Lecture IIa. Alvaro Escribano Department of Economics Universidad Carlos III de Madrid

Transcription

1 Lecture IIa Quality of Infrastructures in Africa: A Robust Assessment on Productivity (TFP) based on Investment Climate Surveys (ICs) Course on Inclusive Growth Analytics by Alvaro Escribano Department of Economics Universidad Carlos III de Madrid The Economic Policy and Debt Department (PRMED) of The World Bank Vienna. January 14, 2010

2 Index: 1. Introduction: Evolution of Per capita Income and Labor Productivity in Africa 2. ICs: Quality of the data base in 26 African countries 3. Evaluation of the impact of Infrastructures on productivity 3.1 Estimation of the Infrastructure impacts on productivity 3.2 Infrastructure effects on aggregate productivity, on average productivity and on allocative efficiency 4. Empirical Results 5. Conclusions 2

3 1 Introduction (I) -- Main objective: obtain a robust assessment of the impact of the quality of Infrastructures on productivity of African manufacturing firms. -- Infrastructure is understood in the ICs to include the quality in the provision of customs clearance, energy, water, transportation, telecoms and ICT. -- Infrastructure is one of the usual key blocks of the country specific investment climate assessment and a significant element for country competitiveness. 3

4 1 Introduction (II) -- We want to identify the main bottlenecks in infrastructures that are affecting productivity (TFP), country by country, after controlling for other important blocks of IC variables: b) Red tape, corruption and crime, c) Finance and corporate governance, d) Quality innovation and labor skills and e) other Control variables -- We use 10 different productivity measures for each country (or total factor productivity), Escribano and Guasch (2005) and Escribano et al. (2007) -- Important econometric elements to be considered when identifying infrastructure bottlenecks based on ICs are: I. Heterogeneity of the countries included in our sample (pool?) II. Endogeneity problem at several dimensions (inputs and IC vars) III. Firm level fixed effects (IC vars) IV. Even if we are interested only on Infrastructures, it is important to control for other important blocks of IC variables V. Measurement errors, missing observations, outliers, etc. 4

5 1 Introduction (III): the countries Sample of 26 Countries from North Africa to South Africa: 1) North Africa or Maghreb: Morocco, Algeria and Egypt 2) Economic Community of West African States (ECOWAS): Mauritania, Mali, Niger, Senegal, Burkina Faso, Benin, Cameroon and Cape Verde 3) The Horn of Africa: Eritrea and Ethiopia 4) East African Community (EAC): Tanzania, Kenya, Uganda and Burundi 5) South African Development Community (SADC): Malawi, Zambia, Namibia, Botswana, Swaziland, Lesotho and Madagascar 6) South Africa Mauritius Heterogeneity among countries -- Geographical factors: insular countries/tropical countries/desert countries/ coastal and landlocked countries, etc -- Social or political factors: civil wars/armed conflicts/early democracies/dictatorships/colonial heritage -- Economic factors: from the most successful (Mauritius) to the poorest (Eritrea). 5

6 1 Introduction (IV) Evolution of the GDP per capita of the countries considered in the report ( ) US $ MUS SWZ ZAF BWA DZA NAM CPV EGY MAR CMR LSO MRT SEN BEN KEN MLI UGA BFA ZMB TZA NER MWI BDI MDG ETH ERI 6

7 ERI 1 Introduction (V) Five-year rate of growth of GDP per capita, ( ) % MUS SWZ ZAF BWA DZA NAM CPV EGY MAR CMR LSO MRT SEN BEN KEN MLI UGA BFA ZMB TZA NER MWI BDI MDG ETH

8 1 Introduction (VI) Evolution of Per Capita Income in Africa relative to U.S, A. GDP per capita MUS SWZ ZAF BWA DZA NAM CPV EGY MAR CMR LSO MRT SEN BEN KEN MLI UGA BFA ZMB TZA NER MWI BDI MDG ETH ERI 8

9 1 Introduction (VII) Evolution of Labor Productivity in Africa relative to U.S, B. Labor productivity (GDP per worker) MUS SWZ ZAF BWA DZA NAM CPV EGY MAR CMR LSO MRT SEN BEN KEN MLI UGA BFA ZMB TZA NER MWI BDI MDG ETH ERI 9

10 1 Introduction (VIII) Evolution of the Demographic Factor in Africa relative to U.S, C. Demographic factor MUS SWZ ZAF BWA DZA NAM CPV EGY MAR CMR LSO MRT SEN BEN KEN MLI UGA BFA ZMB TZA NER MWI BDI MDG ETH ERI Year 1960 Year 1970 Year 1980 Year 1990 Year

11 1 Introduction (IX): Perceptions Percentage of firms that considers telecommunications and electricity as a severe or very severe obstacle to economic performance by country (I) A. Telecommunications B. Electricity Benin Kenya Zambia Niger Ethiopia Lesotho Madagascar Burkina Faso Algeria Malawi Cameroon Tanzania Swaziland Mali Cape Verde Eritrea Botswana Burundi Mauritania Namibia Uganda Mauritius South Africa Senegal Egypt Morocco Percentage of firms Burundi Cameroon Benin Burkina Faso Cape Verde Tanzania Malawi Kenya Madagascar Uganda Ethiopia Mauritania Zambia Eritrea Lesotho Senegal Mali Swaziland Niger Egypt Algeria Mauritius Morocco South Africa Botswana Namibia Percentage of firms 11

12 1 Introduction (X): Perceptions Percentage of firms that considers customs and transport as a severe or very severe obstacle to economic performance by country (II) C. Customs clearance D. Transport Benin Kenya Madagasc Senegal Algeria Cameroon Tanzania Ethiopia Zambia Uganda Niger Malawi Mauritius Mauritania Mali Egypt South Lesotho Cape Burundi Burkina Morocco Namibia Swaziland Botswana Eritrea Percentage of firms Burkina Benin Cameroon Kenya Senegal Zambia Malawi Niger Burundi Madagascar Tanzania Uganda Cape Verde Lesotho Mali Mauritania Namibia Swaziland Ethiopia Mauritius South Africa Botswana Eritrea Morocco Egypt Algeria Percentage of firms 12

13 1 Introduction (XI) A simple illustration based on cross-plots of the relation between GDP and Infrastructure Perceptions in Africa A. Telecommunications vs. GDP per capita B. Electricity vs. GDP per capita % of firms considering telecoms as a severe % of firms considering electricity as a severe o GDP per capita (% of US) GDP per capita (% of US) B. Customs clearance vs. GDP per capita D. Transport vs. GDP per capita % of firms considering customs as a severe % of firms considering transport as a severe o GDP per capita (% of US) GDP per capita (% of US) 13

14 2 Data (I) Summary of IC surveys sorted by geographical areas and classification of countries by pools for estimation purposes N orth A frica E conom ic Com m unity of W est A frican S tates H orn of A frica E ast A frican Com m unity S outhern A frican D evelopm ent Com m unity Y ear of the survey Years of production function variables Total num ber of observations 1 A lgeria E gypt ,629 M orocco ,422 S enegal B enin M ali C ape Verde M auritania* B urkina Faso* N iger* C am eroon* E thiopia** ,142 E ritrea** K enya U ganda Tanzania B urundi M alaw i M adagascar Zam bia Lesotho B otsw ana*** N am ibia*** Final num ber of observations available for regression analysis 2 S waziland*** M auritius S outh A frica ,492 1 Total number of observations is equal to the total number of firms surveyed multiplied by the total number of years. 2 The observations available for regression analysis are the total number of observations minus the observations with any PF variable missing and/ or outlier after the cleaning process. 3 Countries for which no regression analysis was done. *, **, *** mean that the information of these countries was pooled together for regression analysis. 14

15 2 Data (II) Total number of observations available for the production function (PF) variables before and after cleaning missing values and outliers (I) Total number of observations Missing observations of which: firms with one PF variable missing firms with two PF variables missing firms with three PF variables missing firms with four PF variables missing North Africa Western Africa - Economic Community of West African States DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR , Outliers of which: outliers in materials outliers in labor cost outliers in both materials and labor cost Useful observations (no outliers and missing) Missing observations of which: firms with one PF variable missing firms with two PF variables missing firms with three PF variables missing firms with four PF variables missing outliers in materials outliers in labor cost outliers in both materials and labor cost Useful observations (no outliers and missing) 316 1,317 2, Outliers of which: ,629 2,

16 2 Data (III) Total number of observations available for the production function (PF) variables before and after cleaning missing values and outliers (II) Total number of observations Missing observations of which: firms with one PF variable missing firms with two PF variables missing firms with three PF variables missing firms with four PF variables missing Horn of Africa East Africa - East African Community Southern Africa - Southern African Development Community ETH ERI KEN UGA TZA BDI MWI MDG ZMB BWA LSO NAM SWZ MUS ZAF Outliers of which: outliers in materials outliers in labor cost outliers in both materials and labor cost Useful observations (no outliers and missing) Missing observations of which: firms with one PF variable missing firms with two PF variables missing firms with three PF variables missing firms with four PF variables missing outliers in materials outliers in labor cost outliers in both materials and labor cost Useful observations (no outliers and missing) 1, , Outliers of which: , ,492 16

17 2 Data (XIV): Cleaning process Cleaning of Production function (PF) variables: An important number of observations were not available for regression analysis due to missing observations or outliers. --We exclude those plants with missing values in all the production function variables sales, materials, capital stock and labor cost. --We transform outlier observations into missing observations and we proceed as follows: a) we replace the missing values by the corresponding (cells) industry-region-size median of the variable; b) if we do not have enough observations in some cells, we replace them by the corresponding industry-size medians; c) if we still do not have enough observations in those cells, we replace them by the region-size medians; d) if still necessary, in the last step we compute the medians only by size and/or by industry to replace those missing values. --We require to have at least 15 firms per cell to compute the median. 17

18 2 Data (XV): Cleaning process Cleaning of IC variables: --Large amount of observations were not available for regression analysis due to missing values. --Two strategies to solve for this problem: I. Create Industry-Region averages of plant level ICA variables (the response rate of the variable should be larger than 30%) II. Replace only the missing values by the industry-region averages (the response rate of the variable should be larger than 85%) --There must be at least 15 firms per cell to create the averages. --If the number of regions and/or cells is not large enough to create heterogeneity in the industry-region averages, we use the industry-region-size average instead (example: countries with only four industries and two regions). 18

19 3.1 Estimation of the Infrastructure impact on productivity (III) Question: Is it possible to obtain robust infrastructure impacts on productivity based on IC Surveys for several productivity measures? Answer: YES, if we control for fixed effects. 19

20 3.1 Estimation of the Infrastructure impact on productivity (IV) Summary Table of Productivity Measures and Estimated Investment Climate (IC) Elasticities Two Step 1. Solow s Residual Estimation 1.1 Restricted Coef 1.1.a OLS 1.1.b RE 1.2.a OLS 1.2 Unrestricted Coef 1.2.b RE 2.1.a OLS Single Step 2.1 Restricted Coef 2.1.b RE 2. Cobb-Douglas Estimation 2.2.a OLS 2.2 Unrestricted Coef 2.2.b RE 3.1.a OLS Single Step 3.1 Restricted Coef 3.1.b RE 3. Translog Estimation 3.2.a OLS 3.2 Unrestricted Coef 3.2.b RE Total 2 (P it ) measures 4 (IC) elasticities 4 (P it ) measures it 4 (IC) elasticities 4 (P it ) measures 4 (IC) elasticities 10 (P it ) measures 12 (IC) elasticities Restricted Coef.= Equal input-output elasticities in all industries of the three countries Unrestricted Coef.= Different input output elasticities by industry of the three countries OLS = Pooling Ordinary Least Squares estimation (with robust standard errors) RE = Random Effects estimation 20

21 3.2 Infrastructure assessment based on the Olley and Pakes (O&P) decomposition (I) Question: How could we evaluate the impact of each block of IC and C variables on productivity? Olley and Pakes (O&P) decomposition of aggregate productivity: INF effects by Industry, region, size, age of the firm, etc. Allocative efficiency decompositions. When pooling countries to estimate TFP, we can use the O&P decomposition to do country by country, or industry by industry, TFP evaluation. 21

22 4. Firm Perceptions (I) Rankings on firms perceptions (I) A. Total relative weights 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% CPV BDI BFA MRT CMR ETH NER KEN LSO MWI BEN SWZ ERI UGA ZMB TZA MDG SEN NAM MLI DZA BWA ZAF MAR MUS EGY Infrastructures Red tape, corruption and crime Finance Labor skils 22

23 4. Firm Perceptions: Weighted (II) Rankings on firms perceptions (II) B. Average group relative weights 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% CPV CMR BFA BDI ETH NER KEN BEN MRT MWI TZA ZMB LSO MDG UGA SEN SWZ MLI ERI DZA NAM ZAF BWA EGY MUS MAR Infrastructures Red tape, corruption and crime Finance Labor skils 23

24 Empirical Results (III) 14.0 Olley and Pakes decomposition in levels CPV CMR MAR LSO BEN ZAF BFA BWA MUS NAM NER BDI ERI TZA UGA EGY SWZ SEN DZA MWI MLI MDG ETH KEN MRT ZMB Aggregate Productivity Average Productivity Allocative Efficiency

25 4. Empirical Results in logs (IV) Olley and Pakes decomposition in logs ZAF MUS CPV NAM MAR BWA CMR BFA BDI LSO NER SWZ BEN SEN TZA EGY UGA ERI MLI MDG MWI DZA KEN MRT ETH ZMB Aggregate Productivity Average Productivity Allocative Efficiency

26 4. Empirical Results: Demean (V) Demeaned Olley and Pakes decomposition in levels ZAF MUS BWA DZA EGY NAM SWZ MDG MAR ETH KEN SEN ERI MLI BFA CMR NER ZMB MRT BEN UGA MWI TZA Aggregate Productivity Average Productivity Allocative Efficiency

27 4. Empirical Results: Demean (VI) Demeaned Olley and Pakes decomposition in logs ZAF MUS EGY BWA NAM DZA SWZ MAR MDG ETH ERI SEN MLI NER BFA KEN ZMB CMR MRT UGA MWI BEN TZA Aggregate Productivity Average Productivity Allocative Efficiency

28 4 Results (IX): Infrastructures and other IC vars Simulations of infrastructure absolute effects on productivity (20% improvement) A. Aggregate Productivity 100% 80% 60% 40% 20% 0% MWI UGA BEN SEN ETH ZMB ERI MLI CMR DZA MRT BFA NER TZA KEN MAR MDG ZAF EGY SWZ MUS BWA NAM Infrastructures Red tape, corruption and crime Finance and corporate governance Quality, innovation and labor skils Other control variables 28

29 4 Results (XI): Infrastructures and other IC vars Simulations of infrastructure absolute effects on productivity (20% improvement) 100% B. Average Productivity 80% 60% 40% 20% 0% MWI UGA BEN SEN ETH ZMB ERI MLI CMR DZA MRT BFA NER TZA KEN MAR MDG ZAF EGY SWZ MUS BWA NAM Infrastructures Red tape, corruption and crime Finance and corporate governance Quality, innovation and labor skils Other control variables 29

30 4 Results (XII): Infrastructures and other IC var Simulations of infrastructure absolute effects on productivity (20% improvement) 100% C. Allocative Efficiency 80% 60% 40% 20% 0% SEN MWI DZA MLI KEN UGA ZMB TZA NAM BFA ETH MAR CMR MDG MRT ERI SWZ ZAF EGY MUS BEN NER BWA Infrastructures Red tape, corruption and crime Finance and corporate governance Quality, innovation and labor skils Other control variables 30

31 4 Results (XIII): Infrastructures and other IC var Infrastructure absolute effects on productivity: demeaned O&P decomposition in logs 100% A. Aggregate Productivity 80% 60% 40% 20% 0% MWI ERI UGA MLI SEN NAM CMR ETH TZA NER SWZ MRT ZMB BEN MUS BWA DZA MAR EGY BFA MDG ZAF KEN Infrastructures Red tape, corruption and crime Finance and corporate governance Quality, innovation and labor skils Other control variables 31

32 4 Results (XIV): the Infrastructure within the IC Infrastructure absolute effects on productivity: demeaned O&P decomposition in logs 100% B. Average Productivity 80% 60% 40% 20% 0% ERI MWI SEN UGA MLI DZA BEN ETH TZA NER CMR MRT BFA SWZ KEN BWA NAM MUS EGY ZAF MAR ZMB MDG Infrastructures Red tape, corruption and crime Finance and corporate governance Quality, innovation and labor skils Other control variables 32

33 4 Results (XV): the Infrastructure within the IC Infrastructure absolute effects on productivity: demeaned O&P decomposition in logs 100% C. Allocative Efficiency 80% 60% 40% 20% 0% MWI SEN UGA NAM MLI TZA DZA ERI ETH BEN CMR KEN NER SWZ MRT EGY ZMB MAR MDG MUS BFA ZAF BWA Infrastructures Red tape, corruption and crime Finance and corporate governance Quality, innovation and labor skils Other control variables 33

34 4 Results (XXV): Summary of results (3) Infrastructure impact on average log-productivity by key factors Customs clearance includes: days to clear customs to export and import; shipment losses in customs; inspections in customs; wait for an import license. Electricity includes: power outages; avg. duration of power outages; losses due to power outages, wait for an electricity supply; power fluctuations; avg. duration of power fluctuations; cost of electricity from the public grid; cost of electricity from private system, dummy for own generator; electricity from own generator; dummy for own power infrastructures (excl. generators). Water includes: water outages; avg. duration of water outages; losses due to water outages, wait for a water supply; cost of water from the public grid; cost of water from private system, dummy for own water infrastructures; water from own well. Telecomm and ICT: phone outages; avg. duration of phone outages; losses due to phone outages, wait for a phone connection, dummy for ; dummy for webpage Transportation includes: sales lost due to transport delays; sales lost due to delivery delays; shipment losses; low quality supplies; transport delays, dummy for own roads; dummy for own transportation for workers; products with own transport. 34

35 4 Results (XXV): Summary of results (4) 100% Infrastructure impact on average log-productivity by key factors 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% ZMB SEN TZA MLI MDG NAM BWA MWI ETH SWZ UGA ERI KEN NER BFA EGY MRT CMR ZAF DZA BEN MUS MAR Customs clearance Electricity Water Telecomms and ICT Transportation 35

36 4 Results (XXV): Summary of results (4) Infrastructure impact on average log-productivity by key factors 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% MWI UGA BEN SEN ETH ZMB ERI MLI CMR DZA MRT BFA NER TZA KEN MAR MDG ZAF EGY SWZ MUS BWA NAM Customs clearance Provision of electricity Use of generator or power infrastructures Provision of water Use of own water infrastructure Provision of phone Use of ICT Transport services Own transport infrastructures 36

37 4 Results (XXIII): Based on Rankings (1) R an kin g b ased o n p er cap ita G D P R an kin g D B R 0 7 (ran k w ith in sam p le) Summary of results based on Rankings R an kin g A C R 0 7 (ran k w ith in sam p le) D em ean A g g reg ate P ro d u ctivity (ran k) F irm s' p rcp tn s. % ab s. co n trib.. (ran k) P ro d u ctivity A verag e p ro d. % ab s. co n trib. (ran k) A llo cativ e effic ien cy % ab s. co n trib.. (ran k) M U S (1) 3 2 (2 ) 4.2 (1 ) 2.0 (2) 13.9 (2) 24.6 (3 ) 12.4 (4 ) S W Z (2) 7 6 (5 ) n.a 1.4 (7) 22.4 (10 ) 25.6 (4 ) 15.0 (7 ) Z A F (3 ) 2 9 (1 ) 4 (3 ) 2.3 (1) 16.2 (5) 28.6 (6 ) 14.9 (6 ) B W A (4) 4 8 (4 ) 3.4 (6 ) 1.7 (3) 15.6 (4) 17.5 (2 ) 6.7 (1 ) D ZA (5 ) 1 16 (12 ) 2.9 (7 ) 1.5 (4) 18.3 (7) 40.0 (1 4) 46.4 (2 1) N A M (6) 4 2 (3 ) 4.2 (2 ) 1.5 (6) 18.3 (6) 16.5 (1 ) 24.6 (1 5) E G Y (7 ) 1 65 (22 ) 3.7 (4 ) 1.5 (5) 14.0 (3) 26.0 (5 ) 12.9 (5 ) M A R (8) 1 15 (11 ) 3.6 (5 ) 1.1 (9) 9.9 (1) 31.3 (8 ) 20.4 (1 2) C M R (9) 1 52 (18 ) 1.9 (18 ) 0.8 (16) 27.5 (23 ) 41.6 (1 5) 20.3 (1 1) M R T (10) 1 48 (16 ) 2.1 (15 ) 0.6 (19) 25.3 (17 ) 35.4 (1 3) 18.4 (9 ) S EN (1 1) 1 46 (15 ) n.a 0.9 (12) 22.7 (11 ) 58.5 (2 0) 54.3 (2 3) B E N (1 2) 1 37 (13 ) 2.1 (11 ) 0.6 (20) 25.6 (18 ) 59.9 (2 1) 10.9 (3 ) K E N (1 3) 8 3 (6 ) 2.8 (8 ) 1.0 (11) 25.6 (19 ) 31.8 (9 ) 43.9 (1 9) M L I (14) 1 55 (19 ) 2.1 (14 ) 0.9 (14) 21.6 (9) 42.7 (1 6) 44.4 (2 0) U G A (15 ) 1 07 (9 ) 2 (1 7) 0.6 (21) 23.3 (12 ) 60.4 (2 2) 33.3 (1 8) B FA (1 6) 1 63 (21 ) 2.1 (12 ) 0.8 (15) 26.9 (22 ) 35.3 (1 2) 20.9 (1 4) Z M B (17) 1 02 (8 ) n.a 0.7 (18) 24.0 (14 ) 50.6 (1 8) 28.0 (1 7) T ZA (18 ) 1 42 (14 ) 2.7 (9 ) 0.2 (23) 24.3 (15 ) 34.1 (1 0) 25.0 (1 6) N E R (1 9) 1 60 (20 ) n.a 0.8 (17) 26.2 (20 ) 34.2 (1 1) 10.5 (2 ) M W I (2 0) 1 10 (10 ) 2.1 (13 ) 0.4 (22) 24.5 (16 ) 65.9 (2 3) 47.9 (2 2) M D G (21 ) 1 49 (17 ) 2 (1 6) 1.4 (8) 23.5 (13 ) 30.6 (7 ) 19.0 (1 0) E TH (2 2) 9 7 (7 ) 2.3 (10 ) 1.0 (10) 26.7 (21 ) 52.6 (1 9) 20.6 (1 3) E R I (23) 1 70 (23 ) n.a 0.9 (13) 20.7 (8) 46.1 (1 7) 18.0 (8 ) 37

38 4 Results (XXIV): Summary of results (2) Cross-plot: GDP per capita (% of US) and demean aggregate productivity Cross-plot: INF impact on average log productivity and demean aggregate productivity GDP per capita (% of U Demean Aggregate Productivity % INF impact on aggregate log-produ Demean Aggregate Productivity Cross-plot: INF impact on average log allocative efficiency and demean aggregate productivity 60.0 Cross-plot: Ranking on the Easy of Doing Business 2007 and demean agg. productivity 180 Cross-plot: Quality of overall Infrastructure from Africa Competitiveness Report 2007 and demean agg. productivity 4.5 Percentage Absolute contribution to allo Efficiency Ranking Easy of Doing Business Quality of overall Infrastructure ACR Demean Aggregate Productivity Demean Aggregate Productivity Demean Aggregate Productivity 38

39 5 Conclusions (I) The ranking from demeaned aggregate productivity is consistent with the ranking from the Ease of Doing Business (2007) and with the per capita income ranking Within the investment climate variables, the largest contribution of infrastructure is in the Sub-Saharan countries The higher is the contribution of infrastructures on productivity, the lower will be the per capita income of the country. Those results are robust to several productivity measures: Olley and Pakes decomposition in logs or in levels (simulations). 39

40 5 Conclusions (II). The highest (average) productivity impacts of Infrastructures are in relatively poor countries: Senegal, Mali and Malawi, Tanzania.. The impacts of Infrastructures in the relatively most successful countries (say Mauritius, South Africa, North African countries, Swaziland, Botswana) tends to be dominated by the positive IC factors such as use of IC technologies. The countries with the highest Infrastructure impacts tends to be, the most sensitive to changes in the investment climate conditions (simulation exercises). Within the Infrastructure block the most prevalent factor is the low quality of the provision of electricity. The contributions of low quality of the provision of water and phone are in general lower. 40

41 5 Conclusions (III) The use of alternative power infrastructure by firms tends to mitigate the problems that comes from the low quality of the provision of electricity Problems with the transportation of products are negatively related with the productivity in all the cases (with the exception of Botswana, Swaziland and Namibia for which no variables of this group were significant in the productivity regressions) The largest impacts of transportation are in Senegal, Tanzania, Madagascar, South Africa and Zambia To compensate for the bad water infrastructure, firms replace the public provision by own wells, having a positive impact on productivity. 41

42 5 Conclusions (IV) The firm use of own transport infrastructures enhance productivity. The positive effects on productivity of having their own transport facilities are concentrated only in eight countries: Malawi, Tanzania, Kenya, Eritrea, Benin, Senegal and Egypt. Another infrastructure sub- group of IC variables with significant and high impact on the average log-productivity is customs clearance (either with imports or exports) the contributions of this group is negative and very large in most of the countries 42