URBAN MANUFACTURING SURVEY: SAMPLING METHODOLOGY

Transcription

1 APPENDIX 1 URBAN MANUFACTURING SURVEY: SAMPLING METHODOLOGY This appendix describes the design and estimation procedure, including the weight assignment for the respondent establishments of the urban manufacturing survey that was conducted in Sri Lanka. Population Coverage Sri Lanka has nine provinces which are further subdivided into 25 districts. Of these 25 districts, some areas are considered urban, which are either administered by municipality center/urban center (MC/UC 16 MCs and 30 UCs). The rural areas are administered by district secretary local government. The manufacturing survey covered the following sectors: Food and Beverage (ISIC 311, 312, 313), Textile (321), Garments (322), Industrial Equipment (382), and Rubber Products (355). An updated and comprehensive list of all establishments for manufacturing, services or trade did not exist in Sri Lanka at the time of the survey. The closest to this list were the following lists maintained by the Department of Census and Statistics: List of manufacturing establishments for the Western Province (3 districts) developed in cooperation with United Nations Industrial Development Organization (UNIDO). A draft list was consolidated from the registers of the Board of Investment, Ministry of Industries, Ministry of Textile and MEIPIP and was further improved by ground checks. DCS is confident that they have captured all the establishments with five or more employees in the Western Province. List of manufacturing establishments with 25 or more employees that were used as the sampling frame for the Annual Survey of Manufacturing Industries. This list originated from the 1983 Census of Establishments. The 1993 Census of Establishments was not conducted because of the political situation in the country and was updated annually from the registers mentioned above. List of manufacturing establishments with less than 25 employees that were sampled for the Annual Survey of Industries. This list was partial and obsolete since there were no registers that could be used to update it and the births and deaths of establishments under this category are faster than those establishments with 25 or more employees. For this survey, the lists mentioned above were merged and employed as the sampling frame. General Sampling Design A total of 505 establishments were selected. Those 105 establishments that have 500 or more employees were selected with certainty, while others were selected by using the stratified simple random sample design. In general, the strata were defined by the sector X provincial groups. In cases where the stratum sizes in specific sectors were very sparse, sectors were collapsed. Hence, the final stratification was determined by iteratively collapsing small strata until a sufficient strata size was obtained to ensure a 100% replacement in all strata and adequate strata sample sizes. 55

2 URBAN MANUFACTURING SURVEY: SAMPLING METHODOLOGY In the absence of any other auxiliary variables in the list frame that could be used in the sample allocation and selection, sample sizes across strata were determined using proportional allocation. That is, if N h is the population size in stratum h and N is the population size, the first iteration sample size n h in stratum h is derived by n h = N N h 400 Determination of Weights The final weight for respondent k in stratum h -- w hk, is a composite of the base weight, w 1 hk, the non-response adjustment, w 2 hk, and the factor to compensate for coverage errors, w 3 hk, such that: = w w w (1) Base Weight whk 1 hk 2hk 3hk The base weight is the inverse of the probability of selection. The base weight for all large establishments (those with 500 or more employees) is 1 since they were all selected. Table A1.1 presents the base weight for all the small and medium establishments by province and manufacturing sector. Table A1.1 Base Weight for Small and Medium Establishments by Province and Manufacturing Sector (those with less than 500 employees) Province Province code Manufacturing sector Stratum weight Sample size Probability of selection Base weight Western 1 Food and Beverage Western 1 Garments Western 1 Industrial/Agri/Trans Equipment Western 1 Rubber Products Western 1 Textile Central 2 Food and Beverage Central 2 Equipment/Garments/Rubber Central 2 Textile Southern 3 Food and Beverage Southern 3 Equipment/Garments/Rubber Southern 3 Textile Northwest 6 Food and Beverage Northwest 6 Equipment/Garments/Rubber Northwest 6 Textile North Central 7 All Sectors Uva 8 Food and Beverage Uva 8 Textile/Garments/Rubber Sabaragumawa 9 Food and Beverage Sarabagumawa 9 Rubber Products Sarabagumawa 9 Textile/Garments

3 URBAN MANUFACTURING SURVEY: SAMPLING METHODOLOGY Non-response and Sampling Frame Errors Adjustments The weights estimated at the design stage were then adjusted to account for non-response and frame problems. Establishments who refused to participate in the survey were considered as nonrespondents. Frame problems occurred when establishments could not be located, were closed, out of scope (the ISIC classification was not specified correctly), or duplicates. Each of these had a different impact on weight adjustment. Table A1.2 reports the final weights for each strata. Table A1.2 Adjusted Weights for Large Establishments Manufacturing sector Final weight Garments Textiles Food & Beverages Industrial Equipment 1 Rubber Products Table A1.3 Adjusted Weights for Small & Medium Establishments Province Strata Final weight Western Western Western Western Western Central Central Central Southern Southern Southern Northwest Northwest 13 5 Northwest North Central 15 3 Uva Uva Sabaragumawa Sarabagumawa Sarabagumawa

4 URBAN MANUFACTURING SURVEY: SAMPLING METHODOLOGY Estimation Estimates for each of the strata described in table A1.1 could be derived separately. For example, the estimator for a total in stratum h is Yˆ = w y (5) h k hk where y hk is the observed value from the kth sample in stratum h. The estimator for the variance of this 2 sh estimator is v( Yˆ h ) = ( 1 f h ) (6) nh 2 where f h is the sampling fraction and s h is the sample variance in stratum h. hk The estimator for the population total would be Yˆ = w y hk hk (7) h k and its variance estimator is v( Yˆ ) N 2 ( 1 f ) 2 sh = n h (8) n h 58

5 APPENDIX 2 URBAN SERVICES SURVEY (TOURISM AND INFORMATION TECHNOLOGY): SAMPLING METHODOLOGY This appendix describes the sampling design for the investment climate survey conducted in three services sub-sectors of Sri Lanka namely: hotel, tourism (i.e., travel agencies) and IT. Population Coverage The target population for the services survey are establishments with their primary operations in hotels, travel agencies, and information technology (IT), particularly software development. Because there are no comprehensive lists of establishments under these sub-sectors, the lists that became the basis for sampling were developed from different sources. The sampling frame for hotel establishments and travel agencies was developed from the lists provided by the Ceylon Tourist Board, commercial telephone directories, associations of hotel and travel agencies, and the municipal business register for Colombo. For IT establishments, the sampling frame was developed from the associations of IT software developers and the municipal business register of Colombo. General Sampling Design The sample size for the service sector survey was fixed at 100 and equally divided among the 3 services sub-sectors at the onset. Due to sampling methodology constraints, the final sample sizes for the sub-sectors were: 41 hotel establishments and 30 each from travel agencies and 33 IT establishments. The strata were defined differently for each sub-sector, depending on the available auxiliary variables in the sampling frame. For the hotel establishments, the employment variable became the determinant of strata since it was available for most establishments. For IT, both employment size and exporting status were used. Only location data was available for travel agencies. For Hotel and IT sub-sectors, large establishments were selected with certainty. The rest of the sample size was then allocated to the remaining strata using proportional allocation. Establishments were then selected from the non-certainty strata using simple random sampling. For the travel agencies simple random sampling was used. The sample allocation and final weights for each sub-sector are shown in tables A2.1 to A2.3 below. Table A2.1 Sample Allocation and Base Weight for Hotel Establishments Stratum (employment size) Stratum size Sample size No. of respondents Selection probability Base weight Non-response adjustments Frame error adjustments Final weight 200 and above below Total

6 URBAN SERVICES SURVEY (TOURISM AND IT): SAMPLING METHODOLOGY Stratum Table A2.2 Sample Allocation and Base Weight for Tourism Establishments Stratum size Sample size No. of respondents Selection probability Base weight Non-response adjustments Frame error adjustments Final weight Colombo Outside Colombo Total Table A2.3 Sample Allocation and Base Weight for IT Establishments Stratum Stratum (employment & size export status) Sample size No. of respondents Selection probability Base weight Non-response adjustments Frame error adjustments Final weight 100 and above < 100, exporting <100, not exporting Total

7 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY APPENDIX 3 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY The rural investment climate survey in Sri Lanka is one of the first surveys to systematically collect information on the investment climate in which rural formal and informal businesses operate, and on the impact of the investment climate on firm productivity and the decision of rural households to engage in entrepreneurial activity. This survey is part of a pilot of rural investment climate surveys being undertaken by the World Bank. The Sri Lanka rural investment climate survey was conducted between December, 2003 and May, A total of 1327 rural non-farm enterprises and 1,063 rural households with and without enterprises were surveyed as part of a nationally representative sample of rural non-farm enterprises. The Survey Instrument The final survey instrument for the Sri Lanka rural investment climate study consisted of four components (i) a household survey, (ii) an enterprise survey, (iii) a community survey and (iv) a price survey. The household survey collected information on household demographics, sources of incomes, and levels of education. This questionnaire was administered to selected non-farm enterprises that were physically located within households (household based) as well as selected households that did not engage in rural non-farm enterprises. For households not engaged in nonfarm enterprise activities this module also collected data on factors preventing participation in non-farm enterprises. The enterprise questionnaire was completed for each rural non-farm enterprise selected for the survey. The manager or the most knowledgeable person about the firm was interviewed to complete the questionnaire. At the start of this questionnaire each non-farm enterprise was classified as a production, trade or service enterprise based on the sector from which they derived the highest share of sales. The community questionnaire was used to develop community profiles and identify community level characteristics that are important in determining the rural investment climate. This questionnaire was completed by interviewing various community leaders such as the village head, local government officials, the principal of a school etc. A community questionnaire was completed in each of the selected communities (GN Divisions). The price questionnaire gathered data on certain consumer, input and output prices prevailing in the main local market in each community. Sampling Due to the absence of an existing census of rural non-farm enterprises a sample frame was developed for the purposes of this study by the Sri Lankan Department of Census and Statistics in collaboration with the World Bank and AC Nielsen. As a first step, a decision was made to select a sample of 1500 rural non-farm enterprises and 600 households without non-farm enterprises. This sample size allowed national level estimates of the desired statistics given the budget constraints. A two stage sampling approach was adopted. In the first stage approximately 150 Grama Niladari (G.N.) divisions were randomly selected from among the 25 districts of Sri Lanka with probability proportionate to size using a systematic sampling method where the number of rural 61

8 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY housing units in each district was used as a measure of size. Rural areas were defined as per the current Department of Census and Statistics definition that classifies all areas that do not fall under Municipal Councils and Urban Councils as rural areas. There are around 14,000 G.N. Divisions in Sri Lanka of which 12,000 are classified as rural. For the selection of secondary sampling units (SSU's) for the two surveys, a census of all buildings within the boundary of the selected G.N. Divisions in the primary sample was undertaken. In all provinces excluding the North and East, the census (listing exercise) was conducted between September, 2003 and November, The listing in the North and East was completed between October 15, 2003, and March, A total of 65,579 units (residential and non-residential) were included in the listing exercise in 147 G.N.s (figure A3.1). Three G.N.s in the North and East province under LTTE control had to be dropped from the sample because of the inability to conduct the survey in these areas. A total of 27 GNs were selected in the North and East of which 17 were under government control and around 10 were directly under LTTE control. On completion of the listing operation, units with a non-farm enterprise (household based or stand alone) were separated in to the following two groups. Figure A3.1 Rural Sample Points 1) Group A : Enterprises with one or two persons engaged in the economic activity. 2) Group B : Enterprises with three or more persons engaged in the economic activity. For the purposes of the survey rural non-farm enterprise were defined as any income generating activity (trade, production or services) not related to primary production of crops/ livestock/fisheries undertaken either within the household or in any non-housing units. Any value addition (processing) to primary production was considered to be a rural non-farm activity. The total sample frame of rural non-farm enterprises is shown in table A3.1. After separating the establishments into groups A & B, the establishments in each group were sorted in order by industry type, Production, Services and Trade respectively. The sample enterprises to be surveyed were then selected using a systematic sampling procedure. The number of sample enterprises (secondary sample) selected from each sample G.N. were as follows: (i) 5 establishments from Group A; and (ii) 5 establishments from Group B. 62

9 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY Province Table A3.1 Rural Non-farm Enterprise Sample Frame District GN division (code) GN name Number of Number of enterprises enterprises with with more up to 2 workers than 2 workers Western Colombo 430 Kadugoda North 13 1 Western Colombo 450A Walpita 14 1 Western Colombo 474 Hewagama Western Colombo 491A Walpola Western Colombo 500 Brahmanagama Western Colombo 509 Kotuvila Western Colombo 532C Wattegedara Western Colombo 562 Wewala East Western Gampaha 119 Kotugoda Western Gampaha 119/1 Kotugoda Western Gampaha 151 Kovinna Western Gampaha 44A Kaluaggala 14 3 Western Gampaha 53 Nawana West 25 9 Western Gampaha 53A Paragoda South 9 4 Western Gampaha 64A Katana North Western Gampaha 89B Aluthepola East 38 7 Western Gampaha 9 Purana Meerigama 37 9 Western Gampaha 91C Dadonna East 22 3 Western Gampaha 97B Bomugammana South Western Gampaha 9A Maveehena 7 4 Western Kalutara 626A Kottiyawatta 36 0 Western Kalutara 687B Eludila Western Kalutara 694B BiganaThuduwa 43 9 Western Kalutara 708C Potupitiya West Western Kalutara 745 Munhene 59 9 Western Kalutara 787 Walallawita East 53 2 Western Kalutara 800E Dodangoda West-Central Western Kalutara 819H Gamagewatta 57 7 Western Kalutara 832B Ridirekagama 34 2 Central Kandy 1011 Galagoda 16 1 Central Kandy 1061 Giraulla 36 0 Central Kandy 218 Deldeniya 52 4 Central Kandy 474 Kurungudolla 19 3 Central Kandy 535 Uggala Janapadaya 9 0 Central Kandy 604 Meegamawatta 34 2 Central Kandy 62 Hatnagoda 22 6 Central Kandy 712 Malpana Central Kandy 938 Batumulla 27 2 Central Matale E325 Imbuldanda 15 4 Central Matale E371 Palle Weragama ˆ 63

10 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY Province District GN division (code) GN name Number of Number of enterprises enterprises with with more up to 2 workers than 2 workers Central Matale E433A Yatigalpotta 17 9 Central Matale E443 Kaludewa Paranagama 12 2 Central Nuwara-Eliya 315A Ginigathhena Central Nuwara-Eliya 320C Gawaravilla 51 4 Central Nuwara-Eliya 460G Ketubulawa 36 9 Central Nuwara-Eliya 475W Ssummerset Central Nuwara-Eliya 501F Rookwood Estate 14 2 Central Nuwara-Eliya 528F Matatilla 15 0 Southern Galle 113 Batadoowa Southern Galle 143A Halloluwagoda 44 4 Southern Galle 188 Hammeliya 26 7 Southern Galle 228 Malgalla 56 6 Southern Galle 24E Pathirajagama Southern Galle 30B Indipalegoda 62 9 Southern Galle 44B Ganegoda Southern Galle 75H Polhunnawa Southern Matara 241H Kandilpana 32 7 Southern Matara 245A Andaluwa 14 2 Southern Matara 318A Godawa Southern Matara 364 Diyalape Southern Matara 381B Wekada 36 4 Southern Matara 433A Devinuwara West 26 5 Southern Hambantota 178 Hungama Southern Hambantota 319 Dammulla East Southern Hambantota 34 Molkepopathana 32 8 Southern Hambantota 456 Koholana Southern Hambantota 98 Weliwewa 21 5 North Western Kurunegala 1322 Asseduma 23 8 North Western Kurunegala 1343 Bamunumulla 16 5 North Western Kurunegala 1472 Kovulwewa North Western Kurunegala 1591 Indivinna North Western Kurunegala 276 Rasnayakapura North Western Kurunegala 351 Pallekela 28 4 North Western Kurunegala 518 Udawela North Western Kurunegala 669 Mee/udagama 36 1 North Western Kurunegala 804 Dematagahapalessa 14 3 North Western Kurunegala 83 Kumbukwewa 4 1 North Western Kurunegala 850 Galabadagama 37 6 North Western Kurunegala 981 Malddeniya 22 3 North Western Puttlam 485A Hanatotupala North Western Puttlam 512B Maravila South North Western Puttlam 550A Karavitagara East

11 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY Province District GN division (code) GN name Number of Number of enterprises enterprises with with more up to 2 workers than 2 workers North Western Puttlam 615D Sirambiadiya 44 7 North Western Puttlam 629A Kandakuliya North Western Puttlam 653A Thalgaswewa 16 0 North Central Anuradhapura 103 Diviyaudabendawewa 4 2 North Central Anuradhapura 162 Galanbindunuwewa North Central Anuradhapura 324 Bogahawewa North Central Anuradhapura 420 Musalpitiya North Central Anuradhapura 44 Puhudivula 21 3 North Central Anuradhapura 479 Bulnewa North Central Anuradhapura 612 Kollankuttigama 26 2 North Central Polonnaruwa 198 Laksha Uyana North Central Polonnaruwa 270 Malvilla 18 7 North Central Polonnaruwa 8 Damanayaya North Central Polonnaruwa 93 New Town Uva Badulla 12 Ritigahaarawa 6 1 Uva Badulla 1N Hobariyawa 29 6 Uva Badulla 37A Wethalawa 10 3 Uva Badulla 48B Girambe 11 9 Uva Badulla 63B Diyathalawa Uva Badulla 67B Bindunuwewa Uva Badulla 80H Glen Alpin 6 4 Uva Badulla 88H Palagolla 9 2 Uva Monaragala 124B Tjossaira Uva Monaragala 146D Karaville Uva Monaragala 151D Nugayaya 15 9 Uva Monaragala 98B Mudiyala 10 4 Sabaragamuwa Kegalle 136C Thoranagoda 33 7 Sabaragamuwa Kegalle 155D Pathagama Sabaragamuwa Kegalle 174A Yakdehiwatta Sabaragamuwa Kegalle 185A Pallegedara 22 5 Sabaragamuwa Kegalle 214C Moraketiya 69 5 Sabaragamuwa Kegalle 237 Masimbula 38 1 Sabaragamuwa Kegalle 244C Metihakwala 25 1 Sabaragamuwa Kegalle 268A Pinnawala Sabaragamuwa Ratnapura 09A Kotagama 35 2 Sabaragamuwa Ratnapura 106 Yatiyantota Sabaragamuwa Ratnapura 121C Godagampala 36 6 Sabaragamuwa Ratnapura 38 Palliporuwa 40 2 Sabaragamuwa Ratnapura 53A Kavudugama 51 9 Sabaragamuwa Ratnapura 77B Kodapaluwa Sabaragamuwa Ratnapura 99 Lewangama North 38 2 North & East Jaffana J/172 Chulipuram West

12 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY Province District GN division (code) GN name Number of Number of enterprises enterprises with with more up to 2 workers than 2 workers North & East Jaffana J/208 Punnalai Kadduvan North 9 2 North & East Jaffana J/259 Kalviyankadu North & East Jaffana J/294 Navatkuli West 20 1 North & East Kilinochchi KN North & East Mannar MN/153 Mullikulam 4 1 North & East Mannar MN/48 Thalaimannar North 41 4 North & East Mannar MN/73 Pesalai North 32 4 North & East Vauniya 215B Katharsinnakulam 39 1 North & East Vauniya 218 Thandikulam North & East Vauniya 224A 14 0 North & East Mullaitivu M North & East Mullaitivu M North & East Mullaitivu M North & East Batticaloa 107B Kanthipuram 87 4 North & East Batticaloa 192 Eravur North & East Batticaloa 211E Mathuran Keni Kulam 46 0 North & East Amparai 144 Mahaoya North & East Amparai 78D Samanthurai North & East Amparai AP/ North & East Amparai P/ North & East Trincomalee 225I Samawechchathui 16 0 North & East Trincomalee 228F Mullipothanai East 28 1 North & East Trincomalee 31I Pullmoddai 69 6 Total In instances where adequate numbers were not listed under any of the above groups, the balance sample was also selected from the other group. If an adequate number was not available in both groups, the balance was selected from an adjoining G.N. to keep the sample size unaffected. The effective sample was increased by 6 establishments per province to account for anticipated unit non-response. Households without a non-farm enterprise were also identified using the listing information and a sample of 4 such households were selected using a systematic sampling procedure for the household survey. Sample weights for estimation purposes are computed as follows. (This formula applies separately for the two categories A & B of the Establishment Survey and households without enterprises.) Let, g i = No. of G.N.divisions selected from district ί G ίј = Total no. of housing units in the ј th G.N. in district ί. D ί = Total no. of housing units in the ί th district. N ίј = No. of units listed in ј th G.N. in ί th district. n ίј = No. of units selected from the ј th G.N. in the ί th district. 66

13 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY Then the sample weight for the ј th G.N. division in the ί th district W ίј is given by the following formula. W ίј = 1 g i Dί x N ίј G ίј x n ίј If the characteristic y measured in the k th unit, in the ј th G.N. division in the ί th district is given by Y ίјk, the total estimate ŷ at national level is obtained as below. (for categories A & B in the Establishment Survey and for the Household Survey respectively) ŷ = 25 i= 1 gi n ij j= 1 k = 1 Wij x Y ijk Sampling weights computed separately for the two categories A & B are given separately and if the total estimates for the two categories are denoted by ŷ A & ŷ B (Establishment Survey) then the combined national level estimate, ŷ national = ŷ A + ŷ B. The original weights were adjusted for unit non-response before estimation of sample characteristics. The final sample distribution of rural non-farm enterprises and rural households by province is shown in table A3.2 and table A3.3, respectively. Table A3.4 shows the distribution of the non-farm enterprise sample by industry type. 67

14 RURAL SURVEY: INSTRUMENT AND SAMPLING METHODOLOGY Table A3.2 Distributions of Enterprise Sample by Province Province Stand-alone Frequency Household based Total Share of total Western % Central % Southern % North Western % North Central % Uva % Sabaragamuwa % North & East % Total , % Table A3.3 Distribution of Household Sample by Province Frequency Share of total Province Households without non-farm enterprises Household with enterprises Total Western % Central % Southern % North Western % North Central % Uva % Sabaragamuwa % North & East % Total , % Note: The number of households with enterprises and household based enterprises differ between the two samples as some of the businesses had closed down or had re-located to another GN in the time between the listing exercise and the actual survey. Table A3.4 Distribution of Enterprise Sample by Province and Sector Province Production Service Trade Total Western Central Southern North Western North Central Uva Sabaragamuwa North & East Total

15 APPENDIX 4 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE I. Methodology The impact of the IC on firm performance is gauged by using regression analysis. Various versions of the following reduced-form regression equation are estimated: y i = α 0 + α 1 *IC i + α 2 *X i + α 3 *Z i + ε i (1) The dependent variable, y, represents a measure of firm performance. IC represents a measure of the investment climate. X denotes a vector of firm-level control variables. Z is a vector of industry controls (industry dummies) and a location control (whether the firm is located in the capital region or not). In this document, firm performance is measured using seven different variables: (i) average annual sales growth; (ii) average annual employment growth; (iii) average investment rate; (iv) whether or not a firm invested above a certain threshold rate in any of the three sample years; (v) labor productivity; and (vi) two measures of TFP. With the exception of (iv) all the firm performance measures are continuous (or nearly so as in the case of (iii)). The regression equation is therefore estimated by OLS in all cases except when (iv) is the dependent variable. In this latter case, estimation is based on the Probit model. Table A4.1 below provides details on how each of these measures was constructed. Firm performance measure Sales growth Employment growth Investment rate Any new investment Labor productivity* TFP-LP* TFP-OLS* Table A4.1 Definitions of Firm Performance Variables Definition Average annual growth in sales for the past three years (where sales are deflated by industry-specific producers price indexes). Average annual growth in total employment (permanent employees + temporary employees) for the past three years. Average investment rate over the last three years. (Investment rate is defined as investment in plant and machinery divided by the book value of fixed assets.) Whether a firm s net investment in plant and machinery was 5% or more of fixed capital in any of the three sample years. Value added divided by total employment Total factor productivity estimated using the Levinsohn-Petrin 1 estimator applied to a simple Cobb-Douglas value-added production function: lnv it = β 0 + β L *ln(l it ) + β K *ln(k it ) + Year 00 + Year 02 + ω it + η it. V denotes value added, L denotes total employment, K denotes fixed capital, and Year denotes year dummies. The estimate of ω it captures TFP. Total factor productivity estimated using OLS estimator applied to the following Cobb-Douglas value-added production function: lnv ij = β 0 + Σ j D j *[β L *ln(l it ) + β K *ln(k it )] + Year 00 + Year 02 + ε i, where V, L K, and Year are as before, and D denotes industry dummies. The index i represents firms and j denotes four industries (garments and textile; food and beverages; industrial equipment; and rubber products). The estimate of ε it captures TFP. Note: (*) Each of the three productivity measures introduced in the regression equation (1) pertain to the latest of the three sample years,

16 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE The investment climate is captured in a variety of ways. In particular, variables are grouped in terms of their relationship to (i) international integration; (ii) functioning of labor markets; (iii) the state of physical infrastructure; (iv) measures of governance, including administrative and regulatory functioning; and finally (v) finance. Three main econometric problems might bias our results: non response, multicollinearity, and endogeneity. We attempted to correct for all of them. We dealt with non coverage by estimating sampling weights and correcting them for non response and frame problems. All our regressions, except for the Levinsohn-Petrin approach, are weighted hence the estimated values are adjusted for item non response as well. Since a number of investment climate measures are related to one another, including them all in a regression equation would cause difficulties in making robust inference due to multicollinearity issues. Additionally, since not all firms answer every question, including all the investment climate measures together could lead to a large attrition in the sample size. For both reasons, the investment climate measures are included one-by-one. However, as a robustness check we also run a regression where we include together the various investment climate measures which individually show up as having a significant impact on at least one dimension of firm performance. A problem with the type of investment climate analysis carried out here is the potential for endogeneity bias. In particular, establishing the direction of causality from a measure of investment climate to a measure of firm-performance is problematic. To alleviate endogeneity bias, equation (1) is always estimated with controls. These include industry dummies to control for industry-specific effects, a dummy for the Greater Colombo area to control for locationspecific factors, and various firm level characteristics. In addition to the age of the firm, the latter include initial sales in the case of sales growth as a measure of firm performance, initial employment in the case of employment growth, initial capital in the case of the investment related firm performance measures 2, and establishment size dummies for all the productivity-related measures. When we attempted to measure the impact of perception on firm performance a particular set of controls were used. In this case the respondents psychology can play a relevant role in his assessment of the obstacles. Hence the set of controls adopted were the average value of perception in the location of the respondent, the level of education of the respondent himself, and also a dummy variable for interviewer, intended to capture eventual fixed effects due to the style of interview and the personality of the interviewer. II. The Investment Climate Regression Results The results of the IC analysis are consistent with the belief that greater international integration will serve Sri Lanka s urban manufacturing sector well. In particular, the results suggest that better access to imported inputs, specialized skills embodied in foreign workers, and foreign investment would significantly raise the productivity of Sri Lankan firms and generate new investment. There is some evidence suggesting that labor regulations and institutions may be restricting the ability of firms to adjust their workforce in accordance with changes in product demand. To the extent that firms respond to such regulations and institutions by hiring more workers on a temporary basis, the data indicate that firms may be doing so at the expense of 70

17 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE higher productivity. The IC data also indicate significant payoffs from streamlining regulations and reducing regulatory uncertainty. Key regression results are summarized below. International Integration: It is widely acknowledged that the degree of international integration is closely related to how conducive an investment climate is in fostering growth and productivity improvements. A number of channels operate in this direction. First, greater international integration means that firms will have better access to knowledge disembodied and embodied. Second, greater integration also means more channels for financing investment. Finally, firms may be able to become more productive by exporting for example, via learning-by-exporting or by reaping economies of scale made possible by the larger international market. 3 The data from Sri Lankan firms bears out the benefits that international integration brings. International integration facilitates the acquisition of technology in Sri Lanka through two main channels: machinery and personnel. Over half of the sample entrepreneurs consider new machinery and equipment and hiring key personnel as the two most important sources for technological innovations. Those firms show a strong and positive effect of imported new capital goods on their productivity (table A4.2, row 1). Similarly firms with foreign professionals in their workforce are among the top performers in the country in terms of productivity (table A4.2, row 2). In line with results from other countries, the IC survey shows that firms that have at least some foreign ownership and those that export are more productive (table A4.2, rows 3-4). Not surprisingly, having foreign ownership is also found to facilitate investments: firms with foreign ownership are more likely to invest in expanding productive capacity and have a higher investment rate (table A4.2, row 3). It is worth noting that while causality is difficult to establish, the positive association between exporting and productivity is consistent with the learning-byexporting hypothesis. Labor Market Issues: Given the critical importance of labor in production, the functioning of labor markets can have a significant bearing on a country s investment climate. Labor regulations are noted as a major or severe obstacle to operations and growth of firms by around 25% of the firms. This is confirmed by the fact that a similar number of firms express their desire to cut their labor force if they were free to do so. What aspects of labor regulations affect firms? The presence of unions and a dummy variable indicating excess labor of 10% or more were significantly and positively associated with the probability that a firm listed labor regulations as a severe problem. 4 The latter suggests that legal constraints on firing/laying-off workers are a factor contributing to firms difficulties with labor regulations. Do these aspects of labor markets affect firm performance? The presence of a union is associated with lower sales growth and lower employment growth (table A4.2, row 5a) while having excess labor is associated with lower sales growth (table A4.2, row 5d). The latter is suggestive of an adverse impact of labor regulations on firm performance. 5 Although unionization and excess labor are not found to affect productivity adversely, we do find that the larger the share of temporary workers in total employment, the lower is productivity 71

18 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE (table A4.2, row 6). To the extent that firms respond to unionization and restrictions on layoffs by hiring temporary workers, this finding represents an adverse impact of labor regulations on productivity. The evidence is thus certainly consistent with the notion that Sri Lanka s labor regulations have created an insider-outsider distinction among workers. Controlling for industry and education of the workforce, unionized firms do tend to pay more. Thus, labor institutions and regulations may benefit those already with jobs, but they do not help employment and they appear to dampen sales growth. 6 Governance and Administrative Barriers: While our data suggest that corruption is not a particularly serious issue, this does not mean that there are no governance related issues holding back Sri Lanka s investment climate. In particular, regulatory uncertainty is an important issue. If we take the difference between the maximum and the average number of days to clear exports as an index of uncertainty, the analysis clearly shows a significant and negative association between uncertainty and productivity. This is also true in the case of imports (table A4.2, row 7). In addition, our data suggest that there are gains to be had from streamlining regulations: Firms which report a higher share of senior management s time being taken up in dealing with requirements imposed by government regulations are the ones who are less likely to make any significant investment in increasing productive capacity (table A4.2, row 8). Finance: Finance is yet another pressing constraint identified by entrepreneurs in Sri Lanka. Although both access and cost of finance are among the top bottlenecks, the ability to secure a loan seems the most pressing. As a matter of fact internal funds, in the form of retained earnings and equity, remain by and large the principal source of financing for Sri Lankan firms. Unfortunately the survey shows that being more productive is not translated in having easier access to bank loans (table A4.2, row 9c). Hence, while more productive firms need to resort to retained earnings for their investments, less productive firms must rely on their equity to finance new investments (table A4.2, rows 9a and 9g). The survey data suggests that banks appear to be unable to discriminate better performing loan applicants and rely more on the value of collateral when approving a loan application. In fact while only half of the sampled firms had a loan in 2002, almost all of them were required to provide a collateral in the form of fixed assets (land, buildings and machinery). Furthermore the average value of collateral required equals the loan value. Not surprisingly the probability of obtaining a loan is skewed toward large firms. Also financial constraints on firms do seem to be relaxed by FDI. Infrastructure: As noted in the main text of this report, entrepreneurs perception of IC bottlenecks start with complaints about electricity and transportation. The regression analysis confirms the negative impact of Sri Lanka s poor infrastructure on firm performance. Only establishments using a generator (over ¾ of the sample and accounting for 13% of the total energy needs) show a significantly higher level of performance. 7 On the contrary perception of unreliable power supply is not significantly related to firms productivity (table A4.3). Far from being an indicator of good infrastructure, the insignificance of the perception coefficient is justified by the ability of most establishments to protect against power outages and maintain their level of activity. Hence the insignificance of the negative perception of electricity delivery on productivity should not lead us to conclude that electricity is not a major constraint. Although in fact entrepreneurs can to a certain degree protect themselves against power outages, this 72

19 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE phenomenon has an impact of the cost of production and hence on the international competitiveness of Sri Lankan firms. The second infrastructure related complaint identified by respondents is transport. The perception of poor transport facilities is negatively and significantly associated with productivity since entrepreneurs are unable to protect themselves. Firms in Sri Lanka have to maintain a minimum average of 30 days of raw material inventories and lose around 2% of the value of their sales due to delivery delays. This also puts local firms at a disadvantage if we consider that in a country as big as India firms hold half as many days as inventories. The IC data also show that various indicators for the state of the transportation infrastructure have an impact on firm performance. For example, having a paved road leading to the nearest urban center is significantly associated with higher sales growth. Similarly, a poor transportation infrastructure as reflected in a reporting of breakage, theft, and spoilage during shipment is associated with both significantly lower sales growth as well as lower employment growth. 8 Regression Results Using All Key IC Measures Simultaneously: As a check on the robustness of our results, we also run our IC regressions by including simultaneously key IC measures as regressors. Despite the fact that there is a large reduction in sample size by about a 100 firms due to missing observations, the results retain the general flavor of those described above fairly well. In particular, as indicated in table A4.4, imported machinery and equipment continues to have a significant impact on productivity and several other measures of firm performance, unionization and excess labor continue to be associated with weak sales growth (and employment growth in the case of unionization), and a higher share of management time spent dealing with regulations are associated with lower probability of investing in new plant and equipment. Additionally, the pattern of estimates on the various educational-status regressors suggest that formal education at even a primary level is productivity enhancing (the omitted category being workers without even primary education) while the benefits of higher education kick in for firms which are expanding productive capacity by investing in new plant and equipment. What is the quantitative importance of these factors? Focusing on the IC variables which impact firm performance significantly in table A4.4, we can ask to what extent an improvement in each key IC variable translates into better firm performance. Improvements are represented by a one standard deviation increase (or decrease) in the IC variables. Consider the impact of a one standard deviation increase in the share of new imported capital goods on productivity due to easier access to imported inputs. The estimates of table A4.4 indicate that in the case of the L-P measure of TFP, for instance, productivity would increase by 3% for a firm with average productivity. Similarly, what would be the impact of streamlining regulatory requirements so that managers would have to spend less time on them? The investment probit results of table A4.4 indicate that a one standard deviation reduction in the time spent by a manager working in a firm with average attributes would result in an almost fourteen percentage point increase in the probability of investing in new plant and equipment. Of course, it may not be easy to drive down time spent complying with regulations by one standard deviation: a one standard deviation reduction would amount to almost negligible time spent dealing with regulations. While this may not be feasible at least some reduction should be possible and identifying the more cumbersome regulations would seem to hold promise for improving firm performance. 73

20 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE Notes For details, see, Levinsohn and Petrin (2003). A measure of capacity utilization was also used. Its inclusion appeared to make no difference to the main results; however, it led to a reduction in number of observations due to some missing values. As a result, this variable is not included as a control in the regressions reported here. Greater international integration also means greater competition. Increased competition can in turn improve overall productivity by forcing domestic producers to choose between enhancing efficiency or losing market share. The results of this regression are available upon demand. Of course, association does not imply causation. This is especially relevant in interpreting the results of the excess labor regression since even in the absence of any regulatory barriers to firing, an unforeseen reduction in sales (induced by a demand shock, for example) would typically lead to some labor hoarding by the firm. In other words, a firm may anticipate its sales to rebound in the future and so keep its workers on its employment rolls given that the search costs associated with hiring of workers are unlikely to be trivial. The firm would have excess labor, but it would be due to labor hoarding and not regulatory barriers to firing workers. While we cannot disentangle the relative strengths of the labor hoarding versus regulatory barrier channel in explaining our results, data from the services sector IC study are illuminating. In the questionnaire used there, firms were asked to list the various reasons for excess labor, if any, including labor regulations regarding hiring and firing, pressure from labor unions, and labor hoarding. Of the five hotels which report having excess labor (13% of the total sample), four report labor laws as a reason while three report labor hoarding as a reason. Clearly, both forces are in operation and the message we take is that at least a part of the negative association we find between sales growth and excess labor is likely to be driven by labor regulations. Anecdotal evidence gathered during the pilot test of the IC survey shows that for some firms, it may not be so much the restriction on firing workers that is the key problem but instead the uncertainty associated with the costs of retrenchment. One respondent, for example, stressed that the lack of clarity in the computation of severance pay that workers were legally entitled to was a problem. In particular, since the respondent s garment business is subject to uncertainty after the MFA system is phased out, the respondent has been interested in exploring other industries/activities. Not being able to work out exit costs therefore presents special difficulties for the firm. Results are available upon demand. Results are available upon demand. 74

21 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE Table A4.2 Investment Climate Regressions by Major Topic (1) (2) (3) (4) (5) (6) (7) Sales growth Employment growth Investment rate Investment PROBIT Labor productivity TFP-LP TFP-OLS International Integration: (2.40)** (1.88)* (2.03)** (0.21) (0.30) (0.23) (0.52) (0.41) (1.16) (0.64) (2.66)*** (1.20) (0.90) (1.16) (1) Imported new M&E (avg.share) (3.01)*** (1.80)* (2.26)** (3.85)*** (4.43)*** (3.18)*** (2.72)*** Observations R-squared (2) Foreign workers (dummy) (0.61) (0.27) (0.98) (0.03) (2.00)** (2.05)** (1.70)* Observations R-squared (3) FDI (dummy) (1.50) (0.06) (1.96)* (1.81)* (3.53)*** (2.82)*** (2.04)** Observations R-squared (4) Exporter (dummy based on information on when firm started exporting) (2.47)** (0.15) (0.26) (0.82) (2.78)*** (2.65)*** (2.15)** Observations R-squared Labor Market Issues: (5a) Union (dummy) (5b) Strike (dummy) (0.00) (0.05) (0.60) (0.51) (1.09) (0.87) (0.83) (5c) Absenteeism (dummy) (0.85) (0.87) (0.01) (1.53) (1.43) (1.22) (0.71) (5d) Excess labor (dummy) (2.26)** (0.77) (0.36) (1.02) (0.03) (0.25) (0.35) Observations R-squared (6) Share of temporary workers (0.22) (1.14) (0.09) (0.58) (2.41)** (2.41)** (2.22)** Observations R-squared Governance and Administrative Barrier: (7) Uncertainty at customs: Importers (0.50) (1.41) (2.08)** (1.06) (3.13)*** (3.16)*** (3.06)*** Observations R-squared (8) Percent time with govt. administrators Observations R-squared

22 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE (1) (2) (3) (4) (5) (6) (7) Sales growth Employment growth Investment rate Investment PROBIT Labor productivity TFP-LP TFP-OLS Finance: (1.25) (2.56)** (1.28) (1.76)* (0.24) (0.76) (0.96) (1.56) (0.75) (1.00) (1.19) (1.02) (0.88) (9a) Ret. earnings (dummy) (0.57) (0.33) (1.57) (0.04) (2.24)** (2.28)** (2.55)** (9b) Trade credit (dummy) (2.28)** (1.10) (1.60) (0.87) (1.63) (1.39) (1.20) (9c) Bank loans (dummy) (9d) Leasing (dummy) (0.45) (1.15) (0.05) (0.84) (1.27) (1.37) (0.93) (9e) State funds (dummy) (0.29) (0.05) (0.67) (1.24) (0.52) (0.76) (0.76) (9f) Credit card (dummy) (1.07) (1.01) (1.31) (1.21) (0.32) (0.05) (0.22) (9g) Equity (dummy) (0.52) (0.31) (1.84)* (0.14) (2.84)*** (3.53)*** (3.27)*** (9h) Family (dummy) (0.52) (0.06) (0.41) (0.55) (0.26) (0.87) (0.74) (9i) Informal s. (dummy) Observations R-squared Absolute value of t statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% Note: Each cell represents a separate regression, using firm and industry controls which are not shown here. The additional control variables under each performance indicator are: sales growth - initial sales; employment growth - initial employment; investment rate/probit - initial fixed capital; labor productivity and TFP - firm size dummy variables. 76

23 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE Table A4.3 Perception on IC Variables (dummy 1= major or severe problem) (1) (2) (3) (4) (5) (6) (7) Sales growth Employment growth Investment rate Investment PROBIT Labor productivity TFP-LP TFP-OLS Telecom (2.19)** (0.61) (7.19)*** (16.63)*** (1.21) (1.98)** (1.12) Electricity (1.86)* (0.08) (8.10)*** (.) (0.18) (0.28) (0.25) Transport (0.10) (0.94) (0.00) (0.29) (1.12) (1.76)* (1.93)* Tax rates (3.20)*** (2.32)** (6.67)*** (.) (1.32) (2.15)** (1.19) Tax administration (0.89) (0.09) (4.58)*** (.) (0.38) (0.25) (0.94) Labor regulations (1.20) (2.15)** (2.09)** (.) (2.86)*** (3.26)*** (2.53)** Skills (0.91) (3.09)*** (4.53)*** (15.60)*** (1.15) (0.79) (1.00) Access to finance (2.91)*** (2.29)** (5.51)*** (.) (2.26)** (2.89)*** (1.91)* Cost of finance (0.62) (3.90)*** (4.36)*** (2.62)*** (2.33)** (2.18)** (2.90)*** Policy uncertainty (1.27) (1.28) (3.31)*** (.) (0.55) (1.30) (0.98) Macro instability (1.23) (1.65)* (0.77) (.) (0.65) (0.49) (0.73) Corruption (1.86)* (0.26) (5.54)*** (8.21)*** (0.23) (0.31) (0.42) Anti competitive practices (0.90) (3.69)*** (1.42) (3.11)*** (1.25) (1.61) (1.84)* Observations R-squared Note: Controls included but not shown are: (a) average perception in location of firm, (b) education level of respondent, (c) interviewer dummy, and (d) size. The additional control variables under each performance indicator are: sales growth - initial sales; employment growth - initial employment; investment rate/probit - initial fixed capital; labor productivity and TFP - firm size dummy variables. 77

24 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE Table A4.4 Key Non-perception Based IC Measures (included simultaneously) (1) (2) (3) (4) (5) (6) (7) Sales growth Employment growth Investment rate Investment PROBIT Labor productivity TFP-LP TFP-OLS Imported new M&E (avg.share) (1.97)** (0.66) (1.63) (3.76)*** (3.85)*** (2.60)** (1.99)** Foreign workers (dummy) (0.23) (0.15) (0.34) (0.60) (0.18) (0.02) (0.16) FDI (dummy) (1.06) (0.04) (0.61) (0.27) (0.78) (0.41) (0.28) Exporter (dummy based on information on when firm started exporting) (1.05) (0.61) (0.72) (0.38) (1.45) (1.52) (1.18) Union (dummy) (1.66)* (2.71)*** (0.46) (0.19) (0.16) (0.26) (0.51) Excess labor (dummy) (1.87)* (1.51) (0.80) (0.08) (0.80) (1.19) (1.19) Share of temporary workers Elementary: 6-9 yrs. (share of workers) Middle school:10-13 yrs. (share of workers) University: more than 13 years (share of workers) Percent time with govt. administrators (0.84) (1.54) (0.26) (0.24) (1.46) (1.25) (0.84) (1.81)* (0.64) (1.12) (0.96) (2.21)** (2.80)*** (2.73)*** (2.06)** (0.52) (1.16) (1.17) (3.25)*** (3.89)*** (3.52)*** (0.57) (0.59) (4.19)*** (2.38)** (0.94) (0.97) (0.90) (0.57) (1.34) (1.57) (3.13)*** (0.16) (0.18) (0.68) Ret. earnings (dummy) (0.33) (1.32) (1.53) (1.59) (1.01) (0.89) (1.47) Trade credit (dummy) (1.09) (0.57) (1.06) (0.46) (0.84) (0.66) (0.50) Bank loans (dummy) (0.48) (1.94)* (1.18) (1.73)* (0.89) (1.88)* (1.89)* Equity (dummy) (0.80) (0.48) (0.63) (0.54) (2.44)** (3.05)*** (2.90)*** Firm s age (log) (0.68) (1.37) (1.67)* (1.40) (0.06) (0.32) (0.22) Initial real sales (log) (2.93)*** Garments (1.03) (1.17) (1.64) (1.90)* (2.90)*** (2.19)** (1.18) Textile (0.66) (0.59) (0.51) (1.04) (2.49)** (2.07)** (0.10) 78

25 URBAN MANUFACTURING SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE AND FIRM PERFORMANCE (1) (2) (3) (4) (5) (6) (7) Sales growth Employment growth Investment rate Investment PROBIT Labor productivity TFP-LP TFP-OLS Food & Beverages (0.19) (0.81) (0.63) (0.07) (0.86) (0.16) (1.00) (1.44) (0.08) (0.08) (1.08) Industrial equip (0.89) (0.15) (0.17) Greater Colombo (1.02) (0.38) (0.02) (0.51) (1.29) (1.05) (0.78) Initial workforce (log) (1.24) Rubber products Size = Medium (1.62) (0.78) (1.22) Size = Large (2.78)*** (0.43) (1.65)* Initial fixed capital (log) (3.85)*** (3.02)*** Observations R-squared Absolute value of t statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 79

26 APPENDIX 5 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE, AND START-UP How does the investment climate affect enterprise start-up? To analyze this question the effect of the prevailing investment climate on the probability of rural households establishing a non-farm enterprise is estimated. From a policy point of view, it is important to identify the key barriers to enterprise start-up so as to reduce or ease these constraints and thereby facilitate the growth of the rural non-farm economy. The rural IC survey collected data on both households that had just established new non-farm enterprises as well as on those that did not operate any non-farm enterprises, enabling an analysis of the IC on enterprise start-up. To identify obstacles to the startup of a rural non-farm enterprise and examine the magnitude of the impact of each of the constraints, a probit model is specified: Z = α 0 + α 1 (H)+ α 2 (C)+ α 3 (IC) + α 4 (D) + ε (1) In this model, the left hand side variable, Z, is a dummy variable that equals 1 for a household that has recently started a new non-farm enterprise and 0 for a household that does not operate any non-farm enterprises. i H is a vector of household characteristics that could potentially affect a household s economic decision. This vector includes the household head s age, maximum education attained by any household members, household size, land endowment, wealth and household access to informal borrowing, and a dummy variable representing whether or not the household head s parents operated a non-farm enterprise. C is a vector of community characteristics, including distance to a bank, the time taken from the community to the main market, the share of paddy land in total cultivated land and a dummy variable for agriculture as the main source of income in the community. IC is a vector of community level investment climate indicators. Rural enterprises were asked to rank the various business constraints they faced and identify the most important overall constraint. To construct an indicator of the IC in a rural community the perceptions of individual entrepreneurs in a community were aggregated and the constraint that was most frequently identified as being the most serious obstacle facing businesses in that community was identified as the most important overall constraint for the community. In this manner a series of dummy variables were created at the community level for the various investment climate constraints. If more than one constraint was identified as the most important overall problem by equal shares of entrepreneurs, both constraints were assigned a value of 1, the others are assigned to 0. The investment climate constraints in the vector IC include electricity, road access, road quality, cost of financing, bureaucracy of financing, market information access, market demand, availability of transportation facility, telecommunication service and water supply. This set of variables virtually exhausts the list of the most important overall constraints reported by the enterprises. Finally, a vector of provincial dummy variables D is included to control for regional differences. The estimation sample includes all households that did not operate a non-farm enterprise and households with enterprises that were less than 2 years old. The results of this regression are show in table A5.1. Column 2 includes a larger set of IC indicators as compared to Column 1. 80

27 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Does the availability of public goods and a better investment climate result in more investment by existing enterprises? While it is interesting to investigate the effect of the rural investment climate on the startup of new enterprises, it is equally important to analyze the impact of the current rural investment climate in Sri Lanka on the investments and performance of existing enterprises. This is also relevant to the prevailing debate as to whether the government should do more to encourage startups or pay more attention to fostering the growth of existing non-farm enterprises or do something to help both startups and existing enterprises. To examine the impact of the rural investment climate on new investments the probability that existing enterprises make a new investment as well as the effect of the IC on the magnitude of investments is analyzed. A probit model is used to analyze the effect of the IC on the probability of making any new investment and a tobit model is used to analyze the effect on the magnitude of investments. The probit model is very similar to the probit model for the startup analysis (equation 1). The only difference is that now Z stands for the investment decision, 1 for the enterprises who made any new investments in 2003, and 0 for the others. H is now a vector of enterprise characteristics that includes size, industry type, and age of the enterprises, and a set of variables reflecting the characteristics of the top managers (including maximum education level, gender, years of experience and ethnicity). IC and D stand for the same set of variables that are used in the startup analysis. The only difference is that since the sample is now restricted to enterprises that are at least one year old, the individual rankings of the enterprises are included in the regressions without aggregating to the community level. A separate set of regressions is also run where the IC variables are aggregated to the community level to try to control for the endogeneity of individual perceptions. The only difference in the tobit specification is that the dependent variable Z is a continuous variable representing the value of new investments actually made in These results are summarized in table A5.2. Does a better investment climate affect firm performance? To examine how different elements of the investment climate affect performance in terms of partial labor productivity (sales per worker or value added per worker) and total factor productivity (TFP), a reduced form equation is specified as follows: Y = β 0 + β 1 (E) + β 2 (C) + β 3 (IC) + β 4 (D) + η (2) Y represents value added per worker or TFP (all in log terms). E is a vector of enterprise characteristics including age, type, log of the number of employees. C is a vector of community characteristics that are likely to affect the efficiency of non-farm enterprises operations. IC is the same vector of IC constraints as in the previous regressions. And finally D is the vector of provincial dummy variables. Ordinary Least Squares (OLS), correcting for clustering, is used to estimate equation (2). TFP is estimated using a production function approach. In the relevant literature, both total output (total sales adjusted for inventory changes over two periods) ii and value added are used as measures of enterprise s production in the production function equation. A typical Cobb-Douglass production function (when value added is used) can be expressed as: iii lny = γ 0 + γ 1 (lnl)+ γ 2 (lnk)+ γ 3 (E) + γ 4 (D) + µ (3) Where lny is value added in 2003 (in log term). lnl is the total number of workers in 2003, lnk is the value of all the fixed assets, both are measured in log term; E is enterprise characteristics include age and type, and vector D include all the provincial dummy variables. Total factor productivity is calculated as the difference between the real total value added (or real total output) and the predicted value added (or predicted total output) based on the estimated parameters of 81

28 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP equation (3). Tables A5.3 and A5.4 display the regression results. Table A5.5 shows estimates of the Cobb-Douglas production function. Rural Investment Climate Regression Results The Rural Investment Climate and Rural Enterprise Performance Entrepreneurs are most likely to invest in their businesses if they see favorable prospects for future growth and higher profits. Whether or not they make investments depends on a host of factors including their access to capital, their perceptions of consumer demand for the goods and services they produce, and whether they can conveniently hire workers with the skills they need. They also include perceptions of how government policies and regulations affect their bottom line and if the infrastructure that they have access to will support their needs. Analysis of data from the rural survey indicate that many firm characteristics in terms of their size and composition and the skills of the management are important determinants of whether they make new investments. But in addition to these traits, the prevailing investment climate has a pronounced impact on the decision of entrepreneurs to make investments and influences the size of the investments that they make. Compared to the enterprises who just started in 2002, older enterprises (those in the age groups 2-5 years, 5-10 years and more than 10 years) are less likely to invest and tend to invest a smaller amounts (table A5.2) 1. For example, an enterprise between five to ten years age is 17 percent less likely to make a new investment compared to an enterprise that is less than two years old. Compared to enterprises that have one or two workers, those with more than five workers are 11 percent more likely to make a new investment although the amount of investment is not significantly different. As expected, household-based enterprises are significantly less likely to make new investments and they also tend to invest significantly less than stand-alone enterprises. Among all the characteristics of the top manager, years of experience is the only variable that matters in terms of whether or not they make a new investment; an enterprise whose top manager has more relevant experience is more likely to invest and make a larger investment. Among the variables included in the analysis to measure different dimensions of the investment climate, electricity, transport, road quality, water supply and marketing all comes out as important determinations of the whether or not firms invest. Enterprises that perceived electricity as being the most important overall constraint were 10 percent less likely to have made a new investment in the previous 12 months and the size of their investments were significantly smaller. Similarly, the perception of poor road quality and a lack of market information being the most important overall constraint reduced the probability of rural enterprises making a new investment by 11 percent and 17 percent, respectively. Although only a small proportion of businesses reported water as being the most important overall constraint, firms that perceived major problems with water supply were significantly less likely (12 percent) to make a new investment. High interest rates or high transactions costs on loans also reduced the amount invested. Consistent with the effect of perceived road access and quality constraints on new investments, an enterprise located in a community where the most common road surface was mud is 7 percent less likely to invest compared to an enterprise located in a community with gravel and asphalt roads. Although the type of roads is only marginally significant, the size of the investments made by enterprises in communities with mud roads is significantly less than those made by enterprises in communities with gravel or asphalt roads. Other significant predictors of whether or not firms make new investments include the time it takes to reach the closest commercial center and the amount of capital available to the enterprise from informal sources. Entrepreneurs that can borrow more money from informal sources tends to invest a larger amount. 82

29 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP As the perceptions of individual entrepreneurs may be endogenous, it is possible that the results of the analysis of the investment climate perceptions on investments may be biased. In order to see whether the same conclusions hold if endogeneity issues are controlled, data on the perceived business obstacles are aggregated to the community level and the analysis is repeated with the community level perceptions of the overall business environment. Interestingly the results are very consistent with those obtained using the perceptions of individual entrepreneurs. The two most noticeable differences are: (1) the perceived telecommunication, loan procedure and low market demand problems that were not significant in the analysis with the enterprise level perceptions become significant when community level perceptions are introduced; and (2) the coefficients of the perceived constraints are all much bigger than the earlier results. For example, the coefficient for the electricity (road access) changes from to ( to ). This is not surprising given the fact that now the perceived problem is the problem at the community level as opposed to the perception of individual entrepreneurs. To interpret the coefficients of the aggregated perceived constraints, implies that an enterprise is 50 percent less likely to invest if it is located in a community where electricity is perceived as the most important problem by the majority of entrepreneurs as compared to if it were located in a community where electricity is not perceived as the most important overall constraint. An analysis disaggregated by sector reveals that firms are affected in different ways depending on the nature of their business activities. For example while electricity was negatively and significantly correlated with the probability of a production related enterprise making a new investment, it did not have a significant impact on trading and service related enterprises. Rural production enterprises that rated electricity as the number one problem they face were 9 percent less likely to make a new investment as compared to production enterprises who felt that other problems overshadowed poor electricity supply. Production related enterprises that identified poor market demand or lack of market information as major problems were 11 percent and 14 percent less likely to make new investments, respectively. Interestingly, the only IC constraint that comes up significant for trading establishments is lack of market information. Traders who rate lack of market information as the most important problem were 14 percent less likely to make a new investment compared to traders for whom other problems loomed larger. What affects a rural enterprise overall performance in Sri Lanka? Our analysis once again yields some very interesting insights. First, as to be expected, the individual characteristics of rural enterprises such as the experience of managers, the capital intensity of the firms (as measured by its fixed asserts) and the type of firm are significant determinants of productivity as measured by labor productivity (value added per worker) as well as TFP. Not surprisingly, the investment climate also matters. Trading enterprises have significantly higher labor productivity than the production enterprises, but there is no difference in TFP across industry types. Similarly, labor productivity is significantly higher for stand-alone enterprises than household-based enterprises, but TFP is not significantly different. To a large extent, these results are consistent because standalone enterprises also employ more fixed assets than household-based enterprises and likewise, production enterprises utilize more fixed assets than traders. However, enterprises with more informal borrowing capacity and with more experienced and educated managers tend to have both higher labor productivity and TFP. Having access to and using electricity from the grid is associated with a TFP that is 25 percent higher than that of firms not connected to the grid, ceterius paribus. Owning a fixed line or mobile phone, cheaper public transportation costs to the nearest commercial center, the efficiency of the financial sector (as measured by the time taken to clear a check) and access to finance (proxied by the amount that can be borrowed from informal sources) are all significantly 83

30 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP associated with a higher TFP. Owning a fixed line or mobile phone is associated with a 33 percent higher TFP. Enterprises that have access to more efficient financial services had a TFP that was 6 percent higher than firms that had to deal with inefficient financial institutions. In addition to the objective indicators of the prevailing IC, the rural productivity analysis included variables on the perception of entrepreneurs regarding the most important investment climate obstacles faced by them. Although the variables representing perceptions of firms on the most important obstacles are subjective indicators, they provide a relative ranking of the problems faced by different firms and capture other dimensions of access to infrastructure, markets and governance (for example the quality of infrastructure services) that are not reflected in the more objective measures included in the analysis. Perceived problems with electricity, road access, transportation facilities, financing cost and loan procedures, marketing information and low market demand all significantly reduce an enterprise s total factor productivity and labor productivity. It is interesting to note that the magnitude of the impact of the perceived overall constraints on labor productivity and total factor productivity is very similar. The perception that electricity was the most important overall constraint was associated with a 35 percent lower TFP. Similarly poor road quality reduced the level of TFP by 44 percent. When the cost of financing and market demand are perceived as the most important overall constraints they reduce the level of TFP by 43 percent to 53 percent, respectively. Finally, bureaucratic loan procedures and lack of market information also tend to reduce the level of labor productivity and TFP quite considerably however the former is only barely significant at the 10 percent level. Not only does the IC affect the performance of existing enterprises in terms of the investments they make and productivity, but the prevailing climate has a significant impact on whether rural households set-up new enterprises. An analysis of enterprise start-up reveals that enterprises are less likely to be established in communities where access to formal finance sources is limited (as measured by the distance from a community to a bank) and where registration processes are cumbersome (days taken to register an enterprise). On the other hand, households with a large pool of labor and where the family has had prior experience with operating a non-farm enterprise before (as reflected by whether the parents of the household head operated a non-farm business) are more likely to engage in a non-farm enterprise. Being located in a community where electricity is perceived as the most important overall constraint reduces the probability of a household setting up a new enterprise by 18 percent. Similarly, perceptions of low market demand reduces the probability of a non-enterprise household starting up a new enterprise by around 10 percent. Road access, road quality and cost of financing also played a very significant role in determining startup. The probability of a household starting up a business in a community where the cost of financing is viewed as the most important constraint is 6 percent lower than in a community where the cost of financing is not identified as the most important problem. The results also indicate that being located in a community with a larger share of paddy land tends to reduce the probability of individual households starting up a non-farm enterprise. This finding is consistent with the hypothesis that heavy regulation designed to protect paddy production is detrimental to the development of nonfarm business opportunities. Notes 1 Since our focus is to analyze expansion of existing enterprises, we exclude any enterprises that started a business operation in Therefore, enterprises started in 2002 will be the youngest group of the entire sample. 84

31 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Table A5.1 Determinants of Starting Up a Non-farm Establishment (1) (2) Minimum distance to any formal bank *** (2.59) ** (2.36) Maximum amount that can be borrowed from informal sources (log) (1.35) (1.58) Head s age (1.27) (1.41) Maximum level of education attained by any household member (0.13) (0.24) Number of household members with age between 14 and *** (3.57) 0.026*** (3.55) Time taken by the main means of transportation to the main market now (minutes) (0.44) (0.43) Days taken to complete a registration process in your community ** (2.01) ** (2.07) Parents operated non-farm business in the main part of their life 0.071** (2.29) 0.069** (2.31) Household wealth (sum of durable and household values 12 months ago, in log term) 0.016* (1.75) 0.016* (1.77) Per capita total cultivated land owned by household (1.31) (1.37) Share of paddy land of total cultivated land * (1.79) * (1.90) Agriculture as the main income source in the community (0.38) (0.03) Electricity as most important overall constraint *** (6.93) *** (7.09) Road access as most important overall constraint *** (2.74) *** (3.24) Road quality as most important overall constraint * (1.85) ** (2.33) High interest or transaction cost of loan as most important overall constraint ** (2.46) *** (2.94) Loan procedures as most important overall constraint (0.57) (0.89) Lack of market info as most important overall constraint (0.47) (0.03) Low market demand as most important overall constraint *** (4.47) *** (4.73) Telecommunication as most important overall constraint * (1.70) Water supply as most important overall constraint (1.30) Transportation facility as most important overall constraint (0.47) Central Province ** (2.31) ** (2.50) Southern province (0.82) (0.97) North West (0.45) (0.65) North Central (0.49) (0.34) UVA (1.08) (1.27) Sabaragamuwa (0.52) (0.42) North & East Province (0.13) (0.34) Observations Pseudo R-squared Log likelihood Robust z statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 85

32 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Any investment made in year *** (4.29) Amount of investment made in year 2000 (log) Age of establishment 2-5 years *** (2.65) Age of establishment 5-10 years *** (3.38) Age of establishment more than 10 years *** (3.59) Establishment with 3-5 employees (0.82) Establishment with more than 5 employees 0.109** (1.99) Service establishment (1.18) Trade establishment (0.72) Household based enterprises * (1.67) Time taken by the main means of * transportation to the main market now (1.67) (minutes) Mud surface of internal road (1.54) Maximum amount of informal loan can be borrowed (log) (1.31) Education of top manager if there is one (0.53) Years of experience of top manager (1.57) Male top manager (0.50) Sinhalese top manager (0.57) Electricity as most important overall * constraint (1.88) Telecommunication as most important overall constraint (0.98) Water supply as most important overall ** constraint (2.02) Road access as most important overall constraint (1.55) Road quality as most important overall * constraint (1.73) Transportation facility as most important overall constraint (0.51) High interest or transaction cost of loan as most important overall constraint (1.27) Loan procedures as most important overall constraint (1.55) Lack of market info as most important *** overall constraint (2.96) Low market demand as most important overall constraint (1.43) Central Province 0.143** (1.96) Southern Province (0.56) North West 0.282*** (2.88) North Central 0.255*** (2.84) UVA (0.12) Table A5.2 Determinants of New Investment Overall constraints based on perception of individual enterprises Overall constraints aggregated to the community level dprobit tobit dprobit tobit 0.696*** (6.05) *** (3.67) *** (4.54) *** (4.25) *** (3.19) (0.77) 2.158* (1.83) (0.95) ** (2.21) (1.24) ** (2.33) 0.472*** (2.82) (1.21) 0.108** (2.09) (0.49) (1.26) ** (2.07) (0.58) * (1.83) ** (1.98) ** (2.07) (0.34) * (1.81) (1.40) ** (2.06) (1.32) (0.37) (0.84) 7.151*** (4.52) 6.350*** (3.67) ** (2.39) 0.276*** (4.00) ** (2.51) *** (2.94) *** (3.21) (0.78) 0.075* (1.70) (0.80) (1.30) * (1.80) ** (2.22) (0.95) (1.06) (0.23) (1.10) (0.61) (0.52) *** (3.15) ** (2.55) ** (2.37) *** (3.45) (1.11) *** (3.43) *** (3.20) * (1.93) *** (2.75) * (1.92) 0.175** (2.15) 0.143** (2.18) 0.378*** (3.51) 0.348*** (3.60) (1.08) 0.652*** (5.75) *** (3.72) *** (4.50) *** (4.31) *** (3.07) (1.26) (1.37) (1.40) ** (2.21) (1.56) (1.46) 0.437*** (2.65) (0.78) 0.093* (1.86) (0.64) (1.43) *** (4.46) *** (3.12) ** (1.99) *** (4.30) ** (2.40) ** (2.52) *** (4.25) * (1.88) *** (2.67) ** (2.52) (0.67) 3.546** (2.10) 9.408*** (5.56) 8.449*** (4.63) (1.61) 86

33 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Table A5.2 Determinants of New Investment (cont.) Overall constraints based on perception of individual enterprises Overall constraints aggregated to the community level dprobit tobit dprobit tobit Sabaragamuwa 0.131* (1.75) (1.59) 0.169** (2.10) 3.713** (2.05) North & East Province (0.83) 4.562** (2.28) (1.33) 5.385** (2.46) Constant ** (2.29) (0.84) Observations Pseudo R-squared Log likelihood Absolute value of z statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 87

34 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Table A5.3 Determinants of Labor Productivity and Total Factor Productivity Valueadded/worker Total Factor Productivity a Total Factor Productivity b Service Establishment (1.06) (1.21) (1.22) Trade Establishment 0.323*** (2.83) (0.17) (0.21) Age of establishment 2-5 years old (0.11) (0.13) (0.14) Age of establishment 5-10 years old (0.74) (1.25) (1.24) Age of establishment >10 years 0.230* (1.82) (1.48) (1.41) Stand-alone establishment 0.500*** (4.22) (1.30) (1.27) Total fixed assets (log) 0.165*** (3.81) Number of Workers (log) ** ** (2.49) (2.44) Time taken for clearance of a check in this area in the survey month (days) ** ** ** Cost of bus fee to commercial center in the survey month (2.24) (2.31) (2.32) * * (1.81) (1.61) (1.72) Mud surface of internal road (0.64) (0.62) (0.69) Owned fixed line or cell phone 0.264** 0.328** 0.336*** (2.02) (2.50) (2.63) Use electricity from national grid during the past 12 months 0.275** 0.253** 0.248** (2.60) (2.55) (2.52) Electricity line is connected in less than a week (1.49) (1.58) (1.62) Maximum amount that can be borrowed from informal sources 0.043** 0.042** 0.044*** (2.50) (2.45) (2.62) Education of top manager if there is one 0.042** 0.044** 0.041** (2.23) (2.37) (2.16) Years of exprience of top manager (1.47) (1.60) (1.61) Gender of top manager (1.05) (1.28) (1.22) Sinhalese top manager (1.48) (1.50) (1.45) Electricity as most important overall constraint ** ** ** (2.10) (2.24) (2.17) Telecommunication as most important overall constraint (0.38) (0.23) (0.32) Water supply as most important overall constraint (0.99) (1.18) (1.15) Road access as most important overall constraint ** ** ** (2.39) (2.51) (2.37) Road quality as most important overall constraint (0.78) (0.85) (0.76) Transportation facility as most important overall constraint High interest cost of loan as most important overall constraint ** ** ** (2.29) (2.39) (2.25) *** *** *** (3.35) (3.51) (3.32) Loan procedures as most important overall constraint (1.59) (1.61) (1.37) Lack of market info as most important overall constraint Low market demand as most important overall constraint ** ** ** (2.14) (2.07) (2.25) *** *** *** 88

35 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Table A5.3 Determinants of Labor Productivity and Total Factor Productivity (cont.) Valueadded/worker Total Factor Productivity a Total Factor Productivity b (3.61) (3.63) (3.50) Central Province ** (2.11) (0.12) (0.18) Southern Province (0.37) (0.08) (0.05) North Western Province (1.45) (0.59) (0.59) North Central Province *** (2.64) (0.43) (0.47) Uva Province (1.39) (0.27) (0.23) Sabaragamuwa Province ** (1.98) (0.43) (0.42) North and East Province ** (2.40) (0.16) (0.02) Constant 7.895*** (19.83) (0.37) (0.37) Observations R-squared a TFP is calculated from value added production function with constant technology coefficient across production, service and trade b TFP is calculated from value added production function with different technology coefficients for production, service and trade (Industry dummies were interacted with technology coefficients.) Robust t statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 89

36 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Table A5.4 Determinants of Total Factor Productivity by Industry Type TFP is calculated from a value-added production function with industry type interacted with technology coefficients TFP is calculated from a value-added production function which is estimated separately for each industry type Production Service Trade Production Service Trade Age of establishment 2-5 years old (0.91) (0.14) (0.92) (0.28) (0.02) (0.16) Age of establishment 5-10 years old (0.99) Age of establishment > 10 years (0.20) Stand alone establishment (0.02) Number of workers (log) * (1.66) Time taken for clearance of a check in this area in the survey month (days) (1.12) Cost of bus fee to commercial center in the survey month *** (2.72) Mud surface of internal road (0.20) Owned fixed line or cell phone 0.374* (1.75) Electricity line is connected in less than a week (1.37) Maximum amount of credits that can borrowed from 0.069*** informal sources (3.31) Education of top manager if there is one (0.64) Years of experience of top manager (1.21) Gender of top manager 0.315* (1.78) Sinhalese top manager (0.05) Electricity as most important overall constraint ** (2.14) Telecommunication as most important overall constraint 0.601** (2.05) Water supply as most important overall constraint (0.24) Road access as most important overall constraint (1.03) Road quality as most important overall constraint (0.20) Transportation facility as most important overall ** constraint (2.44) High interest cost of loan as most important overall constraint (1.30) (0.68) (0.22) (0.79) (0.09) (1.45) (0.37) (0.85) (1.41) (0.37) (0.10) 0.071** (1.99) (0.30) (1.00) (0.25) (1.63) (1.03) (1.35) *** (3.25) (0.18) (0.32) (0.63) (0.91) * (1.69) (1.33) *** (2.61) (1.14) (0.42) (1.52) 1.050*** (4.93) (0.83) 0.043* (1.75) (1.49) 0.015* (1.69) (0.21) *** (2.79) (0.90) (1.08) (1.59) (0.83) (0.43) (1.01) * (1.87) (0.84) (0.35) (0.99) (1.34) (1.11) *** (2.73) (0.19) 0.383* (1.79) (1.37) 0.069*** (3.31) (0.67) (1.21) 0.315* (1.78) (0.03) ** (2.13) 0.606** (2.07) (0.22) (1.04) (0.19) ** (2.43) (1.30) (0.17) (0.15) (0.14) (0.16) (1.46) (0.34) (0.83) (1.39) (0.37) (0.10) 0.071** (2.00) (0.32) (1.00) (0.24) (1.59) (1.00) (1.31) *** (3.27) (0.18) (0.33) (0.61) (0.44) * (1.85) (0.71) * * (2.34) (1.10) (0.43) (1.51) 1.039* ** (4.86) (0.83) 0.042* (1.70) (1.48) 0.015* (1.68) (0.16) * ** (2.82) (0.93) (1.04) (1.62) (0.84) (0.44) (0.95) * (1.90) Loan procedures as most important overall constraint (0.57) (1.59) (0.02) (0.57) (1.58) (0.01) Lack of market info as most important overall constraint ** (2.06) ** (2.44) (0.60) ** (2.06) ** (2.42) (0.67) Low market demand as most important overall constraint (1.57) *** (2.70) (0.95) (1.58) *** (2.71) (0.99) Central Province 0.715** (2.40) (0.37) (1.24) (1.29) (0.54) (0.40) Southern Province (0.42) (1.46) (0.54) (0.11) (0.38) (0.13) North West (1.07) (0.85) (0.82) (1.02) (0.51) (0.21) North Central (1.07) * (1.71) (0.11) (0.34) (0.68) (0.01) UVA (0.74) (0.06) (0.74) (0.33) (0.36) (0.01) Sabaragamuwa (1.36) (0.52) (0.98) (0.48) (0.18) (0.53) North & East Province 0.585* (1.88) (1.07) (0.73) 0.600* (1.92) (0.78) (1.54) Constant (1.52) (1.14) (0.62) (1.20) (0.62) (0.48) Observations R-squared Robust t statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 90

37 RURAL SURVEY: TECHNICAL APPENDIX ON INVESTMENT CLIMATE, FIRM PERFORMANCE AND START-UP Table A5.5 Estimation of Cobb-Douglas Production Functions Entire sample specification 1 Entire sample specification 2 Production only Service only Trade only Number of workers (log) 0.822*** (8.33) 0.863*** (7.34) 0.525*** (2.88) 0.711*** (3.72) Total fixed assets (log) 0.198*** (4.21) 0.234*** (5.10) 0.238*** (3.71) 0.174*** (3.18) Interaction of number of workers (log) and production enterprises 0.889*** (7.56) Interaction of number of workers (log) and service enterprises 0.523*** (3.03) Interaction of number of workers (log) and trade enterprises 0.762*** (3.94) Interaction of total fixed assets (log) and production enterprises 0.241*** (5.09) Interaction of total fixed assets (log) and service enterprises 0.251*** (3.80) Interaction of total fixed assets (log) and trade enterprises 0.161*** (2.67) Service Establishment 0.242* (1.80) Trade establishment 0.312** 1.328* (2.49) (1.79) Age of establishment 2-5 years old (0.06) (0.05) (0.45) (0.14) (0.69) Age of establishment 5-10 years old 0.283* (1.79) 0.300* (1.87) (0.92) 0.477* (1.70) (0.76) Age of establishment > 10 years 0.428*** (3.37) 0.435*** (3.44) 0.467** (2.45) (1.22) 0.468** (2.20) Stand alone establishment 0.673*** (5.11) 0.649*** (4.99) 0.833*** (5.47) (1.45) 0.528** (2.40) Central Province *** (2.66) *** (2.68) (0.40) (1.19) *** (2.94) Southern Province (0.16) (0.20) (0.10) (1.39) (0.45) North West ** (2.15) ** (2.12) * (1.69) ** (2.46) (0.19) North Central *** (3.41) *** (3.32) (1.49) ** (2.17) (1.56) UVA (0.99) (0.91) (0.16) (0.24) (1.31) Sabaragamuwa * (1.82) * (1.84) (0.21) (1.33) * (1.85) North & East Province * (1.92) * (1.80) (1.46) ** (2.01) (0.62) Constant 7.745*** (14.88) 7.208*** (13.03) 7.147*** (12.83) 8.030*** (9.82) 8.487*** (12.79) Observations R-squared Robust t statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 91

38 APPENDIX 6 RURAL SUMMARY TABLES Table A6.1 Basic Enterprise Characteristics By industry type By number of workers General Information Nation Production Service Trade 1-2 empl. 3-5 empl. > 5 empl. Number of workers (including family members) Enterprises with 3 or more workers (%) Average age of the enterprises (years) How enterprise was started (%) Established by owner Bought Inherited Other % of enterprises that were previously operating in a different location Start up capital % of startup capital from agricultural savings % of startup capital from non-agricultural savings % of startup capital from remittance % of startup capital from asset liquidations % of startup capital from bank loans % of startup capital from private money lender % of startup capital from family/friends % of startup capital from other informal sources % of startup capital from other sources Owner's managerial ability and technical skills Years of prior experience of the owner Ownership type % sole proprietorship in % partnership in % private company limited liability in % public company limited liability in % listed company in % cooperatives in % corporation and government bodies in % others ownership in Other characteristics % registered % stand-alone enterprises % headquarter located in Colombo % previously owned by government/state

39 RURAL SUMMARY TABLE Table A6.2 Use of Utilities and Demand for Business Services By industry type By number of workers Utility use Nation Production Service Trade 1-2 empl. 3-5 empl. > 5 empl. Enterprises using electricity from national grid during past 12 months (%) % using electricity from solar source during past 12 months % using electricity from biomass during past 12 months % using electricity from generator during past 12 months % using electricity from other sources during past 12 months % with no access to electricity source during past 12 months %. that do not use electricity during past 12 months Use water from own well (%) % of water usage from own well Use water from National Water Supply (%) % of water usage from NWS Use water from Pradeshiya Saba (%) % of water usage from PS Use water from individual suppliers (%) % of water usage from individual supplies Use water from other sources (%) % of water usage from other sources % that experienced water shortages Own a phone or cell phone during last year (%) Use fax to communicate with clients during last year (%) Use computer as part of the work during last year (%) Use internet regularly (%) Use regularly (%) Members of any business association (%) Demand for Business Services Engineering service needed (%) Management consultants service needed (%) Marketing services needed (%) Accounting service needed (%) Legal services needed (%) Insurance services needed (%) Information technologies services needed (%) Tech. Support from buyers/suppliers needed (%)

40 RURAL SUMMARY TABLE Table A6.3 Finance By industry type By number of workers Nation Production Service Trade 1-2 empl. 3-5 empl. > 5 empl. Enterprises that have wanted to apply for a formal loan (%) Enterprises that have applied for a loan (%) Enterprise that wanted a loan but did not apply for a loan because (%) Easier to use funds from friends, family & other Interest rate would be too high Duration would be too short Insufficient collateral High cost of application No access to bank Bureaucracy Sources of finance for existing loans (%) Private commercial bank Government commercial bank Rural bank Samurdhi Sanasa Rural Development Bank Friends/neighbors Money lender Other Total Sources of finance for new investments (%) Private commercial bank Government commercial bank Rural bank Samurdhi Sanasa Rural Development Bank Friends/neighbors Money lender Cash-in-hand Other Proportion of enterprises that prepare financial statements (%)

41 RURAL SUMMARY TABLE Table A6.4 Dealings with Government Agencies, Governance Issues and Legal Environment By industry type By number of workers Nation Production Service Trade 1-2 empl. 3-5 empl. > 5 empl. % of enterprise that reported dealing with government agencies % of enterprises that reported dealing with government agencies for: Registration Obtaining/renewing licenses/permits Applying for a construction permit Applying for an electricity connection for industrial use Tax related issues Labor related issues Fire and building safety issues Sanitation/epidemiology issues Environmental regulation Median number of days spent in inspections and meetings with government officials % of enterprises dealing with government agencies reporting making unofficial payments % of enterprises that report laws/regulations that affect their growth and operation are unpredictable or highly unpredictable % of enterprises that agree/strongly agree that laws/regulations can be misinterpreted/manipulated by officials Most important reasons why officials misinterpret/manipulated laws/regulations Officials lack knowledge about the rules/regulations Officials are partial with regards to ethnicity Officials are partial with regards to gender Officials are partial with regards to income status % of enterprises with more than 3 workers that think that enterprises can influence the content of laws and regulations % of enterprises that agree that a contract will protect them from being cheated % of enterprises that agree that the legal system will uphold a contract in a business dispute

42 RURAL SUMMARY TABLE Table A6.5 Share of Firms Assessing Constraints to Operation as Major or Severe by Industry Type and Size (percent) By industry type By number of workers Nation Production Service Trade 1-2 empl. 3-5 empl. > 5 empl. Cost of Finance Loan Procedure Market Demand Electricity Road Quality Road Access Availability of Transportation Market information Water Telecommunication Access to Formal Finance Economic Policy Uncertainty Road Block High Tax Rate Skillful Labor Supply License Cost Registration Cost Registration procedure Figure A6.1 Percent of Enterprises Reporting Major or Severe Constraints Electricity Telecom Water Road access Road quality Financing cost Loan procedures Market information Market demand 96