Financial Silk Road to Africa

Financial Silk Road to Africa Jacopo Ponticelli Andrea Presbitero September 21, 2017 PRELIMINARY and INCOMPLETE Abstract Notwithstanding the increasing importance of Chinese development finance for several African countries, there is still limited systematic evidence of its impact on recipient countries. We provide new micro-level evidence on the effects of Chinese financing on local economies by matching more than 2,400 project-level data on Chinese official development finance with firm-level data on more than 11,000 firms in 32 Sub-Saharan African countries. We exploit differences in Chinese development finance across sub-national regions, and input-output linkages to show that Chinese-financed development projects had a positive effect on firm sales and labor productivity in recipient economies, especially those operating in non-tradable sector and those operating along the production chain of the development projects. Keywords: Development finance; China; Africa; Firms; Growth. JEL Classification: O12; O14; O22; O55. This research is part of a project on Macroeconomic Research in Low-Income Countries (project id: 60925) supported by the U.K. s Department for International Development (DFID). The views expressed in this paper are those of the authors and do not necessarily represent those of the IMF, IMF policy, or of the DFID. Tao Chen and Ariza Gusti provided excellent research assistance. Ponticelli: Northwestern University and CEPR, jacopo.ponticelli@kellogg.northwestern.edu. Presbitero: International Monetary Fund and MoFiR, APresbitero@imf.org. 1

I Introduction China development finance targeting African countries has grown exponentially since the early 2000s. Figure I reports total Chinese lending to African countries, as well as its composition, between 2000 and 2014. As shown, Chinese lending can take different forms: loans issued by Chinese development banks and other state-owned firms, grants, technical assistance and export credits. In particular, bank loans account for 70% of total Chinese development finance in Africa since 2000, and have grown from less than 1Bn USD in 2000 to 16 Bn USD in 2010, with a peak of 29 Bn USD in 2012 (AidData 2017, numbers in constant 2014 USD). Chinese development finance can improve local infrastructure and promote economic growth in African countries. 1 However, despite the increasing magnitude of these financial flows, there is scarce empirical evidence on its effects on recipient economies. Figure I: Chinese Development Finance to African Countries, by Type Billion 2014 constant USD 0 10 20 30 Loans Grant Debt forgiveness Technical assistance/training Export credits Other 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: AidData, 2017 In this paper we study the effect of Chinese development finance on firms operating in recipient economies. Whether development finance should benefit or harm local firms is an open empirical question. The data shows that companies implementing Chinese-financed projects in Africa are often Chinese firms. In addition, the financing of the project itself can be conditional on using Chinese contractors or Chinese imported materials. Therefore, to the extent that development finance is implemented by foreign companies and favors import of Chinese goods, it might have no positive effect on local firms or even penalize those operating in the tradable 1 According to recent estimates, in Africa there are more than 10,000 Chinese firms, and China is the biggest trade partner and the largest bilateral infrastructure financier in the continent (Sun et al., 2017). 2

sector. As a matter of fact, during the period under study, Chinese exports to African countries increased from 20Bn USD in 2000 to 237Bn USD in 2010, and totaled 419Bn USD in 2014 (see Figure II). 2 In addition, there is a positive correlation between flows of development finance and Chinese exports to recipient countries, as shown in Figure III Figure II: Chinese Exports to African Countries 400 300 Billion USD 200 100 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: COMTRADE. One the other hand, there are several channels through which development finance can benefit firms in recipient economies. First, it can improve local well-being in regions where projects are implemented by generating jobs and increasing wages, which should benefit local firms operating in the non-tradable sector. Second, to the extent that some inputs are sourced locally, it can generate positive demand shocks and incentive to upgrade technology for firms operating along the production chain of the development project. Finally, it can ameliorate local infrastructure for example by providing better road connections or more reliable electricity supply which in turn benefits local firms. To test these channels, we use a new loan-level dataset tracking Chinese-financed projects in African countries between 2000 and 2014. We match this dataset with firm-level data from the World Bank Enterprise Survey as well as shipment data at product and destination-country level covering all Chinese exports to African countries. We find positive effects of Chinese-financed development projects on local well-being. Absent reliable measures of income per capita at local level, we use sales by firms operating in the nontradable sector as a measure of local consumption, under the assumption that non-tradable goods tend to be consumed locally. We find that firms located in regions receiving Chinese- 2 African countries constitute a small fraction of Chinese exports to the rest of the world in terms of volume, bust their share has also increased in the last decade, going from 1% of total value of Chinese exports in 2000, to 1.9% in 2010, and to 2.2% in 2014. 3

Figure III: Chinese official development finance and exports to recipient countries ODF from China (as a % of GDP, average 2000-14) 5 4 3 2 1 0 ZWE TON FJI WSM DMA VUT TKM CUB ERI ETH MRT GRD LAO COG GUY VEN NER JAM MDV LKA CIV SLE MOZ KEN GHA TCD BDI CMR AGO MNE MUS CAF BLR RWA COM SSD GNB MLI BOL ECU MWI SDN ZMB SUR LCA GIN MKD SRB KAZ MMR GNQ SYC AFGPAK UGA BIH CPV ARGLSO TZA UZB GAB ALB CRI NICSEN NPL NAM TLS BHS MDG BGD MDA ARM BRA BGR COD RUS LBY NGA PNG YEM AZE BRB IRN UKR PSEROU TTO CHL COL TUR IRQ DZA HTI GEO IND MAR MEX ISR AUS EGY IDN JORMYS KWT STP TUN PER NZL ZAF SYR LBN BRN URY BHR SOM CYP THA Chinese official development finance and exports KHM VNM SGP MNG TJK LBR 0 10 20 30 40 50 Imports from China (as a % of GDP, average 2000-14) MLT BEN DJI TGO KGZ ATG Source: AidData China Global Data and IMF Direction of Trade Statistics. financed development projects experienced larger increases in sales, and these effects are larger for firms operating in the non-tradable sector relative to those operating in the tradable sector. Next, we investigate if Chinese development finance generate positive demand growth for local input-supplier that operate along the production chain of the development project. To this end, we build a measure of exposure of local firms to the Chinese-driven demand shock for inputs. We find that firms with larger exposure to Chinese-finance projects experienced larger increase in sales and labor productivity. Taken together, our results indicate that Chinese-financed development projects had a positive effect on firm sales and labor productivity in recipient economies, especially those operating in non-tradable sector and those operating along the production chain of the development projects. The timing of these effects suggest tat they are not driven by pre-existing trends and that they are relatively short lived. To the best of our knowledge, this is the first paper to analyze with micro-data the effect of Chinese development finance on firms operating in recipient countries. While there are works that look at the macroeconomic effects of Chinese aid in a cross country framework (Busse et al., 2016; Kilama, 2016), Dreher et al. (2016) is the closest to our analysis. They use nighttime light emissions at the regional level in Africa to show a positive short-run impact of Chinese aid on local development across African sub-national units. While their approach is intended to test for the economic consequences of political favoritism, our analysis directly looks at the firm response to an increase in Chinese development finance, exploiting both regional variations and input-output linkages. In this way, we are able to provide novel evidence on the mechanisms through which Chinese financed development projects could affect local economies. 4

Our paper builds also on extensive literature on aid effectiveness (see Rajan and Subramanian, 2011; Clemens et al., 2012; Dreher and Lohmann, 2015; Galiani et al., 2017, among others). 3 Finally, we also contribute to the emerging literature on Chinese lending, which so far mostly focuses on its drivers (Chen et al., 2016; Dreher et al., 2016). II Background and Data Description We merge project-level data on Chinese official development assistance to Africa with firmlevel data from the World Bank Enterprise Survey (WBES). 4 The match is done at the region and sector level. We use the location variable ( a3ax ) of each firms in the WBES and convert it to the standard classification at the first sub-national administrative level (ADM1). We are label to localize almost all (30,313) firms out of the 30,965 firms in 170 ADM1 region in Sub-Saharan Africa. We do the same exercise for Chinese projects: in this case we complement the information available in the original dataset which we convert into ADM1 regions with information on location that we can obtain from the title and the description of the project. We are able to localize 1,363 projects in 282 Sub-Saharan African regions. Figures IV and V show the spatial distribution of firms and Chinese projects, respectively. Figure IV: Firm location in Sub-Saharan Africa Source: World Bank Enterprise Survey; sub-national regions are defined at the ADM1 level. 3 See Qian (2015) for a review of the evidence of the impact of aggregate foreign aid and for a discussion of the problems related to examing aggregate aid flows. 4 We use the Standardized Dataset available online at www.enterprisesurveys.org, downloaded in June 2017. 5

Figure V: Location of Chinese projects in Sub-Saharan Africa Source: AidData; sub-national regions are defined at the ADM1 level. At the sector level, we take the ISIC classification (Rev. 3.1) as reference for the match, since it is already present (at 4-digit) in the WBES. We manually allocate the Chinese investment project to the corresponding ISIC sector (2-digit) using the information included in the title and description of the project. The match is done considering 17 sectors. 5 Data on Chinese development finance to Africa come from AidData: Tracking Chinese Development Worldwide, a new dataset that contains project-level information on Chinese official development finance (loans, grants, debt forgiveness, export credit) to Africa, Asia and Latin America from 2000 to 2014. 6 Africa is the largest recipient, both in terms of number of projects and value, especially since 2011 (Figure VI). Overall, the dataset includes 2,469 projects financed in Africa since 2000, with Zimbabwe, Angola, Tanzania, Ghana, Liberia, Kenya, Ethiopia, and Zambia being the countries that attracted the largest number of projects (Table I). Contrary to the common wisdom that Chinese investments are concentrated in oil, gas, and extractive industries, the sectoral distribution of Chinese projects shown in Table II points to a much more diversified investment pattern across sectors, with a prevalence of 5 The sectors, constructed aggregating the 2-digit ISIC ones, are: Agriculture, forestry and fishing; Mining; Manufacturing; Electricity, gas, and water supply; Construction; Wholesale and sale of motor vehicles; Retail trade, except of motor vehicles; Hotel & restaurants; Transport and storage; Information and communication; Financial intermediation and real estate and renting; Business activities; Public administration; Education, health and social works; Other services; Private household activities; Extra-territorial organizations. 6 The dataset builds on a previous version, which collects data until 2011 and is carefully discussed by Strange et al. (2017). We exclude all projects that are categorized as pledges, and those that have been suspended or canceled. 6

Figure VI: Chinese Development Finance: Number of Projects and Value Number of transactions Values 500 100 Number of transactions 400 300 200 100 Billion 2014 constant USD 80 60 40 20 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Latin America and the Caribbean Southeast Asia Middle East Central and North Asia Africa South Asia Europe The Pacific Latin America and the Caribbean Southeast Asia Middle East Central and North Asia Africa South Asia Europe The Pacific Notes: The left-hand side panel reports the regional distribution of development projects, while the right-hand side chart plots the corresponding vales, in billion of 2014 constant USD; some projects recorded in the dataset do not have information on loan amount. All canceled and suspended transactions are excluded, as well as pledges of official finance. Data on development finance include all flows (loans, grants, debt forgiveness, export credit, etc.). Source: AidData China Global Data. projects in manufacturing, health and social works, and education (Parks, 2015). 7 Since we are interested in improvements in local well-being, and the WBES has a limited panel dimension, we exploit the fact that each survey asks the level of total sales and employment in the year of the survey and three fiscal year before. Additional documentation provided by the World Bank specifies the exact year to which the question refer to. In most cases, the gap between the two points in time is two calendar year, while in a few cases the difference is three ears. For uniformity, we consider all the survey with the same gap in the reporting of current and past levels of employment and sales and we chose two years to maximize the sample size. In this way, we can compute the percentage change in employment, sales and labor productivity (defined as sales per employee) over a three year period. III Empirics In this section we study the effect of Chinese development finance on firms operating in African recipient economies. We investigate two main channels through which development finance can affect local firms. First, we study the effect of development finance on local firms through increases in local demand. To the extent that Chinese investments create new jobs and higher wages in the regions where they are implemented, they can translate into higher consumption and, therefore, higher demand for firms operating in the non-tradable sector. Second, we study the effect of development finance on local firms through input-output linkages. Chinese investment in certain sectors can generate demand for inputs produced by firms in the 7 Similar evidence on the lack of concentration of investment in resource extraction is also discussed by McMillan (2017) with respect to foreign direct investment. 7

Table I: Chinese Development Finance to Africa, Top Recipients Notes: The table reports the top recipients of Chinese development finance, in terms of number of projects, between 2000 and 2014. The last column reports also the total value of projects received by each country, in billion of constant 2014 USD; some projects recorded in the dataset do not have information on loan amount. All canceled and suspended transactions are excluded, as well as pledges of official finance. Data on development finance include all flows (loans, grants, debt forgiveness, export credit, etc.). Source: AidData China Global Data. Country # projects Value Zimbabwe 128 7.1 Angola 117 14.4 Tanzania 103 3.6 Ghana 99 6.8 Liberia 95 0.4 Kenya 93 10.5 Ethiopia 88 10.5 Zambia 87 36.3 Uganda 82 1.9 Sudan 80 8.9 Cameroon 77 4.9 Congo, Rep. 66 2.9 Namibia 65 0.6 Rwanda 61 0.9 Mali 60 10.9 Sierra Leone 60 0.7 Niger 58 1.5 Mozambique 57 2.9 Burundi 52 0.4 Nigeria 50 6.9 recipient economy that operate along the production chain of the development project. III.A Local Demand We start by investigating the local demand channel. Development finance in a given region can improve local well-being by increasing demand for workers employed in the financed project itself, thus increasing local wages. To the extent that higher local income translate into higher consumption, the demand for non-tradable goods should increase, under the assumption that non-tradable goods tend to be consumed locally. In order to test this hypothesis we estimate the following equation: log(y NT irct) t 1,t+1 = α r + α t + βcdf rct + ε irct (1) where i identifies a firm, r identifies a region within a country (the first administrative division, ADM1), c identifies a country, and t identifies time. The variable log(y NT irct) t 1,t+1 in equation (1) is the change in the log of a firm-level outcome between year t 1 and year t+1. We focus on three main firm-level outcomes: sales, employment and labor productivity measured as the ratio of sales over number of workers. In order to test the local demand channel, in this 8

Table II: Chinese Development Finance to Africa, by Sector Notes: The table reports the sectoral distribution of Chinese development finance, in terms of number of projects and value, between 2000 and 2014. Project value is available only for a sub-set of 1,122 project and it is expressed in billion of constant 2014 USD. All canceled and suspended transactions are excluded, as well as pledges of official finance. Data on development finance include all flows (loans, grants, debt forgiveness, export credit, etc.). Whenever possible, each project have been classified into a sector based on the title and description of the project. Sectors are based on the ISIC Rev 3.1 classification. Source: AidData China Global Data. # projects Value Agriculture 155 6.15 Fishing 13 0.64 Mining 24 5.26 Manufacturing 24 2.52 Electricity 158 24.54 Construction 471 62.99 Transport and communication 81 6.69 Financial intermediation 2 0.06 Real estate and business activities 99 0.47 Public administration 87 1.83 Education 291 1.52 Health and social works 493 2.06 Other services 145 2.75 Private household activities 4 0.00 Total 2,047 117.49 specification we restrict our sample to firms operating in the non-tradable sector (N T ). The coefficient of interest is β, which captures the effect of Chinese development finance (CDF) in region c at time t on the change in firm-level outcomes between the year before (t 1) and year after (t + 1) the year of implementation of the development project (t). The variable CDF is measured as number of projects. 8 Table IV reports the results obtained estimating equation (1) for firms operating in the services sector, which tend to be mostly non-tradable. 9 We start by studying the relationship between Chinese development finance and firm-level changes in sales. Columns 1 and 2 report the results. Our estimates indicate that firms operating in regions with a larger number of Chinese-financed development projects experienced a larger increase in sales. The magnitude of the estimated coefficient reported in column 2 suggests that firms operating in regions with one additional project experienced a 20 percent larger increase in sales between the year before and the year after the implementation of the project. Next, we focus on the effect of development finance on firm labor productivity and employment. Labor productivity is measured as the log of sales per worker. As shown, for labor productivity we find effects of similar size as those for sales, while we find small and non statistically significant effects of development finance on 8 We obtain similar and, for certain outcomes, larger magnitudes when using total value of loans as our measure of Chinese development finance. However, we prefer to use the number of projects as our measure of CDF in the baseline specification because for several projects the loan value is missing in the original data. 9 The World Bank Enterprise Survey directly identifies firms operating in manufacturing and services, the latter including wholesale, retail, IT, hotels and restaurants and other services 9

employment. Notice that we can not give a causal interpretation to the coefficients presented in Table IV. Higher firm-level growth in sales of non-tradable goods after the introduction of Chinesefinanced projects could be driven by forces other than the income effect generated by the Chinese-financed project itself. In addition, Chinese development institutions or the recipient country government might select a certain region for a given project based on its current performance. In that sense, our result would be explained by the allocation of development finance to regions that are experiencing larger growth ex-ante. We can partially deal with these concerns by presenting additional evidence consistent with our hypothesis. First, if local firm growth is generated by larger consumption, then we should observe larger growth of firms operating in non-tradable sector relative to the tradable sector. We test this by estimating equation 1 on the sample of firms from the World Bank Enterprise Survey that operate in the manufacturing sector. This is because manufacturing goods tend to be tradable, across regions and across countries. Results are reported in Table V. As shown, the coefficients on the number of Chinese-financed development projects when the outcome is sales growth of manufacturing firms are half the size with respect to those presented in Table IV. Second, we investigate the timing of the effect of Chinese development finance on firm-level outcomes. To this end, we augment equation (1) by adding two-years lag and two-years lead of the number of Chinese financed-projects. The results are reported in Table VI. As shown, firm sales growth between year t 1 and year t + 1 is not a predictor of the number of Chinese-financed projects that will be implemented in a given region (at time t + 2). This result mitigates the concern that our estimates are driven by reverse causality: Chinese-finance projects being allocated to regions with larger sales growth ex-ante. 10 At the same time, our results indicate that, if any, the effect of Chinese-financed projects are relatively short lived, as there is no effect of development finance at t 2 on firm sales growth between t 1 and t + 1. III.B Production Chain In the previous section we showed that Chinese-financed projects had a positive impact on sales of local firms, and that these effects are larger for firms operating in the non-tradable sector. In this section we study the effect of Chinese development finance on firms in recipient economies through input-output linkages. Depending on the sector in which a development project is implemented, it can potentially generate an increase in demand for goods produced by firms that operate along the production chain of that project. In order to test this hypothesis we construct a firm-level measure of exposure to Chinese development finance through inputoutput linkages and estimate the following equation: 10 Aggregate evidence at the country level also indicates that Chinese investment are uncorrelated with growth expectations. Figure VII plots official development flows (as a share of GDP) in year t against the average IMF growth forecast in year t for the period t + 1 to t + 3, and it shows no correlation between the two variables. 10

Figure VII: Chinese official development assistance and growth expectations 20 ZWE ODF from China (as a % of GDP) 15 10 5 0 ERI ERI LBR TGO MUS COM ZWE MRT GNB GIN NER SLE CIV KEN ETH CMR TCD SYCOG BDI CAF COG ZWE KEN MOZ LBR ETH GNQ MRT RWA RWA SDN TZA ZWE COG GNB BDI AGO BEN CPV MLILBR GHA MWI ETH ETH LBR COG CMR AGO CMR COG GHA CIV ETH GHA ZWE COM ZMB MOZ MOZ SDN TCD CAF CMR CPV GNQ LSO COM MLI UGA MOZ GNQ LSOAGO BDI BDI CAF CMR GHAMRT RWA UGA MUS ZMB SYC TGO ERI COD ETH COG COM CMR GHA GIN ZMB GIN SLE ZMB GNB UGA MLI NER CAF AGO AGO BDI BDI ETHGAB GNBMDG GHA LBR GAB LSO COG BDI AGO AGO AGO BWA BWA CAF BDI BDI BDI CIVBWABWA CIV COGCOM GHA GAB LBRGNQ KEN BDI CAF BWABEN CIV CIV CAF BEN BDI CIV CAF BDI BEN CIV CIV CMR COD CMRCMR BEN BWA CIV CMR ETHCPV ERI ERI ERI ETH BDI BWA BEN AGO AGO COD COD COM CPV COM CPV COG COG ERI ERIETH CPV COGCPV ETH COG ETH ETH GAB GAB GHA GHA LBRGHA KEN GNQ GIN GHA GNQ GAB GAB GNB GAB GNB GAB GNB GIN GNB GIN GNB GIN GIN GIN GIN GAB CPV GNB CODCOD GHA ERI CPV ETH GHA GNQ GNQ GNQ KEN KEN KEN KEN KEN KEN KEN LBR LBR LBR LBR KEN LSO LSO LSO MLI LSO MLI MLI MWI MOZ NAM MDG MDG MDGMOZ MOZ MUS MWI NAM NAM RWATZA NER ZWE SDN SEN MLI MOZ MUS MWI NGA MOZ MOZ MUS LSO MDG LSO MDGMDGMDG LSO KEN GNQ KEN MLI LSO LSO MLI MOZ MRT MRT MRT MDG MDG MOZ MRT MUS MWI MUSMUS MWI NAM NAM NAM NER NAM NER NAM SDN SLE UGA SDN NER RWA NER NGA NER NERNER NAM NGA NGA RWA NGA MUS MUSNGA MWI NGA RWA NER NGA SDN SEN RWA RWA RWARWA RWARWA SDN SDN SDN SEN SEN SEN SLE SLE TGO SLE SLE SYC SYC TGO TGO SLE TCD SLE NGA MOZ MRT SDN TCD SYC TGO SYC SYC TGO TCD TGO TZA TGOTGO SENSEN SLE SDN SDN TGO TZA SLESLE TZA TZA ZMB TZA UGA ZAF ZAF ZAFZAF ZWE ZAF ZMBZMB ZMB ZMB ZWE ZMB ZWE ZWE ZWE ZMB UGAUGA UGA TZA SLE UGA TZA TCDTZA COD TZACOG AGO GNQ BWA MOZ SDN GIN MRT 0 5 10 15 Average growth forecast over t+1 and t+3, as in April WEO of year t Source: AidData China Global Data and IMF World Economic Outlook (different April vintages). log(y ijct ) t 1,t+1 = α j + α t + β j w jj CDF j t + ε r (2) where i indexes firms, j indexes sector, c indexes country and t indexes time. Our measure of exposure to development finance through input-output linkages is defined as sector-level. The weight w jj captures the exposure of sector j to sector j through input-output linkages. We define this weight as the share of inputs that sector j buys from sector j divided by the total value of inputs used by sector j. In order to construct these weights, we use the inputoutput table of the United States for the year 2000, which pre-dates the period of analysis of this paper. The measure of exposure is just the weighted sum of the number of development projects implemented in all sectors j that operate downstream with respect to sector j, i.e. all those sectors buying inputs from sector j. Table VII reports the results obtained estimating equation (2). We start by studying the relationship between exposure to Chinese development finance via input-output linkages and firm-level changes in sales. Columns 1 and 2 report the results. Our estimates indicate that firms operating in sectors that are more exposed to development projects experienced a larger increase in sales. The magnitude of the estimated coefficient reported in column 2 indicates that firms operating in sectors with a one standard deviation higher exposure to Chinese-financed 11

development projects through input-output linkages experienced a 21.6% larger increase in sales between the year before and the year after the implementation of the project. These firms experienced an increase in labor productivity of similar magnitude, while we find no effect of the production chain channel on firm employment. IV Final Remarks Chinese lending to Africa has markedly increased over the last decade. While Chinese financing can contribute to infrastructure development and promote growth, it is also fraught with controversy. So far, there is no clear evidence on the effects of Chinese lending on recipient economies. To shed light on this debate, we provide new empirical evidence on the effects of Chinese financing on local economies, based on project-level data on Chinese development finance, matched with firm-level information. First, we find that firms operating in regions that received a Chinese-financed projects experienced a higher sales growth than those located in regions without Chinese projects. As this effect is stronger for service than for manufacturing firms, we interpret it as evidence suggesting that Chinese financing improve local consumption. Second, we show that firms active in sectors more exposed to Chinese development financing growth more in response to an increase in Chinese financing. Both our results indicate that Chinese-financed development projects could have positive effects on local economies, thanks to an increase in firm growth and labor productivity, even though there are no discernible effects on jobs. However, our findings can not be interpreted is a causal way, given that the allocation of projects may not be random, but related to economic prospects. Suggestive evidence on the distribution of projects across countries does not indicate that this is in fact the case. To have a more convincing identification strategy, further analyses will tackle the endogeneity issue in a more formal way. In addition, we will also expand the scope of our work, considering the effect of Chinese financing on trade flows at the firm level. 12

References AidData (2017). Tracking China Development Finance Worldwide Dataset. Busse, M., C. Erdogan, and H. Mühlen (2016). China s Impact on Africa The Role of Trade, FDI and Aid. Kyklos 69 (2), 228 262. Chen, W., D. Dollar, and K. Tang (2016). Why Is China Investing in Africa? Evidence from the Firm Level. World Bank Economic Review Forthcoming. Clemens, M. A., S. Radelet, R. R. Bhavnani, and S. Bazzi (2012). Counting chickens when they hatch: Timing and the effects of aid on growth. The Economic Journal 122 (561), 590 617. Dreher, A., A. Fuchs, R. Hodler, B. C. Parks, P. A. Raschky, and M. J. Tierney (2016). Aid on Demand: African Leaders and the Geography of China s Foreign Assistance. Working Paper 3 (revised), AidData. Dreher, A. and S. Lohmann (2015). Aid and growth at the regional level. Oxford Review of Economic Policy 31 (3-4), 420 446. Galiani, S., S. Knack, L. C. Xu, and B. Zou (2017). The effect of aid on growth: Evidence from a quasi-experiment. Journal of Economic Growth 22 (1), 1 33. Kilama, E. G. (2016). The influence of China and emerging donors aid allocation: A recipient perspective. China Economic Review 38, 76 91. McMillan, M. (2017, July). Chinese investment in Africa. VoxDev, July 21. Parks, B. C. (2015, November). 10 Essential Facts About Chinese Aid in Africa. The National Interest. Qian, N. (2015). Making Progress on Foreign Aid. Annual Review of Economics 7, 277 308. Rajan, R. G. and A. Subramanian (2011). Aid, Dutch disease, and manufacturing growth. Journal of Development Economics 94 (1), 106 118. Strange, A. M., A. Dreher, A. Fuchs, B. C. Parks, and M. J. Tierney (2017). Tracking Underreported Financial Flows: China s Development Finance and the Aid Conflict Nexus Revisited. Journal of Conflict Resolution 61 (5), 935 963. Sun, I. Y., K. Jayaram, and O. Kassiri (2017, June). Dance of the lions and dragons. McKinsey & Company. 13

Tables Table III: Summary Statistics variable name mean median sd N Panel A: Services firms log Sales t 1,t+1 0.424 0.159 2.102 6,585 log Sales L t 1,t+1 0.284 0.024 2.102 6,585 log L t 1,t+1 0.141 0.000 0.413 6,585 CF D t 0.748 0.000 1.111 6,585 Panel B: Manufacturing firms log Sales t 1,t+1 0.525 0.182 2.321 5,170 log Sales L t 1,t+1 0.420 0.075 2.311 5,170 log L t 1,t+1 0.103 0.000 0.434 5,170 CF D t 0.862 0.000 1.214 5,170 Panel C: Production Chain log Sales t 1,t+1 0.477 0.169 2.230 11,344 log Sales L t 1,t+1 0.123 0.000 0.427 11,344 log L t 1,t+1 0.353 0.049 2.225 11,344 CF D t 0.776 0.331 0.894 11,344 Notes: Sources are World Bank Enterprise Survey and AidData - Tracking Chinese Development Worldwide dataset. 14

Table IV: Local Demand Channel: Services Firms (1) (2) (3) (4) (5) (6) VARIABLES log Sales t 1t+1 log Sales log L L t 1t+1 t 1t+1 CDF in the region (n loans) t 0.116** 0.108** 0.113** 0.106** 0.002 0.001 [0.046] [0.044] [0.043] [0.041] [0.007] [0.007] Observations 6,585 6,582 6,585 6,582 6,585 6,582 R-squared 0.039 0.046 0.034 0.041 0.020 0.033 Sector FE Yes - Yes - Yes - Year FE Yes - Yes - Yes - Sector year FE No Yes No Yes No Yes N clusters 44 44 44 44 44 44 Notes: Standard Errors are clustered at the country-year level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1. Table V: Local Demand Channel: Manufacturing Firms (1) (2) (3) (4) (5) (6) VARIABLES logsales t 1t+1 log Sales logl L t 1t+1 t 1t+1 CDF in the region (n loans) t 0.082* 0.082* 0.076* 0.076* 0.004 0.004 [0.042] [0.042] [0.039] [0.039] [0.007] [0.007] Observations 5,170 5,170 5,170 5,170 5,170 5,170 R-squared 0.049 0.049 0.044 0.044 0.014 0.014 Sector FE Yes - Yes - Yes - Year FE Yes - Yes - Yes - Sector year FE No Yes No Yes No Yes N clusters 44 44 44 44 44 44 Notes: Standard Errors are clustered at the country-year level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1. 15

Table VI: Demand Channel: Timing of the Effect (1) (2) (3) (4) (5) (6) (7) (8) (9) log Sales t 1t+1 log Sales L t 1t+1 log L t 1t+1 VARIABLES all firms Serv Manuf all firms Serv Manuf all firms Serv Manuf CDF in the region (n loans) t+2-0.189* -0.236** -0.135* -0.177* -0.220* -0.128* -0.012-0.016-0.007 [0.099] [0.108] [0.073] [0.100] [0.109] [0.071] [0.010] [0.012] [0.010] CDF in the region (n loans) t 0.195*** 0.238*** 0.169** 0.194*** 0.231*** 0.171*** 0.001 0.006-0.004 [0.068] [0.076] [0.065] [0.066] [0.074] [0.063] [0.007] [0.008] [0.007] CDF in the region (n loans) t 2-0.011-0.016-0.031-0.020-0.017-0.046 0.009 0.002 0.016 [0.088] [0.096] [0.088] [0.082] [0.090] [0.082] [0.010] [0.010] [0.011] Observations 9,983 5,550 4,430 9,983 5,550 4,430 9,983 5,550 4,430 R-squared 0.044 0.049 0.048 0.041 0.043 0.045 0.015 0.019 0.013 Sector FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector year FE No No No No No No No No No N clusters 35 35 35 35 35 35 35 35 35 Notes: Standard Errors are clustered at the country-year level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1. 16

Table VII: Production Chain Channel: Baseline (1) (2) (3) (4) (5) (6) VARIABLES log Sales t 1t+1 log Sales log L L t 1t+1 t 1t+1 CDF Exposure - IO Linkages (n loans) t 0.200* 0.242** 0.209** 0.252** -0.012-0.015 [0.105] [0.116] [0.097] [0.106] [0.016] [0.019] Observations 11,344 11,340 11,344 11,340 11,344 11,340 R-squared 0.040 0.046 0.036 0.042 0.017 0.026 Sector FE Yes - Yes - Yes - Year FE Yes - Yes - Yes - Sector year FE No Yes No Yes No Yes N clusters 42 42 42 42 42 42 Notes: Standard Errors are clustered at the country-year level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1. Table VIII: Production Chain Channel: Robustness (1) (2) (3) (4) (5) (6) VARIABLES log Sales t 1t+1 log Sales log L L t 1t+1 t 1t+1 CDF Exposure - IO Linkages (n loans) t 0.235* 0.312** 0.251** 0.314** -0.020-0.008 [0.123] [0.139] [0.111] [0.130] [0.021] [0.021] Exporter 0.292 0.262 0.333 0.285-0.038** -0.023 [0.215] [0.195] [0.227] [0.211] [0.015] [0.020] Med-sized firm, 0.026 0.082-0.064-0.037 0.081** 0.110*** [0.055] [0.056] [0.063] [0.063] [0.035] [0.036] Large firm -0.089-0.023-0.200-0.212 0.102** 0.178*** [0.108] [0.095] [0.144] [0.140] [0.045] [0.052] Government-owned 0.022* 0.021* 0.001 [0.012] [0.011] [0.002] Firm age (ln) -0.129-0.026-0.100*** [0.101] [0.101] [0.010] Sole proprietorship 0.170** 0.107 0.062*** [0.082] [0.081] [0.016] Observations 9,925 8,689 9,925 8,689 9,925 8,689 R-squared 0.050 0.054 0.046 0.049 0.031 0.056 Sector FE - - - - - - Year FE - - - - - - Sector year FE Yes Yes Yes Yes Yes Yes N clusters 38 30 38 30 38 30 Notes: Standard Errors are clustered at the country-year level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1. 17