of real or perceived

quality of country systems, separate from what is measured by Control of Corruption and Civil

Liberties.10 Donors may sometimes infer that wealthier or more rapidly growing countries have

stronger capacity for managing aid, in the absence of more direct evidence. If there are

economies of scale in aid management systems, their quality may tend to be higher in larger

countries. For a given level of expected quality, uncertainty about quality is likely higher in

smaller countries, where donors have obtained relevant information from fewer projects and

have undertaken fewer diagnostic and other analytic studies. If donors tend to be risk averse, this

higher uncertainty can reduce use of country systems.




freedom from government corruption among its criteria. Political freedoms and civil liberties are correlated at .95,
and we prefer the latter to the former to avoid conceptual overlap with Control of Corruption.
10
   Anecdotal evidence from country evaluations suggests that donors sometimes bypass country systems because
they are “slower and more cumbersome� as well as vulnerable to corruption (Wood et al., 2011).


                                                         10
        Use of country systems may be higher in more aid-dependent countries for two reasons.

First, delivering higher volumes of aid may be accomplished in part by disbursing a greater

percentage of it in the form of direct budget support, which by definition uses country PFM

systems. Second, the impact of donor decisions to use (or bypass) country systems on

strengthening (or undermining) government capacity and accountability is likely to be greater

where aid volumes are higher. These potential impacts may also be greater in lower-income

countries, perhaps explaining in part why governments of many middle-income countries

receiving modest levels of aid show relatively little concern over donors‟ choices of aid

management practices (Wood et al., 2011). The net effect of income on use of country systems

is therefore ambiguous.

        Donor characteristics comprise a third set of determinants of trust in country systems. A

first-level distinction, between multilateral and bilateral (i.e. national) donors, reflects their

differing mandates. Multilateral aid agencies were established in part to resolve collective action

problems plaguing bilateral donors. They are better insulated from political pressures to

demonstrate short-term visible results to elected officials and taxpayers (Wood et al., 2011).

Multilaterals “are cases of joint delegation from multiple principals� that may “enable the agency

to commit itself to procedures that would not be easy to implement for a bilateral donor, such as

transparent and competitive procedures for tendering and procurement� (Martens et al., 2002:

21). They also have a comparative advantage in aid activities involving “spillover effects�

which bilaterals “might have difficulty internalizing� (Martens et al., 2002: 65). Donor

coordination is part of the mandate of some multilaterals. The World Bank and UN partner with

the OECD-DAC in its harmonization agenda and in the Paris Declaration monitoring effort. Use




                                                   11
of country systems should therefore be greater in general for multilateral than for bilateral

donors.

          Among multilaterals, a further distinction can be made between the MDBs (multilateral

development banks including the World Bank, IMF and regional development banks) and other

multilateral donors including the UN system and European Commission (EC). The MDBs have

a “country-led business model� that aims to align operations in support of recipient countries‟

national development strategies (World Bank, 2011). The non-MDB multilaterals are a

heterogeneous group, but for multiple reasons we expect their use of country systems to fall short

of the MDBs‟. In the EC, “responsibility rests in the hands of serving politicians from member

states,� so its decisions are less apolitical than other multilateral agencies that “have genuinely

delegated their management to an executive board� (Martens et al., 2002: 47). National

representatives in EC foreign aid decision-making committees devote considerable effort to

pursuing opportunities for their own nation‟s aid contractors (Martens et al., 2002: 193). The

various UN agencies are treated as a single donor in the PDMS, with their data collected and

reported by the UNDP.11 Some UN agencies specialize in humanitarian aid, which is not

included in the PDMS. Other UN agencies provide mostly technical assistance, which rarely

uses country PFM systems.

          Among bilateral donors, the distinction between OECD-DAC donors and non-DAC

donors may be important.12 Use of country systems is expected to be higher for DAC donors, as

the DAC is the leading advocate of Paris Declaration objectives including increased use of




11
   Some bilaterals also have multiple aid agencies but the survey collects and reports data only at the national level,
e.g. for the USA and not separately for USAID, the MCC, etc.
12
   Non-DAC bilaterals in the sample include only Turkey and Korea. The latter is now a DAC member, but it was
not during the 2005-2010 period covered by the surveys.


                                                          12
country PFM systems. The DAC‟s peer reviews of its members‟ aid programs now include

assessments of their consistency with Paris Declaration principles and objectives.

       The DAC donors can be divided further, between the “Nordic Plus� group and others.

Nordic Plus donors include Denmark, Finland, Norway, Sweden, Ireland, the Netherlands, and

the United Kingdom. The group‟s purpose is to improve complementarities among its members,

through division of labor based on comparative advantages (NORAD, 2006; de Renzio, 2005).

By reducing the number of sectors and countries each donor operates in, transactions costs for

recipients can be reduced, at the price of reduced visibility for the donors. Nordic Plus

membership can thus be viewed as a proxy for low skepticism of aid‟s effectiveness among their

domestic constituencies. Empirical support for this hypothesis may be interpreted as merely

showing that donors committed to the harmonization elements of the PD agenda tend to be

committed also to its “alignment� elements. At a minimum, however, the Nordic Plus dummy

can usefully test for any significant variation among DAC bilaterals in use of country systems.

       The vertical funds (sometimes called global funds) comprise a last set of donors. These

donors have limited sector-specific mandates, such as the environment or specific diseases. In

the PDMS, most aid from vertical funds is accounted for by the Global Fund to Fight AIDS,

Tuberculosis and Malaria. It disbursed 2.2% of the aid represented in the survey - more than the

Netherlands or Canada. Vertical funds, particularly in health, are often viewed as being

motivated by concerns over global public goods, not integrated into national strategies, and

reluctant to use country systems (Wood et al. 2011: 34-35; World Bank, 2006: 22). For these

reasons vertical funds may be associated with lower use of country systems in general.

       For the DAC bilateral donors, we can go beyond these donor group dummies and attempt

to measure domestic constituents‟ trust in aid effectiveness in general using data from public




                                                13
opinion surveys. We expect stronger public support for development aid to increase a bilateral

donor agency‟s use of country systems. Where support for aid is relatively weak, aid agency

officials will be under more pressure to show that the funds they are provided produce visible

results directly attributable to its efforts, so will provide more aid in the form of projects using

parallel systems and less in the form of budget support.

       Data for testing this hypothesis are available from three different public opinion surveys:

Gallup International‟s 2002 “Voice of the People� survey, the 1995-1998 round of World Values

Surveys (WVS), and the 2004 Eurobarometer. These surveys each cover a somewhat different

sample of donor countries, as shown in Appendix Table A2. All 24 bilateral donors are

represented in one or more of the surveys, but only four (Finland, Germany, Spain and Sweden)

are included in all three. The question inquiring about support for development aid is worded

somewhat differently in each survey. The percentage of respondents indicating greater support

for aid in the WVS is correlated at .85 with the corresponding percentage from Gallup

International (GI), and at .49 with the percentage supporting aid in Eurobarometer. Support for

aid in GI and Eurobarometer, however, are correlated at only .18. The two donors with the

weakest support for foreign aid in both the WVS and GI are the U.S. and Japan (neither is

included in Eurobarometer).

       To maximize the information content from these three surveys and to preserve the sample

size, we construct an index of public support for aid. Values from each survey are first

standardized to have a mean of 0 and standard deviation of 1. Index scores, reported in Table 2,

are then computed as the unweighted mean of the index components that are available for each

country. The alpha coefficient of index reliability from this procedure is a very respectable .72,




                                                  14
with an average inter-item correlation of .46. Most of our tests employ this index, but we also

show how results are affected by instead using each of the three surveys in isolation.

         Year dummies are also included in the regressions where appropriate. The Paris

Declaration along with its associated monitoring indicators and follow-up conferences is

designed in part to create “peer pressure� among donor agencies toward meeting its numerical

targets.13 If pressure from peers, recipients, or civil society is increased as intended, coefficients

for year dummies included in the analysis should exhibit a significant positive trend.

         Donors choose whether or not to use country systems. Their choices, however, reflect

some combination of donor and recipient country characteristics. We estimate regressions of the

general form:

                             UCSijt  �Z ijt  �X it  �M jt 1  Yt  uijt        (6)

where UCSijt is the share of donor i ‟s aid to recipient j in year t that is recipient-managed, Z ijt

are regressors that vary by donor, recipient, and year; X it are regressors that vary by donor and

year but are recipient-invariant; M jt are regressors that vary by recipient and year but are donor-

invariant; and Yt is a vector of year dummy variables. Recipient variables are lagged by one

year. Donor variables are mostly time-invariant, including donor public opinion, which is

constructed from surveys conducted between 1 and 15 years prior to the years for which the

PDMS measures use of country PFM systems.

         The dataset can be treated as an unbalanced panel, with anywhere between 1 and 80

observations per donor. We can exploit this structure of the data to conduct stronger tests by

controlling for donor-year, recipient-year, or donor-recipient fixed effects. For example, in

13
  International donor conferences on harmonization bring senior managers of aid agencies “in close contact with
colleagues from other agencies, pushing them to align with recognized international best practice and not be seen as
laggards� (de Renzio, 2005: 11).


                                                         15
testing the robustness of results for Control of Corruption and other recipient characteristics, with

the inclusion of donor-year fixed effects estimates are informed only by cross-recipient variation

in the data within each of the 102 (= 34 x 3) donor-year panels. These regressions take the form

                              UCSijt  �Z ijt  �X i Yt  �M jt 1  uijt (7)

where X i Yt is a vector of donor-year fixed effects. The analysis similarly adopts a stringent and

highly conservative approach to determining statistical significance of estimates. All tests,

unless otherwise specified, correct standard errors for two-way clustering (Cameron, Gelbach

and Miller, 2011), for non-independence of errors within clusters of observations pertaining to

each donor and to each recipient.



3. Results

       Equation 1.1 of Table 1 presents results from the baseline model, which includes

observations from all 34 donors, by not including the public opinion variable that pertains only to

bilaterals. Most coefficients have the hypothesized signs, and the majority of them are

statistically significant at the .05 level. In countries where a donor has a larger share of the aid

market, its aid is more likely to be managed via country PFM systems. The coefficient of .45

indicates that for each 10 percentage point increase in the donor‟s share of aid in a country, its

use of country systems increases by 4.5 percentage points.

       Control of Corruption and Civil Liberties also have positive and significant coefficients.

Control of Corruption is measured in standard deviation units, so use of country systems is

nearly 9 points higher, other things equal, in a recipient scoring 1 standard deviation higher than

another recipient. Each 10-point increment in the civil liberties index (scaled from 0 to 100) is

associated with an increase in use of country PFM systems by 5.5 percentage points.



                                                   16
        Donors use country systems more in larger and more aid-dependent recipients. Neither

per capita income nor the recent growth rate of income is significant.

        The “donor type� dummies in general produce results consistent with expectations.

Relative to the omitted category of other DAC bilaterals, use of country systems is about 17

percentage points higher on average for the “Nordic Plus� group, and 15 points higher for the

MDBs. Use of country systems is 15 percentage points lower for the non-DAC bilaterals14, and

(somewhat surprisingly) 9 points higher on average for vertical funds, than for the omitted

category. However, the latter difference falls far short of statistical significance at conventional

levels. “Other multilaterals� do not differ from the reference group.

        Coefficients for the year dummies are negative, indicating that use of country systems is

higher by 3 or 4 percentage points in 2010, other things equal. These dummies are not

significant, however. These coefficients may underestimate the actual trend, as anecdotal

evidence indicates that survey coordinators in later rounds of the survey were able to hold donors

to account for more accurate reporting and stricter application of the indicator criteria (OECD,

2011; OECD, 2009a).

        Equation 1.2 adds the public opinion index. This variable is undefined for the

multilateral donors and vertical funds, so its inclusion reduces the sample size by about two-

fifths. A 1-standard-deviation increase in support for aid is associated with a 6.6-point increase

in use of country systems by bilateral donors on average, significant at the .01 level. This result

is consistent with the prediction that stronger domestic support for overseas aid allows donors to

be less risk averse and to place a relatively high weight on broader, longer-run development

objectives as opposed to visible and short-term outputs.

14
  The non-DAC bilateral category in our sample includes only Korea and Turkey, but the result appears to
generalize more broadly. In 2010, 42.9% of all aid from DAC bilaterals used country PFM systems, compared to
6.6% from Korea and Turkey, and only 4.9% from the other non-DAC bilaterals excluded from the regressions.


                                                      17
       Results for most other variables in equation 1.2 are very similar to those in equation 1.1.

The most notable difference is that the coefficient for non-DAC bilaterals increases in magnitude

from -15 to -23. Controlling for their very high levels of support for aid, Korea and Turkey stand

out even more strongly as negative outliers on use of country systems. It is not clear how to

interpret the public opinion results for these two countries, however. Korea and Turkey were

historically aid recipients, and their aid outflows are still small. The public opinion index for

them is measured using only the GI 2002 survey, in which 87% of respondents in each country

thought that “the wealthier nations should give more financial help to the poorer nations.� Many

respondents in both countries might not have considered their own country to be among the

“wealthier nations,� and some (particularly in Turkey) may even have considered their own

country to be among the “poorer nations� on the receiving end. In equation 1.3 we therefore

drop the 52 observations in which Korea or Turkey are the donors. Results for the public

opinion variable (and for the other regressors) are unaffected by this change in sample, however.

       The remainder of this section addresses three potential problems with the basic results

reported in equations 1.1-1.3. First, those estimates are potentially affected by sample selection

bias. Second, there may be omitted variables correlated with use of country systems. Third,

coefficients for some key regressors may incorporate endogeneity bias.

       There are two sources of potential selection bias in the data. First, aid-recipient

governments self-select into (or out of) the PDMS. Of the 151 eligible aid recipients on the

official DAC list, only 80 participated in one or more rounds of the survey. Second, not all

donors choose to provide aid to all recipients in all years (or, in some cases, recipients may

choose not to accept aid). In fact, for the vast majority of all possible donor-recipient pairs, no

aid flows are recorded, as shown in Appendix Table A3. For 2010, the year with the highest




                                                 18
participation by recipient countries, pairing each of the 34 donors in our sample with each of the

76 recipients would imply a total of 2584 potential aid relationships. This figure represents only

50.3% (row 7 of the table) of the 5134 total potential aid relationships implied by pairing each of

the 34 donors with each of the 151 eligible DAC recipients. Of the 2584 potential pairings in the

survey, a positive aid flow is recorded in only 998 cases, or 38.6% (row 8). These 998 active aid

relationships account for only 19.4% (row 9) of the 5134 total pairings of the 34 donors with

each of the 151 DAC recipients. For the earlier survey years of 2005 and 2007 these percentages

were even lower. Pooling the three survey years (last column of Table A3), nearly 64% of

potential observations are censored by recipient nonparticipation in the PDMS (although the

majority of these would be censored anyway due to zero aid flows), and more than 60% of the

remaining pairs are censored by donors‟ provision of zero aid.

       There is some evidence suggesting that the censored observations may differ

systematically from the uncensored observations, potentially creating a selection bias in our use

of country systems tests. Use of country systems averages 38.3% for 2010 among all

observations corresponding to the initial (2005) set of PDMS recipient participants. The average

for 2010 is only 28.5% for recipients that joined the survey in 2007, and is only 26.5% for

recipients that did not participate until 2010. Very few recipients dropped out, but many were

added over the three rounds, and use of country systems is markedly higher in the earlier

volunteers. Uncensored observations differ significantly from censored observations on all of

the recipient-level regressors in the analysis: they are larger, poorer, more aid-dependent, grow

more slowly, and score less well on Control of Corruption and Civil Liberties.

       Correcting for potential selection bias using the Heckman approach requires identifying

exogenous variables that affect selection by recipients into the PDMS, and by donors and




                                                19
recipients into active aid relationships. Selection regressions are reported in Appendix Table A4.

The sample in Equation A4.1 is the full set of 15,402 pairings of the 34 donors and 151 eligible

DAC recipients for each of the three survey years.

        Selection variables include HIPC status, colonial tie, and distance from the donor capital

or headquarters city (in log of kilometers). Countries already intensively engaged with donors to

qualify for debt relief may have a greater interest in results, as well as lower participation costs,

so we expect a dummy for countries reaching the HIPC completion point to be positively related

to survey participation.

        Colonial ties are shown to influence aid allocations by donors across recipients in Alesina

and Dollar (2000), among other studies. We create a colonial tie dummy variable equal to 1 for

all donor-recipient pairs (such as UK-Ghana) where the recipient was once part of the donor‟s

colonial empire.15 A colonial tie is present in 4.9% of the PDMS observations but in only 1.9%

of the censored cases. Physical distance has also been shown to influence donor‟s aid decisions

(e.g. Frot and Perrotta, 2011). Distance is less meaningful and its effect may be weaker for

multilaterals, however, as some of them have global mandates (UN, World Bank) or interests

(EC). We therefore create two separate distance variables, one for bilaterals (with multilaterals

coded 0) and another for multilaterals (with bilateral coded 0).

        Appendix Table A4 reports probit regression results for the selection variable, coded 1

for uncensored and 0 for censored observations. The coefficient estimates presented in the table

represent marginal effects, evaluated at the means of all of the other regressors. Equation A4.1

includes data from all three survey years. The selection variables are all highly significant. The

probability of a donor-recipient pair being uncensored increases by 20 percentage points if they


15
 Colonies are assigned only to their last colonial master; e.g. Rwanda and Burundi are assigned only to Belgium,
which occupied those parts of German East Africa in 1916.


                                                       20
have a colonial tie, by 26 points if the recipient had reached the HIPC completion point (for full

irrevocable debt cancellation), and by 11 points if it had reached the HIPC decision point (for

interim debt relief). Greater geographic distance separating the donor from the recipient

significantly increases the probability of censoring, as expected. The selection tests also include

all of the regressors from equation 1.1 that do not reduce the sample size in A4.1, i.e. those that

are available for all 15,402 data points. Pairs in which the donor is a non-DAC bilateral are

significantly more likely to be censored, and larger recipients are less likely to be censored. Year

dummies confirm that censoring has diminished with each subsequent round of the PDMS.

       The three remaining equations in Table A4 report results for 2005, 2007 and 2010

respectively. Country size, colonial ties and distance matter more in the later years, while HIPC

status matters more in the earlier years. The pseudo-R2 for this selection model drops from .26

in 2005 to .23 in 2007 and .16 in 2010, suggesting that the sample becomes more representative

as additional recipient countries join later rounds of the survey.

       To test and correct for potential sample selection bias in our use of country systems

regressions, we compute inverse Mills‟ ratios from equations A4.2-A4.4, allowing the impact of

selection variables to vary by survey year (Wooldridge, 2002). Equation 1.4 of Table 1 adds the

inverse Mills‟ ratio (IMR) to the base specification of equation 1.1. The IMR is inversely related

to the predicted probability of selection, so its negative coefficient estimate in equation 1.4

implies that use of country systems is positively related to the selection probability. It is not

significant at conventional levels, however, so there is only weak evidence of selection bias.

Moreover, the addition of the IMR to the model has little effect on results for key variables such

as the donor‟s share of aid. For simplicity, therefore, we do not correct for potential selection

bias in reporting additional robustness tests in Tables 2 through 6. Note that these tests all




                                                 21
include donor-year, recipient-year, or donor-recipient fixed effects, and the potential for selection

bias is reduced (although not eliminated) in the presence of fixed effects (Wooldridge, 2002).

       Table 2 tests the robustness of results for Control of Corruption and other recipient

characteristics. The possibility of omitted variable bias is addressed in two ways. First, each

regression includes donor-year fixed effects, so estimates are informed only by cross-recipient

variation in the data within each donor-year panel of observations. This method effectively

controls for any omitted donor characteristics, even if they vary by survey year. Second, these

tests control for additional regressors that vary by recipient or by recipient-donor pairs.

       Equation 2.1 differs from equation 1.1 only by replacing donor type variables (and year

dummies) with donor-year fixed effects. This change increases the coefficient on donor aid

share from .45 to .71, and the t-statistic increases from 2 to 3.5. Results for other variables,

including Control of Corruption and Civil Liberties, change very little.

       Equation 2.2 adds a dummy variable for “fragile states� (listed in OECD, 2011: 104).

This dummy is significant at the .05 level, and use of country systems is 11 percentage points

lower on average in these countries. Controlling for the fragile states dummy, coefficients for

Control of Corruption and Civil Liberties diminish slightly, but remain significant.

       Equation 2.3 controls for HIPC status and colonial ties. In using them as selection

variables in Table A4, we assumed they had no direct effect on use of country systems.

However, donors might perceive a larger reputational stake in the development of their former

colonies, and internalize more of the benefits of investments in strengthening their PFM systems.

The more intensive engagement of with donors, including relatively high levels of budget

support, could imply increased use of country systems in HIPC countries, other things equal.

Neither the HIPC variables nor colonial ties are significant in equation 2.3, however, supporting




                                                 22
their validity as selection variables in Table A4. Moreover, their inclusion in equation 2.3 has

little effect on results for donor aid share and most of the recipient-level variables. The

coefficient for aid dependence declines somewhat when HIPC indicators are included, as the

HIPC countries tend to receive more aid than most others.

       Equation 2.4 adds the percentage of aid for each donor-recipient pair that is in the form of

direct budget support (DBS), as reported in the PDMS. Budget support by definition uses

country PFM systems. When this variable is included, the regression in effect is testing for

relationships between the regressors and use of PFM systems for project aid and other non-DBS

aid. Coefficients that were significant in equation 1.1 remain significant (although only at the

.10 level for Control of Corruption) in equation 1.4, but are reduced in magnitude.

       If using country systems strengthens them while bypassing them weakens them (OECD,

2011; 2008a), then quality of country systems is endogenous. We address this potential problem

in several ways. First, any beneficial effects on quality from using PFM systems will be on the

PFM systems themselves: budget execution, financial reporting, and auditing procedures. Our

system quality indicators - Control of Corruption, supplemented by Civil Liberties – are defined

far more broadly, and are likely affected only with a lag. Second, we lag the system quality

indicators by a year in our tests. Third, we test for the presence of heterogeneous effects among

donors that should be present – or absent - if coefficients on Control of Corruption are biased

upward to a significant degree by reverse causation.

         Our basic identification strategy is premised on differences among donors. First, we

distinguish donors by size. The impact of a percentage point increase in a donor‟s use of country

systems should vary with a donor‟s aid share. Each $1 increase should have the same impact, so

1 percentage point will imply a larger impact for larger donors. A sufficiently small donor could




                                                 23
safely take system quality as given in its decisions. In contrast, there is no particular reason to

expect the causal impact of quality of country systems on use of country systems to vary by

donor size. Therefore, if coefficients on control of corruption in Tables 1 and 2 mostly reflect

endogeneity bias, the strength of association should vary positively with donor size. We add the

appropriate interaction term in equation 2.5 to test this hypothesis.16 Results are highly

inconsistent with the argument for reverse causation: the coefficient on the interaction term is

negative (and significant at the .10 level for a two-tailed test), not positive.

         In our second variant of this identification strategy, we distinguish donors by selectivity

in aid allocations. Specifically, we posit that donors that are more sensitive to quality of country

policies and systems when allocating aid will also be more sensitive to the quality of systems

when deciding whether to rely on them for managing their aid. In contrast, there is no particular

reason to expect the strength of any reverse causation from use of country systems to vary with

donors‟ sensitivity to policy in allocating aid. An empirical finding of heterogeneous impacts in

this instance can be explained better by the causal argument than by reverse causality (from use

of systems to system quality). Equation 3.1 includes an interaction term, to test the hypothesis

that Control of Corruption will affect use of country systems more for donors that are more

policy selective in allocating aid among recipients.17 For this purpose, we use the donor policy

selectivity scores from Knack, Rogers and Eubank (2011), based on aid disbursements in 2007.

The interaction coefficient is positively signed, as hypothesized, and significant at the .05 level.




16
   The interaction term is equal to the product of the deviations of Control of Corruption and donor aid share from
their respective means. The coefficient on Control of Corruption thus indicates its marginal effect conditional on the
mean value of donor aid share.
17
   The interaction term is equal to the product of the deviations of Control of Corruption and policy selectivity from
their respective means. The coefficient on Control of Corruption thus indicates its marginal effect conditional on the
mean value of policy selectivity. The coefficient on policy selectivity is subsumed by the donor-year fixed effects,
because it is donor-invariant.


                                                         24
       Taken together, our measurement and identification strategies for estimating the impact

of the quality of country systems do not entirely rule out endogeneity bias from our estimates.

However, these strategies coupled with empirical results from equations 2.5 and 3.1 suggest that

the direction of causality is predominantly from quality to use of country systems.

       Next, we address the possibility of omitted variable or endogeneity bias in the coefficient

for donor aid share. Two potential objections are easily answered. First, the donor aid share

coefficient is not merely showing that larger donors use country systems more. Tests in Tables 2

and 3 include donor-year fixed effects, so results imply that a given donor makes more use of

country systems in those recipients where its share of aid is larger, as predicted. Second, higher

quality country systems might induce donors both to increase their aid and to manage a larger

share of it through country systems. If quality of systems is imperfectly measured by Control of

Corruption and Civil Liberties, the donor aid share coefficient might appear to be affected by

simultaneity bias. Donor aid shares will be unchanged, however, if all donors increase their aid

proportionately, so this argument for bias is largely unfounded.

       The simultaneity critique does apply if we allow donors‟ perceptions of corruption (or

other aspects of the quality of country systems) to vary for a given recipient. Donors with

relatively high risk perceptions may respond by reducing use of country systems and reducing

aid – which will now reduce the aid share, because not all donors are reducing their aid. In

equation 3.2 we control for risk perceptions that vary across donor-recipient pairs, by introducing

donor-recipient fixed effects. Not incidentally, these pair dummies also control for effects of any

historical, commercial or other ties that might increase a donor‟s interest or reputational stake in

a recipient‟s long-run development. In this specification estimates are informed only by

variation over time within each donor-recipient pair. This test is particularly stringent, as some




                                                 25
regressors may exhibit little variation over the three periods, covering only a 5-year time span.

Nevertheless, most variables significant in previous tests remain significant in equation 3.2. The

exception is Control of Corruption, which varies little over time for most recipients. The

coefficient for donor aid share is somewhat lower in equation 3.2 than in tests using donor-year

fixed effects, but it remains positive and significant at the .01 level.

        We are not aware of any plausible argument regarding reverse causation from use of

country systems to donor aid share. Nevertheless, the remainder of Table 3 addresses this

possibility, again by exploiting differences among donors as an identification strategy. The

theoretical argument concerning the impact of donor aid share on use of country systems applies

more strongly to bilateral donors in general than to multilaterals. Multilaterals‟ decisions are less

sensitive to political pressures and arguably to their ability to claim credit for results. Their

allocations (and hence aid shares in a country) are determined partly by their mandates, including

funding formulas established by their membership. For these reasons, a donor‟s aid share should

be a better proxy for its reputational stake in a country‟s development for bilaterals.

        These arguments are supported in Table 3. Equations 3.3 and 3.4 run the specification of

equation 2.1 separately for bilateral and multilaterals, respectively. While donor aid share

coefficients are positive and significant in both cases, the magnitude is three times as large for

the bilaterals. Equation 3.5 retains the combined sample but introduces an interaction term,

equal to the product of donor aid share and a multilateral dummy. The coefficient on this

interaction is significant at the .05 level.

        Table 4 addresses omitted variable bias in results for donor type variables that may be

present in Table 1. Most notably, these tests control for recipient-year fixed effects, so rule out




                                                  26
bias from missing recipient characteristics. Results on donor type variables and donor aid share

in equation 4.1 are surprisingly similar to those in equation 1.1, which included no fixed effects.

        Equation 4.2 tests robustness of donor type results to controlling for differences in

donors‟ priorities or mandates, as proxied by the sectoral composition of their aid. These

variables, constructed from the DAC‟s Creditor Reporting System, vary by donor and year, but

not over recipients. They potentially matter in either of two ways. First, sectoral priorities may

be correlated with other aid delivery preferences. Second, aid in sectors such as infrastructure is

mostly delivered in the form of projects, so less likely to use country systems than (for example)

education aid, which often takes the form of sectoral budget support to help fund teacher

salaries.18 The five sectoral variables in equation 4.2 are all significant, relative to the base

category of all other sectors (including multisectoral aid). Infrastructure, health and governance

aid have negative signs. Donors providing larger shares of their aid for governance reform may

be the ones that perceive greater governance-related risks, accounting for the negative

relationship with use of country systems. Of the donor type variables, non-DAC bilaterals are no

longer significantly different when the sector shares are added, but this is due in part to missing

sectoral data for Turkey: the non-DAC bilateral dummy is in effect a Korea dummy in equation

4.2. The coefficient on vertical funds increases in magnitude and significance, controlling for

the fact they mostly provide health aid, which is associated with lower use of country systems.

        Equation 4.3 adds three more proxies for donor preferences, but at the expense of

dropping the multilaterals due to missing data. Technical assistance as a share of aid is not

significant. Contributions to multilateral donors as a share of total aid predicts reliance on

country systems: donor countries that are reluctant to delegate to multilaterals tend to be the


18
  The PDMS does not collect information on sector, so we cannot directly compare use of country systems across
sectoral categories.


                                                      27
same ones that are reluctant to delegate the management of aid funds to country PFM systems.

Donors that tie more of their aid to purchases from home-country suppliers are significantly less

likely to use country systems. Inclusion of these variables does not materially affect results on

donor aid share, or on the Nordic Plus dummy.

       Equation 4.4 adds a USA dummy. Corruption concerns are voiced often in U.S. debates

on foreign aid, in Congress and in the media, and Congressional earmarks and reporting

requirements severely restrict the use of budget support or reliance on recipients‟ procurement

and other PFM systems in managing project aid (OECD, 2006). Equation 4.4 confirms that U.S.

aid use country systems at a far lower rate than other DAC bilaterals, even when controlling for

donor preferences via sectoral composition, tying, etc.

       Equation 4.5 replicates equation 4.1, but replaces the “other multilateral� dummy with

separate dummies for the two donors belonging to this residual category of multilaterals that are

neither MDBs nor vertical funds. The UN and EC differ in important respects as noted in section

3. Coefficients for the UN and EC show that use of country systems is lower in the former and

higher in the latter, relative to the omitted group of DAC bilateral. The negative UN effect (but

not the positive EC effect) disappears when sector shares are added to the model (results not

shown in table).

       Table 5 addresses the donor public opinion measures in more detail. As in equation 1.4,

Korea and Turkey are dropped, but Table 5 regressions (as in Table 4) control for recipient-year

fixed effects. In equation 5.1, public support for aid is significantly and positively associated

with use of country systems, with a coefficient very similar to those in equations 1.1 and 1.2. A

one standard deviation increase in support is associated with an increase of about 7 percentage

points in use of country PFM systems. Equations 5.2-5.4 respectively substitute for the index its




                                                 28
three separate component measures, from surveys covering different sets of donor countries and

using somewhat different question wording. Coefficients are positive and significant using GI

(equation 5.2) and WVS (equation 5.3), but not for Eurobarometer (equation 5.4). Each 10-point

increase in the percentage of GI respondents (or 5-point increase for WVS respondents) is

associated with an increase in use of country PFM systems of more than 4 percentage points.

         Equation 6.1 shows that results for the index of public support are robust to the inclusion

of sector shares and other donor-level control variables from Table 4. Public support is weakest

for the USA, providing a possible explanation for its extremely low usage of country systems as

shown in equation 4.4. The USA dummy coefficient remains strongly significant, however,

when it is tested together with public support for aid in equation 6.2. The coefficient magnitude

of -34 is not so different from the -37 in equation 4.4, so the outlier status of the USA is largely

unexplained by the substantive variables in the analysis.19 The USA is an influential observation

however in testing the link between public support for aid and use of country systems. The

public support index is significant only at the .06 level in equation 6.2, with a coefficient only

half as large as in equation 6.1.

         Public support for aid is potentially endogenous to its effectiveness, which may in turn be

influenced by donor‟s use of country systems. The opinion surveys were all conducted between

1995 and 2004, prior to the 2005-2010 period for which the PDMS measures use of country

systems. This timing sequence reduces but does not eliminate the potential for reverse causation.

We therefore instrument for public opinion in equation 6.3, with first-stage results shown in



19
  The large negative USA coefficient reflects its relatively tepid support for some of the PD principles. Due to U.S.
opposition the PD omitted targets for budget support, and weakened targets for use of country procurement systems
(Mokoro Ltd., 2008b: 23). In contrast, the U.S. led the effort to include a provision on “managing for results� in the
PD because “reporting on results is critical to demonstrating aid effectiveness and to sustaining public and
Congressional support for U.S. assistance� (USAID, 2006). However, USAID‟s own reporting system “focuses
mainly on physical deliverables� such as number of schools or clinics, instead of outcomes (OECD, 2006).


                                                         29
equation 6.4. The exclusion restriction assumes donor country size and number of former

colonies predict support for aid but do not independently affect use of country PFM systems.

Citizens of smaller countries, or of countries with a more active history of colonization, may in

general be better informed about poverty outside their own borders and feel a greater obligation

to alleviate it (Paxton and Knack, 2011). Total number of former colonies may be correlated

with a dummy for whether each donor-recipient pair represents a former colonial relationship.

We thus control for a colonial tie dummy in the 2SLS test.

       As shown in the first-stage diagnostics in equation 6.4, the exogenous instruments have

strong predictive power for public opinion. Donor country population has the expected negative

sign, and number of former colonies has the expected positive sign. The applicable test of over-

identification does not reject the assumption of instrument validity (p value = .30). The

coefficient on public opinion in the second stage regression (equation 6.3) is positive and

significant. The coefficient size is very similar to those in OLS tests in equations 1.4 and 5.1.

While there may be alternative possible interpretations of this relationship, results are consistent

with our argument that stronger domestic support for aid allows donors more freedom to pursue

medium- or long-run development objectives less constrained by the need to demonstrate visible

outputs and avoid any risks.



4. Conclusions

       This paper identifies and tests three broad explanations for donor agencies‟ inability or

unwillingness to rely more on country systems. First, and most fundamentally, country systems

are often not very trustworthy so their use implies significant reputational and fiduciary risks to

donors. We provide empirical evidence supporting the proposition that use of country systems is




                                                 30
strongly linked to their quality, as proxied by indicators of corruption perceptions and civil

liberties. Knack and Eubank (2009) obtained a similar result using an indicator of PFM quality

that is produced by World Bank staff.

       Second, donors‟ trust in country systems is influenced by their mandates and by their

constituents‟ faith in the development effectiveness of aid. Multilateral donors (particularly the

development banks) use country systems more than bilaterals. There is substantial diversity

within each category, however, and in bilaterals popular support for aid is associated with greater

use of country systems. Where voters are more skeptical of aid, donors tend to micro-manage

aid to produce tangible, visible outputs that can be plausibly attributed to its own activities.

       Third, we find that a donor‟s use of country systems is greater when it has a more

“encompassing interest� in a recipient‟s development. When a donor‟s share of all aid provided

to the recipient is higher, it internalizes more of the benefits of its use of country systems. These

three key findings are robust to a variety of corrections for possible sample selection, omitted

variables and endogeneity bias. Results for donor aid share (and Civil Liberties) are even robust

to the inclusion of fixed effects for donor-recipient pairs, where estimates are informed only by

variation in the data over the three survey years within each unique pair.

       In contrast, pressure from other donors, recipients or civil society groups created by the

Paris Declaration campaign appears to be relatively unimportant in explaining use of country

systems. Coefficients from year dummies fail to show a significant positive trend over the 2005-

2010 period, controlling for quality of country systems and other variables.

       If donors‟ risk tolerance is mostly unaffected by the Paris Declaration campaign, broad-

based and sizeable increases in the use of country systems may depend largely on improving

their quality. However, successful reform of PFM systems tends to occur only over long time




                                                  31
periods, and involves political as well as technical challenges, with resistance from certain public

officials who benefit from maintaining less-transparent systems and procedures (OECD, 2009b;

Allen, 2009; IEG, 2008).

       Concentrating a typical donor‟s aid in fewer countries and sectors would reduce the

enormous transactions costs imposed on recipient governments. It would also tend to increase

donors‟ use of country systems, as the leading donors in a recipient would internalize a greater

share of the benefits. Even if donors‟ aid shares were all unchanged in a country, but divisions

of labor by sector were sharpened, donors would have less incentive to micro-manage aid. The

lead donor for education, for example, would have a strengthened reputational stake in education

outcomes, and similarly for the lead donor in health. Incentives for them would shift away from

delivering donor-managed, successful-looking projects, and toward working with governments

to deliver improved sector-wide development outcomes. Concentrating aid in fewer countries or

sectors, however, potentially reduces a donor agency‟s visibility, with possibly adverse

consequences for its budget (see Andersen, 2000: 193).

       Of course, the urgency of pursuing such measures for the purposes of increasing use of

country systems depends on learning more about the latter‟s effects on development outcomes.

When donors choose to rely on country systems or bypass them, how important are the

associated “learning-by-doing� externalities? While the Paris Declaration‟s premise that the

quality of aid matters as much as its quantity is surely correct, research on the implications of

donor‟s aid management practices for aid quality is still in its infancy. Further research using the

PDMS data can test for the effects of different categories of aid – based on whether it is

delivered in ways consistent with PD principles - on governance and other outcomes.




                                                 32
                                        REFERENCES

Alesina, Alberto, Dollar, David, 2000. Who Gives Foreign Aid to Whom and Why? Journal of
       Economic Growth 5, 33-63.
Allen, Richard, 2009. The Challenge of Reforming Budgetary Institutions in Developing
       Countries. IMF Working Paper WP/09/96. Washington DC: International Monetary
       Fund.
Andersen, Ole Molgard, 2000. Sector Programme Assistance. In: Tarp, F. (Ed.), Foreign Aid and
      Development. Routledge, London.
Anderson, Edward, 2011. Aid Fragmentation and Donor Transaction Costs. University of East
      Anglia, School of International Development.
Angrist, Joshua D., Pischke, Jorn-Steffen Pischke, 2009. Mostly Harmless Econometrics: An
      Empiricist‟s Companion. Princeton University Press, Princeton.
Angeles, Luis, Neanidis, Kyriakos C., 2009, Aid Effectiveness: The Role of the Local Elite.
      Journal of Development Economics 90(1), 120-134.
Azam, Jean-Paul, Devarajan, Shantayanan, O‟Connell, Stephen A., 1999. Aid Dependence
      Reconsidered. World Bank Policy Research Working Paper 2144. World Bank,
      Washington, DC.
Birdsall, Nancy, Kharas, Homi, 2011. Quality of Official Development Assistance Assessment.
       Center for Global Development, Washington, DC.
Bjornskov, Christian, 2010. Do elites benefit from democracy and foreign aid in developing
       countries? Journal of Development Economics 92(2), 115-124.
Brautigam, Deborah, Knack, Stephen, 2004. Aid Dependence, Institutions and Governance in
       Sub-Saharan Africa. Economic Development and Cultural Change 52(2), 256-85.
Cameron, A. Colin, Gelbach, Jonah B., Miller, Douglas L., 2011. Robust Inference With Multi-
      Way Clustering. Journal of Business and Economic Statistics 90(3), 414-27.
Collier, Paul, 2006. Is Aid Oil? An Analysis of Whether Africa Can Absorb More Aid. World
        Development 34(9), 1482-97.
De Renzio, Paolo, 2005. Incentives for Harmonization and Alignment in Aid Agencies. ODI
      Working Paper 248. London: Overseas Development Institute.
Djankov, Simeon, Montalvo, Jose G., Reynal-Querol, Marta, 2009. Aid With Multiple
      Personalities. Journal of Comparative Economics 37, 217-29.
Easterly, William, 2007. Are Aid Agencies Improving? Economic Policy, 52, 634-68.


Easterly, William, Pfutze, Tobias, 2008. Where Does the Money Go? Best and Worst Practices
       in Foreign Aid. Journal of Economic Perspectives 22(2), 29-52.


                                               33
Frot, Emmanuel, Perrotta, Maria, 2011. Aid Effectiveness: New Instruments, New Results?
       Stockholm Institute of Transition Economics. Working Paper No. 11.
IEG, 2008. Public Sector Reform: What Works and Why. World Bank Independent Evaluation
       Group, Washington DC.
Jansen, Erik G., 2009. Does Aid Work? Reflections on a Natural Resources Programme in
       Tanzania. U4 Issue 2009:2. U4 Anti-Corruption Resource Centre Issue, Bergen.
Knack, Stephen, 2009. Sovereign Rents and Quality of Tax Policy and Administration. Journal
       of Comparative Economics 37(3), 359-71.
Knack, Stephen, Eubank, Nicholas, 2009. Aid and Trust in Country Systems. Policy Research
       Working Paper 5005. World Bank, Washington DC.
Knack, Stephen, Rahman, Aminur, 2007. Donor Fragmentation and Bureaucratic Quality in Aid
       Recipients. Journal of Development Economics 83(1), 176-97.
Knack, Stephen, Rogers, F. Halsey Rogers, Eubank, Nicholas, 2011. Aid Quality and Donor
       Rankings. World Development 39(19), 1907-17.
Knack, Stephen, Smets, Lodewijk, 2012. Aid Tying and Donor Fragmentation. World Bank
       Policy Research Working Paper 5934. World Bank, Washington DC.
Martens, Bertin, Mummert, Uwe, Murrell, Peter, Seabright, Paul, 2002. The Institutional
      Economics of Foreign Aid. Cambridge University Press, Cambridge.
Mokoro Ltd., 2008a. Putting Aid on Budget: A Study for the Collaborative Africa Budget
      Reform Initiatvie and the Strategic Partnership with Africa (Synthesis Report). Mokoro
      Ltd., Oxford.
Mokoro Ltd., 2008b. Putting Aid on Budget (Literature Review). Mokoro Ltd., Oxford.
Moore, Mick, 1998. Death Without Taxes: Democracy, State Capacity, and Aid Dependence in
      the Fourth World. In: White, G., Robinson, M. (Eds.), Towards a Democratic
      Developmental State. Oxford University Press, Oxford.
NORAD, 2006. Nordic Plus: Practical Guide to Delegated Cooperation. Norwegian Agency for
    Development Cooperation, Oslo.
OECD, 2011. Aid Effectiveness 2005-10: Progress in Implementing the Paris Declaration.
     OECD, Paris.
OECD, 2009a. Aid Effectiveness: A Progress Report on Implementing the Paris Declaration.
     OECD, Paris.
OECD, 2009b. Managing Development Resources: The Use of Country Systems in Public
     Financial Management. OECD, Paris.
OECD, 2008. 2008 Survey on Monitoring the Paris Declaration: Effective Aid by 2010? OECD,
     Paris.
OECD, 2006. Peer Review of the United States. OECD Development Assistance Committee,
     Paris.
Olson, Mancur, 1982. The Rise and Decline of Nations. Yale University Press, New Haven.



                                              34
Paxton, Pamela, Knack, Stephen Knack, 2012. Individual and Country-Level Factors Affecting
       Support for Foreign Aid. International Political Science Review 33:2 (forthcoming,
       March).
USAID, 2006. Guidance to USAID Field Missions: Frequently Asked Questions on Delivering
     Aid More Effectively. U.S. Agency for International Development, Washignton, DC.
     (http://pdf.usaid.gov/pdf_docs/Pdach113.pdf)
Williamson, Tim, Agha, Zainab Kizilbash, 2008. Stumbling Blocks or Building Blocks? The
       Effectiveness of New Approaches to Aid Delivery at the Sector Level. ABIA Project
       Working Paper No. 6. Overseas Development Institute, London.
Wood, Bernard, 2011. The Evaluation of the Paris Declaration Phase 2: Final Report. Danish
      Institute for International Studies, Copenhagen.
Wooldridge, Jeffrey, 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press,
      Cambridge MA.
World Bank, 2011. The World Bank and Aid Effectiveness: Performance to Date and Agenda
      Ahead. World Bank: Washington, DC.
World Bank, 2006. Integrating Global Partnership Programs With Country-Led National
      Programs. Background Document for the International Workshop on Global Programs
      and the Paris Agenda. World Bank, Washington DC.
World Bank, 2003. World Development Report 2004: Making Services Work for Poor People.
      World Bank and Oxford University Press, Washington DC.
World Bank, 1998. Assessing Aid: What Works, What Doesn‟t, and Why. World Bank Policy
      Research Report. World Bank and Oxford University Press, Washington, DC.




                                             35
                                                Table 1
Equation                            1.1               1.2              1.3             1.4
                                                Public support   Korea, Turkey
Variation                        Baseline                                           Heckman
                                                for aid added      dropped
D share of aid in R (%)           0.453**          0.713**         0.712**           0.448**
                                    (2.02)          (2.09)           (2.05)            (2.07)
Log of R per capita income         -1.860           -2.739          -2.829            -0.565
                                   (-0.99)          (-1.14)         (-1.14)           (-0.30)
Income growth rate in R             0.459            0.134           0.011            0.721*
                                    (1.05)          (0.29)           (0.03)            (1.72)
Log of R population              9.105***        10.820***        11.484***         7.412***
                                    (5.16)          (3.99)           (4.05)            (4.05)
Log of aid per capita in R       8.516***        11.066***        11.813***         6.953***
                                    (3.02)          (2.67)           (2.75)            (2.75)
Civil liberties in R             0.554***         0.631***         0.617***         0.517***
                                    (3.60)          (3.33)           (3.04)            (3.43)
Control of corruption in R        8.754**          8.195**          8.892**          7.944**
                                    (2.49)          (1.98)           (2.10)            (2.38)
Multilateral devp. bank         15.292***                                          15.277***
                                    (2.78)                                             (2.81)
Vertical fund                       9.044                                              7.724
                                    (0.91)                                             (0.78)
Other multilateral (UN, EC)        -1.568                                             -5.188
                                   (-0.19)                                            (-0.63)
Non-DAC bilateral               -14.922***       -22.920***                        -12.139***
                                   (-3.85)          (-4.65)                           (-2.55)
Nordic Plus                     16.647***         14.582***        14.476***       17.078***
                                    (3.33)           (3.17)           (3.17)           (3.45)
Public support for aid in D                        6.569***         6.562***
                                                     (2.59)           (2.61)
Year=2007                         -3.782            -1.541           -1.826           -3.794
                                  (-1.46)           (-0.57)          (-0.65)         (-1.52)
Year=2005                         -3.531            -1.120           -1.544           -2.506
                                  (-1.00)           (-0.29)          (-0.38)         (-0.68)
Inverse Mills ratio                                                                   -6.710
                                                                                     (-1.59)
Constant                           -158.33          -193.52          -205.36         -124.47
                                   (-3.96)           (-3.21)         (-3.26)         (-3.08)
No. of observations                  2110             1274             1222            2110
No. of Donors, Recipients           34, 77           24, 75           22, 75          34, 77
R2                                   0.16              0.20            0.19             .16
 Dependent variable is percentage of donor aid to government managed by recipient
 government‟s public financial management systems. T-statistics, reported in parentheses below
 point estimates, are based on standard errors adjusted for non-independence of errors within both
 donor and recipient clusters of observations, with *** p<0.01, ** p<0.05, * p<0.1.



                                                36
                                              Table 2
Equation                          2.1           2.2             2.3            2.4            2.5
Fixed effects                D-Y pairs       D-Y pairs      D-Y pairs      D-Y pairs       D-Y pairs
D share of aid in R (%)        0.712***      0.737***        0.703***        0.493**       0.740***
                                  (3.52)        (3.67)         (3.50)         (2.50)         (3.71)
Log of R per capita              -1.895        -3.021          -0.003         -0.812         -1.835
Income                           (-0.98)       (-1.48)         (-0.01)        (-0.48)        (-0.95)
Income growth rate in R           0.543         0.335          0.723*          0.361          0.547
                                  (1.30)        (0.77)         (1.65)         (1.01)         (1.31)
Log of R population            9.451***      8.939***        8.391***       7.997***       9.418***
                                  (5.40)        (5.18)         (5.01)         (5.40)         (5.39)
Log of aid per capita in R     9.192***      9.097***        7.274***       7.041***       9.232***
                                  (3.30)        (3.31)         (2.99)         (3.28)         (3.32)
Civil liberties in R           0.574***      0.497***        0.510***       0.411***       0.570***
                                  (3.67)        (3.65)         (3.46)         (2.88)         (2.72)
Control of corruption in R      8.672**       6.947**         7.455**         5.899*        8.910**
                                  (2.40)        (2.05)         (2.20)         (1.78)         (2.50)
Fragile state                                -11.217**
                                               (-2.30)
HIPC completion                                               6.456
                                                              (1.41)
HIPC decision                                                 1.807
                                                              (0.28)
Colonial tie                                                  2.627
                                                              (0.85)
Direct budget support                                                        0.554***
(share of aid)                                                                (11.55)
Control of corruption x                                                                      -0.248*
Donor aid share                                                                               (1.72)
No. of observations                2110           2110           2110           2079           2110
No. of Donors, Recipients         34, 77         34, 77         34, 77         34, 77         34, 77
  2
R                                  0.31           0.31           0.31           0.40           0.31
  Dependent variable is percentage of donor aid to government managed by recipient
  government‟s public financial management systems. T-statistics, reported in parentheses below
  point estimates, are based on standard errors adjusted for non-independence of errors within both
  donor and recipient clusters of observations, with *** p<0.01, ** p<0.05, * p<0.1.




                                                 37
                                              Table 3
Equation                         3.1            3.2             3.3            3.4            3.5
Fixed effects                 D-Y pairs      D-R pairs     D-Y pairs      D-Y pairs      D-Y pairs
Donor sample                     All            All         bilaterals   multilaterals      All
D share of aid in R (%)       0.705***       0.455***       1.119***        0.344**       1.102***
                                (3.45)         (4.32)         (3.28)         (2.52)          (3.39)
D share x multilateral                                                                    -0.737**
                                                                                            (-2.14)
Log of R per capita             -1.899        27.336*         -2.704          -1.691        -2.279
Income                          (-0.96)         (1.85)        (-1.10)         (-1.05)       (-1.20)
Income growth rate in R          0.577         16.942         33.144         78.766*        51.958
                                 (1.35)         (0.41)         (0.73)          (1.90)        (1.23)
Log of R population           9.731***        75.629*       10.970***       7.584***      9.365***
                                 (5.38)         (1.73)         (4.16)          (4.86)        (5.21)
Log of aid per capita in R    9.485***        9.582**       11.430***        5.504**      8.838***
                                 (3.33)         (2.34)         (2.83)          (2.51)        (3.14)
Civil liberties in R          0.592***         0.756*        0.629***       0.476***      0.561***
                                 (3.70)         (1.73)         (3.20)          (3.21)        (3.57)
Control of corruption in R     8.955**         -0.005        8.202***      10.020***       8.926**
                                 (2.40)        (-0.85)         (1.87)          (2.57)        (2.44)
Control of corruption x        1.062**
governance selectivity           (2.22)
Year=2007                                        -1.258
                                                (-1.37)
Year=2005                                        -0.261
                                                (-0.15)
No. of observations               2049            2110           1274          836            2110
No. of Donors, Recipients        33, 77          34, 77         24, 75        10, 77         34, 77
  2
R                                 0.30            0.78           0.31          0.31           0.31
  Dependent variable is percentage of donor aid to government managed by recipient
  government‟s public financial management systems. T-statistics, reported in parentheses below
  point estimates, are based on standard errors adjusted for non-independence of errors within both
  donor and recipient clusters of observations, with *** p<0.01, ** p<0.05, * p<0.1.




                                                 38
                                              Table 4
Equation                         4.1           4.2           4.3             4.4            4.5
Method                         R-Y pairs     R-Y pairs     R-Y pairs       R-Y pairs      R-Y pairs
D share of aid in R (%)          0.423**   0.479**          0.770***       1.059***        0.381*
                                   (2.03)    (2.55)           (2.77)         (4.07)        (1.89)
Multilateral devp. bank        15.866*** 26.607***                                       16.108***
                                   (2.88)    (5.50)                                        (2.96)
Vertical fund                      8.373  40.131***                                         8.162
                                   (0.89)   (10.16)                                        (0.87)
Other multilateral (UN, EC)       -1.883      6.353
                                  (-0.24)    (1.10)
Non-DAC bilateral              -15.132***    -4.556          17.911          -5.329      -0.127***
                                  (-3.75)   (-1.18)           (1.26)        (-0.45)        (-4.06)
Nordic Plus                    15.135*** 18.422***         15.209***      15.241***      0.131***
                                   (3.24)    (3.73)           (2.70)         (3.52)         (2.97)
Infrastructure share                      -0.271***         -0.331*          -0.228
                                            (-2.61)          (-2.12)        (-1.13)
Health share                              -0.708***          -0.263          -0.172
                                           (-11.12)          (-0.41)        (-0.29)
Education share                            0.730**            0.371          0.236
                                             (2.38)           (1.44)         (0.78)
Humanitarian share                          0.740*            0.683        1.142***
                                             (1.83)           (1.24)         (3.31)
Governance share                          -0.600***          -0.242         -0.447*
                                            (-3.20)          (-0.95)        (-1.77)
Tech. assistance share                                        0.176         0.232*
                                                              (1.44)         (1.81)
Multilateral contribution                                   0.432**          0.082
share of aid by bilateral                                     (2.29)        (0.056)
Tied aid                                                    -0.289*          0.013
                                                             (-1.86)         (0.12)
USA                                                                       -37.230***
                                                                            (-6.21)
UN                                                                                       -11.520***
                                                                                            (-3.37)
EC                                                                                         8.463**
                                                                                             (2.16)
No. of observations                 2213         2132           1307          1307            2213
No. of Donors, Recipients          34, 80       33, 80         23, 78        23, 78         34, 80
  2
R                                   0.30         0.36           0.39          0.41            0.31
  Dependent variable is percentage of donor aid to government managed by recipient
  government‟s public financial management systems. T-statistics, reported in parentheses below
  point estimates, are based on standard errors adjusted for non-independence of errors within both
  donor and recipient clusters of observations, with *** p<0.01, ** p<0.05, * p<0.1.

                                                 39
                                                Table 5
Equation                            5.1                   5.2          5.3               5.4
Fixed effects                    R-Y pairs          R-Y pairs       R-Y pairs        R-Y pairs
                                                                    World Values Eurobarometer
Survey                         Index of all 3      Gallup 2002
                                                                      1995-98              2004
D share of aid in R (%)             0.619**          0.604**            0.529           1.422***
                                     (2.07)           (2.02)            (1.34)            (3.42)
Nordic Plus                       13.103***         16.573***           8.138          11.106***
                                     (3.01)           (3.45)            (1.31)            (2.91)
Public support for aid in D        6.897***          0.442**          0.835***            0.091
                                     (2.70)           (2.05)            (3.05)            (0.53)
No. of observations                   1289             1089               639               795
No. of Donors, Recipients            22,78            17, 77             9, 77            15, 71
R2                                    0.38             0.39              0.39              0.48
 Dependent variable is percentage of donor aid to government managed by recipient
 government‟s public financial management systems. T-statistics, reported in parentheses below
 point estimates, are based on standard errors adjusted for non-independence of errors within both
 donor and recipient clusters of observations, with *** p<0.01, ** p<0.05, * p<0.1.
                                               Table 6
Equation                                 6.1                6.2            6.3              6.4
Method                                   OLS               OLS            2SLS          first stage
Fixed effects                         R-Y pairs          R-Y pairs      R-Y pairs       R-Y pairs
D share of aid in R (%)                0.797***           1.046***      0.660***         -.016***
                                          (2.82)            (3.84)        (4.01)          (-5.26)
Nordic Plus                           15.780***          15.392***     13.287***        -0.303***
                                          (2.69)            (3.29)        (5.87)          (-5.76)
Public support for aid in D            5.270***             2.768*      6.738***
                                          (3.31)            (1.94)        (3.13)
Infrastructure share                     -0.068             -0.098
                                         (-0.39)            (-0.45)
Health share                             -0.186             -0.170
                                         (-0.30)            (-0.28)
Education share                         0.419**              0.288
                                          (1.58)            (0.97)
Humanitarian share                        0.772           1.146***
                                          (1.53)            (3.49)
Governance share                         -0.348            -0.469*
                                         (-1.33)            (-1.91)
Tech. assistance share                   0.219*            0.234**
                                          (1.94)            (2.00)
Multilateral contribution                 0.268              0.024
Share                                     (1.56)            (0.15)
USA                                                     -33.906***
                                                            (-4.96)
Colonial tie                                                             -5.765          0.449***
                                                                         (-1.48)           (4.62)
Log of D population 1995                                                                -0.299***
                                                                                          (-12.45)
No. of former colonies                                                                   0.025***
                                                                                          (18.63)
No. of observations                        1268            1268              1289           1289
No. of Donors, Recipients                 22, 78          22, 78            22, 78         22, 78
  2
R                                          0.40            0.42              0.09           0.41
Overid test p value                                                           .30
Angrist-Pischke weak-ID p value                                                             < .0001
F test of excluded instruments                                                               176.8
F test p value                                                                              < .0001
  Dependent variable in columns 1-3 is percentage of donor aid to government managed by
  recipient government‟s public financial management systems. Dependent variable in column 4
  is index of public support for aid in donor country. T-statistics, reported in parentheses below
  point estimates, are based on standard errors adjusted for non-independence of errors within both
  donor and recipient clusters of observations in equations 1 and 2, and within donor-recipient pair
  clusters in equations 3 and 4, with *** p<0.01, ** p<0.05, * p<0.1.




                                                   41
                                           Appendix

                              Paris Declaration Monitoring Survey
                                      Indicator Definitions



Use of country PFM systems (PD indicator 5a): Simple average of following three sub-
indicators, as a percentage of all aid for the public sector.

Use of national budget execution procedures

Three of these four criteria must be met to qualify:
   1) Funds are included in the annual budget approved by the legislature;
   2) Funds are subject to established country procedures for authorization, approval and
       payment of funds;
   3) Funds are deposited and disbursed through the established treasury system;
   4) Opening of separate bank accounts for donor funds is not required

Use of national financial reporting procedures

Both criteria must be met to qualify:
   1) No separate accounting system is required to satisfy donor‟s reporting needs;
   2) No separate chart of accounts is required to record the use of donor funds

Use of national auditing procedures

To qualify, funds are subject to audits by the country‟s Supreme Audit Institution using its
auditing cycle and standards, and additional auditing arrangements are not requested in normal
circumstances.

Direct budget support

To qualify, funds must be transferred to the national treasury and be managed using national
budgetary procedures, and not earmarked for specific uses. They may be nominally earmarked
for a broadly-defined sector such as education (sector budget support).




                                               42
                                           Table A1
                                        Summary statistics
                                                  mean Std. dev.   Min.    Max.      N
Use of PFM systems (%)                             33.3    35.0      0      100    2213
D share of aid in R (%)                            6.99    9.13  0.00001    83.6   2213
Log of R per capita income                         6.31    0.90    4.48     8.37   2110
Income growth rate in R                            2.95    2.70   -7.43     9.88   2110
Log of R population                               16.55    1.37   11.55    19.30   2110
Log of aid per capita in R                         3.82    0.86    1.46     6.08   2110
Civil liberties in R                               32.0    10.4      6       53    2110
Control of corruption in R                        -0.63    0.43   -1.43     1.05   2110
Multilateral development bank                      0.14    0.35      0        1    2213
Vertical (global) fund                             0.11    0.31      0        1    2213
Other multilaterals                                0.14    0.35      0        1    2213
Non-DAC bilateral donors                           0.03    0.16      0        1    2213
“Nordic plus� donors                               0.18    0.39      0        1    2213
Public support for aid (index)                    -0.15    0.85   -1.49     2.00   1289
Public support for aid (Gallup)                    62.7    13.6     45       91    1089
Public support for aid (World Values)              17.4     9.7     6.7     35.5     639
Public support for aid (Eurobarometer)             33.1     8.5     19       58     795
Direct budget support                              12.4    22.8      0      100    2079
Infrastructure (% of D aid)                        13.7    13.4      0      63.1   2132
Education (% of D aid)                              8.2     6.1      0      34.7   2132
Health (% of D aid)                                 9.6    15.6      0      89.3   2132
Humanitarian (% of D aid)                           5.7     5.2      0      28.9   2132
Government/civil society (% of D aid)              10.4     7.3      0      38.8   2132
Technical cooperation (% of D aid)                 17.1    11.9     1.4     47.4   1307
Multilateral contributions (% of D aid)            32.4    11.5    10.1     73.5   1307
Tied aid (% of D aid)                              12.8    16.9      0      98.1   1307




                                           43
                                           Table A2
                       Public Support for Foreign Aid in Donor Countries



survey           WVS 1995-8          Gallup International       Eurobarometer 2004        Index
                                             2002
donor           "very much for"     should “give more aid�       aid share of budget
                      aid                                            “too small�
Australia           17.7%                    63%                           .               -0.28
Austria                .                       .                        21%                -1.17
Belgium                .                       .                        36%                 0.51
Canada                 .                     64%                           .               -0.36
Switzerland            .                     71%                           .                0.12
Germany             18.2%                    68%                        25%                -0.29
Denmark                .                     53%                        35%                -0.29
Spain               33.1%                    91%                        58%                 2.00
Finland             15.2%                    61%                        31%                -0.34
France                 .                       .                        35%                 0.39
UK                     .                     69%                        38%                 0.36
Greece                 .                       .                        30%                -0.16
Ireland                .                     87%                        30%                 0.53
Italy                  .                     83%                        30%                 0.39
Japan                9.8%                    45%                           .               -1.32
Korea                  .                     87%                           .                1.22
Luxembourg             .                     71%                        19%                -0.64
Netherlands            .                     50%                        33%                -0.58
Norway              21.9%                    63%                           .               -0.05
New                 11.9%                      .                           .               -0.75
Zealand
Portugal                 .                   89%                       22%                  0.15
Sweden                35.5%                  69%                       29%                  0.50
Turkey                   .                   87%                          .                 1.22
USA                    6.7%                  45%                          .                -1.49
Gallup: Do you think that the wealthier nations should give more financial help to the poorer
nations or are they giving enough now? (Should give more/are giving enough now)
WVS: Some people favor, and others are against, having this country provide economic aid to
poorer countries. Are you personally: very much for/for to some extent/somewhat against/very
much against
Eurobarometer: Do you think that the share of its budget that the [respondent‟s country]
Government dedicates to development aid is: too big/too small/about right




                                               44
                                          Table A3
                                     Sources of Censoring

Row #   Year                         2005         2007      2010   Total
1       Donors                       34           34        34
2       Recipients                   34           55        76
3       DAC recipients               151          151       151
4       D-R pairs in survey          1156         1870      2584   5610
5       D-DAC recipient pairs        5134         5134      5134   15402
6       D-R active pairs in survey   473          730       998    2201
7       (4) as % of (5)              22.5         36.4      50.3   36.4
9       (6) as % of (4)              40.9         39.0      38.6   39.2
10      (6) as % of (5)              9.2          14.2      19.4   14.3




                                             45
                                             Table A4
                                   selection regressions (probit)
Equation                       5.1          5.2            5.3              5.4
Years                          All           2005            2007           2010
Multilateral devp. bank         0.004         0.007            0.002         0.028
                                (0.04)        (0.07)           (0.02)        (0.14)
Vertical fund                   0.056         0.016            0.033         0.158
                                (0.44)        (0.14)           (0.25)        (0.69)
Other multilateral (UN, EC) 0.233             0.135            0.228         0.362
                                (1.36)        (0.78)           (1.18)        (1.39)
Non-DAC bilateral               -0.065*** -0.048***            -0.067***     -0.066***
                                (-5.48)       (-4.43)          (-4.36)       (-3.66)
Nordic Plus                     -0.015        0.009            -0.014        -0.050
                                (-1.44)       (1.35)           (-1.25)       (-3.08)
Log of R population             0.034***      0.021            0.039***      0.041***
                                (5.19)        (3.71)           (4.75)        (4.84)
Colonial tie                    0.203***      0.107***         0.206***      0.285***
                                (8.89)        (5.77)           (7.73)        (8.32)
Distance from capital city      -0.031*       -0.013           -0.039**      -0.045**
(bilateral)                     (-1.91)       (-0.88)          (-2.34)       (-1.97)
Distance from donor HQ          -0.031*** -0.013*              -0.039***     -0.046**
city (multilateral)             (-2.70)       (-1.76)          (-2.92)       (-2.27)
HIPC completion                 0.264***      0.309***         0.276***      0.227***
                                (8.61)        (7.95)           (8.12)        (5.96)
HIPC decision point             0.114**       0.083*           0.214***      -0.002
                                (2.48)        (1.93)           (3.70)        (-0.03)
Year=2007                       -0.039***
                                (-3.63)
Year=2005                       -0.085***
                                (-6.22)
No. of observations             15402         5134             5134          5134
No. of Donors, Recipients       34, 151       34, 151          34, 151       34, 151
Pseudo R2                       0.21          0.26             0.23          0.16
  Dependent variable is selection, coded 1 for donor-recipient pairs if positive aid flows are
  recorded in the PDMS and 0 otherwise. Coefficients are marginal effects, evaluated at the means
  of all other independent variables. T-statistics, reported in parentheses below point estimates, are
  based on standard errors adjusted for non-independence of errors within recipient clusters of
  observations, with *** p<0.01, ** p<0.05, * p<0.1.




                                                  46