«The Effectiveness of Industrial Policy in Developing Countries: Causal Evidence from Ethiopian Manufacturing Firms Tewodros Makonnen Gebrewolde, ...»
Department of Economics
The Effectiveness of Industrial Policy in
Developing Countries: Causal Evidence
from Ethiopian Manufacturing Firms
Tewodros Makonnen Gebrewolde, University of Leicester
James Rockey, University of Leicester
Working Paper No. 16/07
The Eﬀectiveness of Industrial Policy in
Developing Countries: Causal Evidence from
Ethiopian Manufacturing Firms
Tewodros Makonnen Gebrewolde
University of Leicester
University of Leicester
Prioritizing the growth of particular sectors or regions is often part of LDC growth strategies. We study a prototypical example of such policies in Ethiopia, exploiting geographic and sectoral variation in the form and scale of the policy for identiﬁcation. Using product-level data on Ethiopian manufacturing ﬁrms we show that the policy was unsuccessful: There was no improvement in productivity, productive assets, or employment. The policy failed due to its negative eﬀects on productivity of the entry of new ﬁrms and existing ﬁrms diversifying.
Moreover, subsidised loans and tax-breaks led to an increase in capital but not in machinery.
Keywords: Industrial Policy, Ethiopia, Manufacturing.
JEL-Codes: H25, H81, O14, O25 Corresponding author email: email@example.com. We thank Martin Foureaux Koppensteiner, Sourafel Girma, and Jochen Mierau for their comments. We also thank seminar participants at the CSAE Conference 2015; the 13th International Conference on the Ethiopian Economy; Royal Economic Society Sussex; Leicester; Nottingham;
Queen’s Belfast; and the Centre Emile Bernheim, Solvay Brussels School. Any errors are ours alone.
It is thus that every system which endeavours, either, by extraordinary encouragements to draw towards a particular species of industry a greater share of the capital of the society than what would naturally go to it, or, by extraordinary restraints, to force from a particular species of industry some share of the capital which would otherwise be employed in it, is, in reality, subversive of the great purpose which it means to promote. It retards, instead of accelerating the progress of the society towards real wealth and greatness; and diminishes, instead of increasing, the real value of the annual produce of its land and labour. Smith (1776) Despite Adam Smith’s scepticism, the provision of subsidies and tax-breaks to particular industries has long been a favoured tool of governments hoping to steer and accelerate the development of their economy. This is particularly true of the governments of the Least Developed Countries (LDCs) (see, Rodrik, 2009). Such subsidies are a particular form of Industrial Policy (IP) – one that is extremely prevalent but poorly understood. This paper recasts Smith’s claim as a question: is this form of IP eﬀective? It provides the ﬁrst causal analysis of IP in an LDC and in particular it studies whether the governments of LDCs may improve economic performance through the use of targeted subsidies and tax-breaks. Such policies reﬂect an important strain of thought in Development Economics; the ‘big-push’ arguments originally due to Rosenstein-Rodan (1943) and rejuvenated by Murphy et al. (1989). This tradition argues that such encouragement of activity in speciﬁc sectors and places can lead to suﬃcient scale and agglomeration economies to deliver sustained growth. We study a prototypical IP introduced by the Ethiopian Government as part of its growth strategy in 2002. In common with many such policies, the Ethiopian policy provided for a series of targeted tax-breaks, and other support including concessionary loans. Sectoral and geographic variation in the scale and form of the policy provide for a natural experiment that we exploit to identify causal estimates. We ﬁnd that it did not lead to improvements in the productivity or scale of existing businesses, and that whilst it led to the entry of new, lower productivity businesses, this did not lead to any any accompanying agglomeration externalities.
That it failed is particularly disappointing as there are reasons to believe the potential beneﬁts of IP will be larger in LDCs than elsewhere.
It is also disappointing because there are also reasons to think that the costs of an unsuccessful policy are also high. Indeed, many of the theoretical arguments for industrial policy, such as those of Bulow and Summers (1986), Rebelo (1991) and Rodrik (2009), depend on imperfections in product, labour, and capital markets that can be expected to be larger in LDCs. Equally, the concerns that such imperfections are hard to identify and address are perhaps more plausible in LDCs.
Yet, there is a lack of econometric evidence one way or the other. This paper addresses this deﬁcit, and provides the ﬁrst micro-econometric analysis of an industrial policy in an LDC. One of the complications of understanding the eﬀects of an industrial policy is that any given policy can be expected to have several countervailing consequences.
We formalise these diﬀerent consequences with a simple analytical framework in which there are both direct eﬀects on ﬁrms’ productivity, capital, and employment levels, but also spillovers between ﬁrms. We then take this to a rich product-level dataset on the universe of Ethiopian manufacturing ﬁrms to tease apart these diﬀerent consequences. These data allow us to understand not only the impact on ﬁrms’ growth, but also on the entry of ﬁrms, and the entry of existing ﬁrms into new product lines. Crucially, they also allow us to exploit the intersection of variations in the sectoral and geographical eligibility criteria, to obtain causal estimates from a diﬀerence-in-diﬀerence-in-diﬀerence estimator.
In line with the theory, we ﬁnd evidence that the direct impact of the policy on productivity is negative, and that whilst there are positive spillover eﬀects, these are to small to compensate for the direct costs.
We further show that whilst the policy does encourage additional investment out of retained proﬁts, the productivity of capital falls.
Drilling down further into the data reveals that this investment is concentrated on only indirectly productive assets, such as buildings and vehicles, rather than directly productive machinery. We interpret this a form of savings in the face of high prevailing inﬂation rates rather than investment in production per se. The policy’s failure to generate additional employment can also be explained by this lack of investment in machinery. A calculation of the costs of the policy suggests that the policy was expensive, with forgone tax revenues alone equivalent to 0.5% of GDP or 5% of annual Government Spending – for comparison, the entire manufacturing sector accounts only for 5% of GDP.
Whilst we focus on Ethiopia, the policy we study is similar in form to those implemented in several other countries in Sub-Saharan Africa (SSA). Table 1 summarises the structure of the Industrial Policy of a selection of eight SSA countries as described by Marti and Ssenkubuge (2009). The table separates IP into three categories: Trade related policies, the literature on which is discussed below; Sector Speciﬁc Support, which is the focus of this paper; and inducements for Foreign Direct Investments, itself the subject of a large literature. Each of these policies is then categorised on the basis of whether it has a substantial Tax or Duty/Tariﬀ component, and whether other forms of government support were provided. The table reveals the broad consistency in the form of IP in these eight countries. Speciﬁcs of the policies are provided in Table A6 in the Appendix. Focussing on sectoral support, all of the countries other than Cameroon provided additional support to speciﬁc sectors. Cameroon provided support solely by exempting certain sectors from tax, while Ghana and Kenya employ both strategies. Ethiopia is unusual in that it provides diﬀerent reductions in tax depending on location. The precise form of the ‘other’ support varies; but it normally involves, as in Ethiopia, a combination of concessionary loans, alongside infrastructure and training support. That the Ethiopian policy also oﬀers tax-breaks based on location means that we are also able to test the eﬃcacy of tax incentives. Thus, the Ethiopian policy we study is both representative of policies implemented elsewhere. The combination of geographic tax-breaks and sectoral support allows us to test the full range of policies. For both of these reasons, our ﬁnding that it was ineﬀective may therefore have implications for the design of policy in other LDCs.
This paper builds on previous work that estimates the causal eﬀects of similar combinations of tax-breaks and subsidies designed to stimulate output in economically depressed regions of rich nations. Busso et Table 1: Summary of Industrial Policy in Eight Subsaharan Africa Countries.
al. (2013) study the US Federal Empowerment Zone (EZ) program and ﬁnd that it increased employment without costs in eﬃciency or eﬀects on prices. Similarly, Criscuolo et al. (2016) ﬁnd in their study of the impact of the EU Regional Selective Assistance scheme on UK ﬁrms that both employment and investment increased. Moreover, they too ﬁnd that this happens at little cost to productivity. On the other hand Gobillon et al. (2012), studying the French EZ programme, ﬁnd that the eﬀects are small and transitory. Moreover, Neumark and Kolko (2010), who study a similar Californian policy, ﬁnd it too to be ineﬀective.
As well as a diﬀerence in context between countries such as the the US (Busso et al., 2013), France (Gobillon et al., 2012), or the UK (Criscuolo et al., 2016) and that of countries such as Ethiopia, there is also a necessary diﬀerence in emphasis. This may be seen through the lens of the analysis of Albouy (2009). He exploits the fact that the incidence of US federal taxation is unevenly distributed to provide evidence that local tax rates are a long-run determinant of output and employment levels. This suggests a mechanism through which EZ zones can boost local output. In the US, such policies represent a small fraction of GDP and thus need not pay for themselves, but rather can be seen as an eﬃcient form of redistribution. On the other hand, the Ethiopian policy is, in relative terms, expensive. This means, that even if a policy of tax-breaks for particular sectors of Ethiopian Manufacturing Industry is expected, given the previous evidence, to be successful, it should come at a cost elsewhere in terms of the necessary additional taxation. Thus, what it means for IP to be eﬀective is diﬀerent in one of the LDCs, such as Ethiopia, and a rich nation, such as France or the U.S., and our focus on TFP reﬂects that.1 Put diﬀerently, in the LDC context, eﬀective IP is IP that can pay for itself.
Our focus on productivity is in common with the literature on the another form of IP, protectionism and infant industries, which also has often focussed on less-developed countries. Harrison (1994) revisiting Krueger and Tuncer (1982) argued that sectors of Turkish manufacturing that had enjoyed more protection had (in fact) also exhibited faster productivity growth. By now, there is increasing support for the opposite conclusion. Topalova and Khandelwal (2011) provide causal evidence that reduced tariﬀs in India led to improved ﬁrm-level productivity. Goldberg et al. (2010) show that the same reduction in tariﬀs led to an increase in the number of products available. Blonigen (2015) looks at the impact of protecting an important sector, in his case steel, on other sectors. He ﬁnds that there are large costs for sectors using steel. Of particular relevance for this paper is that he ﬁnds the costs of the policy are highest in LDCs. Harrison and RodríguezClare, Andrés (2010) review both the theoretical and empirical literature and suggest that there is little evidence that Industrial Policy based on tariﬀs, quotas and subsidies is eﬀective. They do suggest, however, that more subtle policy may be successful. Nunn and Treﬂer (2010) provide quantitative evidence of such subtlety. They show that countries in which the tariﬀ structure favours skill-intensive industries grow faster, but that three-quarters of this eﬀect is due to the endogeneity between tariﬀ-structures and domestic rent-seeking. Along similar lines, Aghion et al. (2015) argue theoretically and empirically that protectionist policies can be productivity-enhancing when they are targeted at sectors that are already competitive, or when they are designed to encourage competition. They present evidence from Chinese 1 IP could also be eﬀective if it led to growth through the reallocation of capital to more productive activities. In Section 6 we test for this and ﬁnd no evidence of such eﬀects.
manufacturing ﬁrms that the interaction of sector-level competition and subsidy is positive and signiﬁcant. This paper contributes to this literature by providing micro-econometric evidence based on extremely rich data about the eﬀects of tax-breaks in an LDC. We are able to isolate the causal mechanisms through which the policy aﬀects ﬁrms. This provides insights into why the policy we study was less successful than that examined by Aghion et al. (2015). One reading of our ﬁndings in the context of this literature is that we provide new evidence for why ‘blunt’ policies are ineﬀective.
The paper is organized as follows. Section 2 introduces a simple analytical framework with which to organise our ideas and to derive the hypotheses that the remainder of the paper will test. Section 3 discusses the particular policy we study and outlines key features of the Ethiopian context. Section 4 introduces the data we employ, Section 5 speciﬁes the empirical strategy we use to identify the causal eﬀects of the policy, and Section 6 presents the results. Section 7 presents calculations of the cost of the policy. Section 8 closes the paper.