92287 Implementing Enterprise Surveys in Latin America and the Caribbean1 REV. 8/2014 Methodology of the Enterprise Surveys 3. A consistent definition of the survey’s sampling universe of inference in every country ensures Beginning in 2005, the World Bank’s Private Sector data comparability. This universe consists of the Development Vice Presidency undertook an evaluation nonagricultural, formal, private sector economy. Based of its Enterprise Surveys program and decided to shift to on the International Industrial Classification (revision CARIBBEAN SERIES NOTE NO. 10 a global survey approach from a country level approach. 3.1), it includes the following sectors: The global approach would enable the institution to both š Manufacturing (D). study regional business environments and substantiate international comparisons. To date, over 65,000 firms š Construction (F). in 119 countries have been surveyed using this global š Wholesale and retail trade and repair services (G). methodology. The result is full comparability of data across š Hotels and restaurants (H). countries and regions.2 š Transport, storage, and communications (I). The World Bank has conducted many survey type studies š A subsector of computer and related activities (72) of the private sector, dating back to the 1980s. However, most of these studies focused on specific issues in one Only firms with five or more employees are included, country. In many instances, surveys aimed to address and firms with 100 percent state ownership are only specific, short-term goals, and were therefore limited excluded. Only registered firms are included, and in scope, for instance studying only a few manufacturing registration is defined as registration for taxation sectors or only large firms. These short-term goals AND THE purposes. limited the uses of the information and the possibility of comparative analysis with other surveys. 4. In each country the sample design of the Enterprise Surveys follows a stratified, random sampling strategy The global systematic approach of the World Bank in all countries, where the sample of firms is divided LATIN AMERICA Enterprise Survey program has now addressed these into L mutually exclusive and exhaustive strata. The issues, providing business environment indicators that are L strata are based on three elements: geographical comparable across countries. This approach is based on location, firm size, and sector of activity (Table 1). the following principles: Stratification supports analyses at each level, and increases the precision of overall estimates. 1. The Enterprise Analysis Unit (EAU) implements surveys worldwide to provide the public good of information. The Unit’s findings are made public, Sample Design while the anonymity of respondents is preserved. The To enable statistically robust analyses, the optimal target audience is World Bank Group staff, World Bank sample size for each level of stratification is defined to clients, researchers, and the general public. However, achieve a minimum level of precision of 7.5 percent for 90 the general public has access to aggregate indicators percent confidence intervals for estimates of population WORLD BANK GROUP only. All information collected is released as soon as its proportions (percentages) and estimates of the mean of quality and accuracy have been certified by the EAU. the logarithm of sales. The optimal sample size is 120 firms 2. Regional implementation of the survey guarantees for each level of stratification.3 homogeneity of implementation, indicators, and methodology, as well as economies of scale in Stratification by firm size divides the population into three management. The result is full comparability of data strata: small (5–19 employees), medium-size (20–99 across countries and regions. employees), and large firms (100 or more employees). 1 Geographical distribution is defined to reflect the efforts were made to re-contact enterprises there as well. distribution of nonagricultural economic activity in each In all, 3,356 of the firms surveyed (including in Brazil and country. For most countries, this implies including areas Costa Rica) were panel firms. typically clustered in and around main urban centers. Stratification by sector of activity is based on the size of the Representativeness of the ES Samples economy, as measured by Gross National Income (GNI) in The sampling methodology of the Enterprise Surveys is 2008. Very small countries (less than $15 billion GNI) are designed to generate samples suitable for achieving two stratified into two groups: manufacturing and the rest of main objectives: to benchmark the investment climate of the nonagricultural economy. Small countries ($15 billion individual economies around the developing world; and to to $100 billion) are stratified into retail, manufacturing, analyze firm performance and derive data to determine and the rest of the nonagricultural economy. For medium- how the business environment affects productivity and job size countries ($100 billion to $500 billion), the four creation. most important manufacturing industries are grouped into a stratum, the remaining manufacturing industries To achieve these objectives, representative samples are grouped into another, and retail and the rest of the are needed that can substantiate inferences for the nonagricultural economy compose the final two strata. nonagricultural, formal, private sector. Two major Finally, for large countries (GNI greater than $500 billion) impediments are generally associated with generating six or more manufacturing industries are selected as representative samples of a target population: inaccurate individual strata, each one on its own; the remaining sample design and non-response bias. Inaccurate sample manufacturing industries are grouped into a residual design occurs when samples are designed using either stratum. Among services, retail and other selected sectors inappropriate sampling methodology or incomplete (such as IT and wholesale) are singled out as strata on information about the target population. Non-response their own, and the remaining services sectors are grouped bias can take the form of item non-response or survey into a residual “other services” stratum. non-response: the former occurs when sampled firms do not answer specific questions of the survey; the latter A key goal of the Enterprise Surveys is to collect panel data when firms fail to respond to the survey altogether. by interviewing the same enterprises in sequential rounds. Whenever a previous round of an Enterprise Survey is Incomplete information about the target population available, focus is placed on collecting panel data. To can result in inaccurate sample design. To avoid this ensure the new sample is also representative of the problem and ensure sample representativeness, a critical current economy, the presence of panel firms is capped at requirement for the Enterprise Surveys is the availability 50% of any given cell. Consequently, each cell is designed of a list that identifies every firm, establishment, and to be one-half panel and one-half fresh (previously economic unit within the target population—the sample non-interviewed) firms. However, the full sample of the frame. Accurate sample frames ensure that each individual first wave of the survey is contacted to determine whether economic unit can be identified unambiguously. each firm is still in operation or if they exited the market, were acquired by other firm, merged with other firm, etc. Given the importance of sample frames, a careful Within each cell, panel firms that are actually interviewed evaluation of potential lists is made at the start of each are randomly selected. survey. Sample frames are evaluated primarily for their completeness with respect to industry. In addition, In the 2010 round of Enterprise Surveys in Latin America the frames must be less than three years old, contain and the Caribbean, 15 of the countries had a previous information about the number of employees, and include round in 2006—Argentina, Bolivia, Chile, Colombia, current contact information. Available sample frames may Ecuador, El Salvador, Guatemala, Honduras, Mexico, contain information that does not match perfectly with Nicaragua, Panama, Paraguay, Peru, Republica Bolivariana Enterprise Survey definitions, requiring adjustments to the de Venezuela, and Uruguay. Though prior survey rounds in standard methodology. Brazil (2003) and Costa Rica (2005) were not implemented under the global methodology of the Enterprise Surveys, 2 IMPLEMENTING ENTERPRISE SURVEYS 10 Methodology Sample frames are often incomplete or outdated, and size, and location. Potential substitutions are selected missing information must be completed in order to simultaneously with the original sample at the onset of locate establishments and confirm their characteristics. the project to ensure the same probability of selection. The suitability of a sample frame is determined through Substitutions can be made only for well-justified reasons: a careful validation process. Sometimes, particularly establishments that are determined to be out of business, in very small countries, the validation process yields have changed their line of business such that they are no information that requires adjustments to universe figures. longer in the target population, cannot be located, cannot In Barbados, for instance, the validation process confirmed be motivated to finalize the appointment after several that more firms existed in a cell than previously reported sincere attempts, or that simply refuse to participate. in official numbers. In such cases, universe estimates are Substitutions are made only with World Bank approval. adjusted accordingly, subject to later refinement during the interview process. In many cases, identifying and Several measures are used in all countries to enhance obtaining a valid sample frame is a joint effort between participation and minimize non-response bias. A the World Bank and the survey contractor. Such efforts personalized invitation letter is sent to each selected may include consultation with local contacts, additional firm a few weeks before the implementing contractor research, cooperation with the country’s statistical agency contacts them. The letter aims to motivate participation and chambers of commerce, and visits by the World Bank by describing the goals of the survey, providing the name Enterprise Surveys team. of the implementing agency hired for the study, and reassuring potential respondents about the confidentiality To reduce item non-response, a strict quality control of the information collected. Also, potential respondents process identifies non-responses and contacts firms are offered a summary report of the main findings from to complete the data. Item non-response issues are previous rounds of the survey in their country. Finally, to especially critical when it comes to accounting data, thank respondents for their participation, they receive a such as sales and costs, used to assess firm performance. brief snapshot of their country, the Country Profile, based To account for item non-response, when possible, the on the newly collected survey data. sampling strategy factors-in a 25 percent non-response rate per stratum, so that there are enough valid responses In every country, implementing contractors, along with the to compute performance indicators with the required Enterprise Surveys team, are required to work with local precision level. This implies a total of 160 interviews per business associations or similar institutions to ensure stratum. However, due to budget constraints and the fact broad-based participation in the survey. In a few countries, that only medium-size and large countries have enough a media campaign is also used to encourage participation. observations to complete 160 interviews per industry, this sample size adjustment is only implemented for those Questionnaire countries. A major challenge in implementing a global survey Another method to reduce item non-response is the use of is dealing with business environment peculiarities. a web application. Interviews with missing financial data Homogeneous concepts and definitions are necessary to are identified and the respondents are invited to provide enable global comparisons, and the Enterprise Surveys the information through a private website. This method, therefore use a homogeneous, global questionnaire across used in 15 countries surveyed in the 2010 wave, resulted all regions. Yet, this approach poses challenges when in a 1 percent increase in responses for sales data and a dealing with the characteristics of each country’s business 3 percent increase in the number of firms for which total environment — challenges that must be considered factor productivity could be computed. when implementing the questionnaire. For example, in some countries, the banking system may offer particular Bias resulting from survey non-response is minimized by forms of loans or accounts that may not be listed among strict controls on the original sample selection and by the possible answers to global finance questions. These using firm substitutions. For each sampled establishment, issues can be managed by working closely with the local substitute firms are selected, by stratum, with exactly contractors in charge of data collection to make sure that all the same characteristics in terms of sector of activity, key concepts are understood correctly by the interviewees. IN LATIN AMERICA & THE CARIBBEAN 3 Despite the drive to achieve homogeneity, a minimum š Section J – Business-Government Relations: quality of level of tailoring of the questionnaire is nonetheless of public services, consistency of policy, regulatory desirable to reflect certain issues that are specific to each compliance costs (management time, bribes). region or country. Indeed, one of the major challenges š Section L – Labor: employment, training, and skills. from a global perspective is the tradeoff between global comparability and coverage of issues at the regional or š Section M – Business Environment: ranking of general country level. This challenge is met by allocating about obstacles. one-fourth of the questionnaire to reflect characteristics of š Section N – Performance: numbers and figures needed the region and country. to estimate performance or productivity. A final challenge in designing a multicultural, firm-level It should be noted that section F, Capacity, is only survey is accounting for local sensibilities. For example, in utilized for manufacturing firms. Additionally, section some countries discussing gender issues in management or P, Business Development Services, is included only in the employment may be inappropriate, in others, questions about Latin America and Caribbean questionnaire. corruption or the use of bribes can be highly controversial. There are also differences in attitudes toward questions The Enterprise Survey questionnaire is designed to be regarding the efficiency of government or the role it should administered in face-to-face interviews with managers, play, and in the willingness of firms to reveal financial results. CEOs, or firm owners. The establishment is the primary All of these challenges may limit the scope of issues that can sampling unit. The Enterprise Surveys are implemented in be tackled globally and were considered when designing a two stages. In the first, a screener questionnaire is given to homogeneous, global questionnaire. determine if an establishment is eligible. The screener is designed to be administered over the phone, but alternative The Enterprise Surveys questionnaire calls for a variety strategies can be used when phone communication is not of qualitative and quantitative information, organized by feasible. The second stage is the face-to-face interview. topic into 14 sections: Special attention is paid to the length of the questionnaire. š Section A – Control Information: information collected in Experience shows that the quality of information collected the first stage of implementation. is inversely correlated to the length of the questionnaire. š Section B – General Information: characteristics of the Consequently, interviews are designed to last no longer establishment. than one hour. š Section C – Infrastructure and Services: power, water, Accurate translation of the questionnaire and its manual transport, and communication technologies. into the local language is critically important. In addition š Section D – Sales and Supplies: imports, exports, and to using back-translations into English, translations are supply and demand conditions. checked by the contractor and World Bank staff whenever possible. Lawyers and accountants are consulted to š Section E – Degree of Competition and Innovation: preserve the original English meaning of legal terms and number of competitors and innovation activities and accounting definitions. A pilot study is also used to test for R&D. language errors, and to time how long it takes respondents š Section F – Capacity: use of production capacity, hours to answer the questionnaire. The questionnaire is finalized of operation. using the information obtained from the pilot study. š Section G – Land and Permits: land ownership, land The structure and format of the questionnaire are critical access issues. to ensuring that the information collected accurately š Section I – Crime: extent of crime and losses due to reflects the views and opinions of the participants. The crime. questionnaire is designed to flow in a logical order, š Section K – Finance: sources of finance, access to credit. with questions grouped by topic and similar questions positioned across sections in a consistent manner. For š Section P – Business Development Services: establishment’s example, questions about the degree of obstacles faced are demand and use of selected business services. always asked at the end of each topic section. In addition, the same language and structure are used for similar 4 IMPLEMENTING ENTERPRISE SURVEYS 10 Methodology questions. For example, in the infrastructure section, the levels, costs, and the replacement value of fixed assets. same set of questions is asked about power outages as These tests identify values that are more than three about water shortages—specifically, the number, duration, standard deviations from the country average. In a number and losses due to these phenomena. In addition, skip of cases, certain variables are assumed to be distributed patterns are widely used throughout the questionnaire to log-normally, and outlier tests are run on the distribution of avoid unnecessary questions. the logged values.4 Further tests are conducted using the distribution of performance ratios (for example, the ratio of Special attention is paid to the enumerators’ handling of annual sales to annual cost of employment) and analyzing sensitive questions when administering the survey. As with the distribution of key variables at the firm level within a any survey, enumerators must refrain from trying to direct country. Identified outliers are presented to contractors, who the answers in order to avoid introducing interpretation are requested to review the questionnaires to determine if a bias. The questionnaire visually identifies any question data entry error occurred or if a callback is necessary. that is sensitive in nature or that requests the respondent’s opinion. Enumerators are trained to read these questions exactly as they are written, without explanations. When Panel Information asked for the meaning of any term or when faced with a Difficulties in obtaining panel data arise when firms respondent who is confused, enumerators should read cease operations, move, change contact information, the full question again, as it is written. They should not or even change sector of activity. In order to record this attempt to change the wording or rephrase the question information, all firms included in the first wave of the survey using synonyms or commonly used expressions. The rest are re-contacted to determine their current status and of the questions are hard-data questions, designed to eligibility. In 2010, 56 percent of firms interviewed in 2006 obtain objective facts and quantitative data. For this type of were confirmed to have remained in operations. Only 4 question, enumerators are allowed to probe the respondent percent were confirmed to have ceased operations, while using standard techniques, such as emphasizing the 19 percent could not be reached.5 Contact information key concept or asking a follow-up question in order to for 4 percent of firms yielded responses indicating the complete an incomplete answer. enterprises were no longer in the Enterprise Surveys universe, while 17 percent of firms previously interviewed Quality Control refused to confirm information on their current operations. The Enterprise Surveys implementation process employs rigorous quality controls. Data are received from local Weighting Methodology contractors in three batches—at 10 percent completion, Whenever stratification is used as a sampling strategy, 50 percent completion, and 100 percent completion. observations must be weighted in order to make inferences After each of these deliveries, a series of tests detects to the population or subsectors of the population. The any patterns or values that violate logical assumptions, weights correct for the different probabilities of selection identifies invalid answers, and flags the inconsistent use introduced by the stratification. In countries that did of skip patterns. No responses are changed during this not contain panel firms,6 the standard Enterprise Survey process, but contractors are requested to review flagged weights were used. In countries with panel firms, values and identify the problem. In certain cases, this adjustments for the priority of selection given to panel process requires re-contacting a respondent by phone or firms were made for both fresh and panel firms. in person. All weights are calculated for each combination of the After the 50 percent delivery, simple response rates for three levels of stratification: firm size, location, and sector key performance variables are calculated. If these response of activity. Base weights are given by the universe figure rates are determined to be low, contractors are asked to (N) divided by the number of realized interviews in that cell re-contact firms. At this stage, the preliminary distribution (n). Base weights assume that cell-level universe figures are of completed interviews is also analyzed against the initial completely accurate. However, during fieldwork, evidence survey design. At the completion of fieldwork, an additional generally arises that suggests the universe numbers may analysis is used to flag outlier values. Outlier tests are run be incorrect—for example, it is common to find firms in the on firm performance variables, including sales, employment IN LATIN AMERICA & THE CARIBBEAN 5 sample frame that are ineligible. As a result, universe numbers Implementation of the Global Methodology in must be adjusted, and this adjustment is given by the number the LAC Region of establishments confirmed eligible as a proportion of the number of contacts issued. The adjustment factor is then The Latin America and the Caribbean (LAC) region is the multiplied by the base weight to provide the sampling weight. first region to be surveyed using the global methodology over multiple periods. In 2006, firms in 15 of the region’s For countries where panel firms were part of the fieldwork, countries were surveyed using this methodology. In adjustments are made for the varying probabilities of 2010, 14,657 firms were surveyed in 31 countries, 3,535 selection for both panel and fresh firms. The calculation of of which had been interviewed previously in 2006. Thus, weights for fresh firms follows the straightforward process LAC is the first region to have a regional report dedicated described above. However, the universe figures must be to analyzing what businesses are experiencing over time adjusted for the presence of panel firms. For fresh firms, in using Enterprise Surveys data. these cases, universe figures in individual cells are given by removing the number of firms interviewed in the previous The 2010 Enterprise Surveys for Latin America and the round from the initial universe estimate (Nfresh = N − Npanel ). Caribbean covers 31 countries (Table 1). Fieldwork took place in 2010 and 2011, except for Brazil, where the survey The calculation of panel firm weights likewise follows was implemented in 2009. For all countries, the period of the standard weighting methodology. For these firms, reference for most of the questions is the last completed the universe estimate is given by the number of firms in fiscal year and therefore the 2010 ES refers to fiscal year a cell based on answers provided in the previous round. 2009. For the 15 countries included in the 2006 ES wave, Base weights are calculated using these universe figures. the period of reference is fiscal year 2005. In the case of Since the Enterprise Surveys record the eligibility status of Brazil, the survey’s period of reference is fiscal year 2008. all firms previously interviewed, not just those that were randomly selected, this information is incorporated in the calculation of an adjustment factor. The adjustment factor for panel firms is the proportion of eligible panel firms among all those contacts available (among all panel firms, not simply those selected for interviews in the current round). Finally, an adjustment is made to account for the initial weight of panel firms in the first round. This panel adjustment therefore links the current weight to the probability of firms entering the sample in the first round. This adjustment is given by the average of the first-round weights of all firms in a cell according to answers given in the first wave. For example, a panel firm that is drawn in 2010 as a medium manufacturer in a certain city receives a panel adjustment of the average 2006 weight of all firms that were classified as medium manufacturers in that city (based on their 2006 answers). For further information on this methodology, see the Sampling Note at http://www. enterprisesurveys.org. 6 IMPLEMENTING ENTERPRISE SURVEYS 10 Methodology TABLE 1 ENTERPRISE SURVEYS IN LATIN AMERICA AND THE CARIBBEAN: NUMBER OF FIRMS INTERVIEWED AND LEVELS OF STRATIFICATION BY COUNTRY7 Stratification Level Firm Size Sector of Activity Location Country Small Medium Large Total Brazil 678 750 374 1,802 Food, textiles, apparel, leather, Amazonas, Bahia, chemicals, machinery & equipment, Brasilia DF, Ceara, vehicles, furniture, other Goias, Maranhão, manufacturing, retail, IT, other services Mato Grosso, Minas Gerais, Paraiba, Parana, Pernambuco, Rio LARGE Grande do Sul, Rio de Janeiro, Santa Catarina, São Paulo Mexico 502 472 506 1,480 Food, apparel, chemicals, rubber & Coahuila, Estado de plastics, fabricated metal, machinery Mexico, Guanajuato, & equipment, furniture, other Jalisco, Mexico DF, manufacturing, retail, IT, other services Nuevo Leon, Puebla, Veracruz Argentina 336 395 323 1,054 Food, textiles & apparel, chemicals Buenos Aires, Chaco, & rubbers/plastics, basic metals/ Cordoba, Mendoza, fabricated metals/machinery & Rosario equipment, other manufacturing, retail, other services Chile 303 394 336 1,033 Food, textiles & apparel, chemicals Antofagasta, Los & rubbers/plastics, basic metals/ Lagos, Santiago, fabricated metals/machinery & Valparaiso MEDIUM equipment, other manufacturing, retail, other services Colombia 349 326 267 942 Food, textiles & apparel, chemicals Barranquilla, Bogota, & rubbers/plastics, basic metals/ Cali, Medellin fabricated metals/machinery & equipment, other manufacturing, retail, other services Peru 318 379 303 1,000 Food, textiles & apparel, chemicals Arequipa, Chiclayo, & rubbers/plastics, basic metals/ Lima, Trujillo fabricated metals/machinery & equipment, other manufacturing, retail, other services Bolivia 124 145 93 362 Manufacturing, retail, other services Cochabamba, La Paz, Santa Cruz Costa Rica 7 SMALL 199 216 123 538 Food, minerals/basic metals/ San Jose, Rest of the fabricated metals/machinery & Country equipment, other manufacturing, retail, hotels & restaurants, other services Dominican 116 133 111 360 Manufacturing, retail, other services Santo Domingo, Rest Republic of the Country Continued Next Page IN LATIN AMERICA & THE CARIBBEAN 7 TABLE 1 ENTERPRISE SURVEYS IN LATIN AMERICA AND THE CARIBBEAN: NUMBER OF FIRMS INTERVIEWED AND LEVELS OF STRATIFICATION BY COUNTRY (CONTINUED) Stratification Level Firm Size Sector of Activity Location Country Small Medium Large Total Ecuador 128 139 99 366 Manufacturing, retail, other services Azuay, Guayas, Pichincha El Salvador 124 120 116 360 Manufacturing, retail, other services San Salvador, Rest of the Country Guatemala 221 185 184 590 Food, textiles & apparel, other Guatemala City, Rest manufacturing, retail, other services of the Country Honduras 182 111 67 360 Manufacturing, retail, other services San Pedro Sula, Tegucigalpa, Rest of the Country Jamaica 140 169 67 376 Manufacturing, retail, other services Kingston, Rest of the SMALL Country Nicaragua 154 124 58 336 Manufacturing, retail, other services Managua, Rest of the Country Panama 129 161 75 365 Manufacturing, retail, other services Panama City, Rest of the Country Paraguay 128 160 73 361 Manufacturing, retail, other services Asuncion, Central Trinidad and 166 103 101 370 Manufacturing, retail, other services Port of Spain, Rest of Tobago the Country Uruguay 244 208 155 607 Food, textiles & apparel, other Canelones, manufacturing, retail, other services Montevideo Venezuela, RB 160 113 47 320 Manufacturing, retail, other services Caracas, Maracay, Valencia Antigua and 94 51 6 151 Manufacturing, services Entire country Barbuda Bahamas, The 70 55 25 150 Manufacturing, services Entire country Barbados 49 62 39 150 Manufacturing, services Entire country Belize 79 61 10 150 Manufacturing, services Entire country VERY SMALL Dominica 103 43 4 150 Manufacturing, services Entire country Grenada 99 41 13 153 Manufacturing, services Entire country Guyana 51 72 42 165 Manufacturing, services Entire country St. Kitts and Nevis 82 60 8 150 Manufacturing, services Entire country St. Lucia 79 55 16 150 Manufacturing, services Entire country St. Vincent and 110 38 6 154 Manufacturing, services Entire country the Grenadines Suriname 66 77 9 152 Manufacturing, services Entire country Total Firms 5,583 5,418 3,656 14,657 Interviewed 8 IMPLEMENTING ENTERPRISE SURVEYS Analytical Groupings 10 Methodology TABLE 2 COUNTRY SIZE CLASSIFICATION For analytical purposes countries are generally categorized 2009 GNI into large, medium-size, or small Caribbean countries, Country Name (current USD) depending on the overall size of their economy8 (based on in millions 2009 GNI in current USD). The overall market size of an Brazil 1,562,412 LARGE COUNTRIES economy frequently determines important aspects of the business environment, including the scale and efficiency Mexico 864,907 of regulation, the size of private sector demand, trade Venezuela, RB 323,481 patterns, and other aspects vital to business operations. Argentina 297,639 Note that three Caribbean countries—Dominican Republic, Jamaica, and Trinidad and Tobago —are classified as Colombia 226,344 “medium-size countries,” whereas Belize, Guyana and Chile 149,316 Suriname are included in the “small Caribbean countries” Peru 119,439 group (see Table 2). Ecuador 50,638 In addition to analyzing firms by country groupings, the data is also analyzed by characteristics of interest. Table Dominican Republic 45,019 3 lists and describes firm groupings that are used for Guatemala 36,572 MEDIUM-SIZE COUNTRIES analytical purposes. Uruguay 30,651 Note that for global comparisons in the Latin America Costa Rica 28,362 and the Caribbean Series Notes, 118 countries are used in Panama 23,059 total: 31 in LAC, 39 in Africa, 11 in EAP (excluding China as the data was unavailable), 29 in ECA, 6 in SAR (excluding El Salvador 20,113 India as the data was unavailable), and 2 in MENA. No Trinidad and Tobago 18,712 MENA region averages are presented as there are only two countries. Only surveys implemented under the Bolivia 16,666 Enterprise Surveys Global Methodology are used in global Paraguay 13,954 comparisons. Honduras 13,641 Jamaica 11,933 Nicaragua 5,955 Bahamas, The9 6,941 SMALL CARIBBEAN COUNTRIES Barbados 3,517 Suriname 3,257 Guyana 2,009 Belize 1,232 Antigua and Barbuda 1,202 St. Lucia 1,057 Grenada 695 St. Vincent and the Grenadines 683 St. Kitts and Nevis 640 Dominica 452 Source: World Development Indicators IN LATIN AMERICA & THE CARIBBEAN 9 TABLE 3 CLASSIFICATIONS BY FIRM CHARACTERISTICS Group Description Firm size Firms are categorized into three size groups: small (<20 employees), medium (20-99 employees), and large (more than 100 employees). Exporter and Non-Exporter A firm is an exporter if at least 10 percent of sales are derived from direct or indirect exports. Foreign ownership and Domestic-owned A firm is considered to have foreign ownership if at least 10 percent of ownership is held by foreigners; otherwise the firm is domestically owned. City size Large cities are those with a population of at least one million; cities with populations of less than one million are considered small. Women-run A firm is “Women-run” if the top manager is female. Women-owned A firm is “Women-owned” if at least one of the owners is female. Sector Sectors are grouped by industry stratifications (See Table 1.) Endnotes 6. Antigua and Barbuda, The Bahamas, Barbados, Belize, Brazil, Costa Rica, Dominica, the Dominican Republic, Grenada, Guyana, Jamaica, St. Lucia, St. Kitts and Nevis, St. Vincent and the 1. Lead authors: Rita Ramalho, Jorge Luis Rodriquez Meza, and Grenadines, Suriname, and Trinidad and Tobago. Federica Saliola, with the collaboration of the LAC report team. 7. Firms are categorized according to the reported numbers of full- 2. The ten notes in the Latin America and Caribbean Region series time equivalent employment, taking into account permanent and use only Enterprise Surveys data collected under the global temporary employees. methodology. Please note that additional (older) data, not collected under the global methodology, are also hosted in the 8. In certain cases, in response to various requests, sample sizes Enterprise Surveys website. were expanded to allow for further stratification. This was the case in Costa Rica, Guatemala, and Uruguay, which are 3. For a justification of the optimal sample size per level categorized as small countries. of stratification, see the Sampling Note at http://www. enterprisesurveys.org/~/media/FPDKM/EnterpriseSurveys/ 9. Due to geographical proximity, survey design, and shared Documents/Methodology/Sampling_Note.pdf economic aspects, the Bahamas is categorized as the Caribbean. 4. In cases where zero values occur within the distribution, the References distribution used is the log of the variable plus one—that is, log(variable + 1). Sampling Note at http://www.enterprisesurveys.org. 5. Specifically these cases included firms that gave no reply, had an out-of-order line, a toneless line, a non-existent phone number, a fax line, an answering machine, or an incorrect address. Enterprise Surveys provide the world’s most comprehensive firm-level business environment data in developing economies. An Enterprise Survey is a firm-level survey of a representative sample of an economy’s private sector. The surveys cover a broad range of business environment topics including access to finance, competition, corruption, crime, gender, infrastructure, innovation, labor, performance measures, and trade. The World Bank has collected this data from face-to- face interviews with top managers and business owners in over 130,000 companies in more than 135 economies. Firm-level data and summary indicators are available on the website. www.enterprisesurveys.org 10 IMPLEMENTING ENTERPRISE SURVEYS