«The Effectiveness of Industrial Policy in Developing Countries: Causal Evidence from Ethiopian Manufacturing Firms Tewodros Makonnen Gebrewolde, ...»
2 Analytical Framework To ﬁx ideas and clarify our hypotheses, it useful to construct a simple analytical framework. We extend that of Criscuolo et al. (2012) who focus on an expression for the cost of capital owing to Hall and Jorgenson (1967), King (1975), and Ruane (1982). They use this to study the implications of EU Regional Assistance for employment and capital utilization. We extend this approach to obtain predictions for the eﬀects of the policy on TFP. In particular, our framework formalises the intuition that IP will lower average productivity as previously nonviable ﬁrms will enter the market. It also embodies the notion that, particularly in LDCs, that there may be increasing returns to scale as increased output and competition can improve average productivity through spill-overs and other agglomeration externalities.
We consider a highly stylised economy comprised of a continuum of ﬁrms, each able to produce Ai ∈ (0, A+ ) units of output given K units of capital and L of labour. These units are normalised such that i K di = L di = 1. Firms are otherwise identical and each ﬁrm must pay capital i and wage costs of ρK + ψL, where ρ 0 and ψ 0 and Value Added Tax rate of 1 τ 0.
The cost of capital is given by the Hall and Jorgenson (1967) formulation:
where 1 θ 0 is the depreciation allowance, r 0 is the interest rate, and 1 δ 0 the depreciation rate. In equilibrium, not all ﬁrms choose to produce, and in particular, ﬁrm i produces iﬀ it makes weakly positive
We denote the Ai that satisﬁes this condition exactly as A∗. Every ﬁrm has a CES production technology Yi = f (Ai, K, L) with elasticity of substitution σ. Ai is the ﬁrm speciﬁc TFP term given by the product
where Bi 0 is the level of TFP of ﬁrm i that would obtain in the absence of agglomeration externalities and φ 0 implies that there are positive
agglomeration externalities. It follows that total output is given by:
For simplicity, we treat the funding for any tax-reduction as being obtained from elsewhere, in the context of Ethiopia perhaps from development assistance. It follows immediately that a tax-relief policy
has the following consequences:
If φ 3 then the spillover eﬀect is suﬃciently large that the additional agglomeration eﬀect due to the new ﬁrms entering more than oﬀsets the eﬀects of their lower average productivity, and ∂Y
0. This is a very simple statement of the notion common to ∂τ many of traditional ‘big-push’ arguments for IP: If φ is suﬃciently large then average productivity will increase, and the policy will have had an unambiguously positive impact. Whilst, caution is necessary in drawing quantitative conclusions from such a simple model that here agglomeration externalities need to be cubic suggests that alone they may often be insuﬃcient.2
3. Unemployment falls. As in Criscuolo et al. (2012), we can use the chain-rule to write the elasticity of employment as the product of the elasticity of capital with respect to the tax and the elasticity of
employment with respect to the cost of capital:
in employment. On the other-hand, if for some reason an increase in capital were mis-invested in a low-productivity asset then we should not expect much increase in employment. We shall see that this is the case below.
The overall consequences of the policy will thus depend on the distribution of the (latent) productivities of ﬁrms in the economy, the relative importance of agglomeration externalities, and the skill with which additional capital is invested. We shall see that in the case we study, that agglomeration externalities are insuﬃcient to oﬀset the lower productivity of entering ﬁrms, and that capital tends to be directed towards assets that are more fungible rather than productive.
Here we ignore how the tax-breaks are ﬁnanced. This is reasonable if they are paid for by cuts to non-productive expenditure elsewhere, from additional foreign-aid, or deﬁcit spending. The generalisation to a general-equilibrium model, where the policy must be ﬁnanced from other taxation, or cuts in productive government expenditure in the tradition of Barro (1990) produces the same qualitative predictions at the cost of some additional complication.
3 Industrial Policy in Ethiopia Between 1974 and 1991, Ethiopia endured decades of drought, war, and political instability under the communist regime known as the Derg. During this era there was little industrial production and private enterprise was discouraged. This changed in 1994 with the promulgation of a new constitution. Since then, Ethiopia has been following an industrial development strategy named Agricultural Development Led Industrialization (ADLI). It focusses on improving agricultural productivity to both release labour for the industrial sector and increase agricultural incomes to serve as a strong market for the industrial sector’s products. The overall strategy has so-far comprised three ﬁve-year plans since 2000. The ﬁrst plan was called the Sustainable Development and Poverty Reduction Plan (SDPRP) and began in 2000.3 3 The second ﬁve-year plan is called the Plan for Accelerated and Sustainable Development to End Poverty (PASDEP) and ran for the next ﬁve years. The ﬁnal phase, The subject of this paper is speciﬁc aspects of the SDPRP to enhance private sector development. Speciﬁcally, in 2002, the government announced a revised schedule of incentives and tax-breaks. The strategy was explicitly designed to encourage manufacturing sectors that were labour intensive and that utilised Ethiopian agricultural products (see, Ministry of Finance and Economic Development, 2002). Firms were
eligible for tax breaks as follows:
• If a ﬁrm exports 50 percent or more or supplies 75 percent to an exporter it received 4 years income tax exemption.
• Exports 50 percent it recieved 2 years income tax exemption.
• Companies not around Addis Ababa gained 1 additional year of tax exemption.
• All enterprises were eligible to customs duty exemption on capital goods.
These investment incentives do not diﬀerentiate between speciﬁc industries. They do, however, diﬀerentiate ﬁrms based on location and export volume. The number of ﬁrms with such export volume is small and these ﬁrms are almost exclusively long-standing and government owned. Instead, we focus our attention on the eligibility of ﬁrms more than 100km outside the centre of Addis-Ababa for an additional tax-break. Ethiopia is divided in 9 administrative regions.
This division is based on ethnicity. These regions are further divided into 68 administrative zones which are in turn divided into 560 woredas (districts). Figure 1 plots these diﬀerent regions and a central circle depicts the 100km zone that deﬁnes our treatment. This shows that even though Addis-Ababa is the key locus of economic activity, this area is small given the size of the country. When we come to test for agglomeration externalities, we will treat woredas as our unit of analysis.
In addition to these general investment incentives, and in line with the development plans discussed above, speciﬁc sectors were targeted the Growth and Transformation Plan (GTP) ﬁnished in 2015.
for direct support. The selection of these sectors is mainly based on their linkages to the agricultural sector, labour intensity and export potential.
These sectors are: textiles and garments; meat and leather products;
agro-processing; and construction. The details of this support are described in Appendix C. The structure of the policy and support oﬀered is outlined in Figure 2. Treated sectors had access to concessionary loans, as well as initiatives to improve the supply of trained workers and other technology support. Our treatment is the intersection of the two arms of the policy – being outside Addis-Ababa and in a supported sector.
By focusing on those ﬁrms that have received the most support we are giving the policy the best chance of being successful. That some of the sector-speciﬁc support is speciﬁcally designed to boost productivity, and thus likely to oﬀset the predictions of declining productivity outlined in Section 2, improves these chances further.
3.1 Identiﬁcation So that estimates of the policy’s eﬀectiveness are not biased upwards, we need to be sure that the policy was not targeted at ﬁrms most likely to beneﬁt from it. Similarly, to avoid the concern of Rodrik (2009) that estimates may be biased downwards because aid goes to ﬁrms that most need it, we must be sure that the policy was also not targeted on this basis. Inspection of the policy proclamation (Ministry of Finance and Economic Development (2002)) shows that the overall objective of these measures is clear: it is to increase the linkages between agriculture and industry; to increase employment, and to increase exports. Thus, the sectors targeted were chosen solely on the basis of whether they make use of Ethiopian agricultural produce, or are labour intensive. It is clear that all of the targeted sectors; Meat and Leather, Textiles, AgroBusiness, and Construction ﬁt this description. Importantly, none of them involves a product where Ethiopia may be expected to have a particular competitive advantage (or disadvantage). Thus, whilst the government must be keen to boost productivity there is no evidence that the choice of targeted sectors was made on the basis of maximizing TFP growth.4 Indeed, such a strategy of ‘picking winners’ is always fraught with diﬃculty, and particularly so given the context of Ethiopia at the turn of the century. Moreover, the reverse strategy of supporting losers is not consistent with the Ethiopian political context, or aﬀordable given its budget constraints.
4 There is also no evidence that these sectors were chosen for political economy reasons, and we have no evidence that there was systematic corruption in the delivery of the policy.
Inspection of a map of the region around Addis-Ababa shows that the 100km threshold is outside of the city and of no-obvious geographical importance – it clearly reﬂects the usual preference for round numbers than any particular economic or geographic reality.5 There are also relatively few ﬁrms near the threshold that might be expected to relocate. Secondly, property rights are technically all held by the Government in Ethiopia and thus the opportunity of ﬁrms to relocate is extremely limited. Thus, there is no reason to suspect that the choice of threshold geographic threshold was endogenous. Finally, one might be concerned that the ﬁrms subject to the geographic treatment are systematically diﬀerent. There is little reason to believe this to be the case as most ﬁrms are engaged in low value added production using homogenous agricultural produce as inputs. Moreover, we include ﬁrm ﬁxed eﬀects and in the Appendix show that our results are robust to controlling for region-speciﬁc time trends. Thus, we can be clear that both arms of the policy and their intersection are exogenous.
4 Data The data used in this study were obtained from the Ethiopian Large and Medium Scale Manufacturing Enterprises Census that is conducted annually by the Central Statistics Agency of Ethiopia. It contains the universe, and is hence an unbalanced panel, of ﬁrms for 14 years from 1996-2010. Initially, there are close to 600 ﬁrms in 1996. By 2010, there are around 1900. The ﬁrms are categorised into 54 industrial classiﬁcation (ISIC) codes. Table A7 in the Appendix reports the average number of ﬁrms in each category over the period.
As well as being available for all ﬁrms, the data are extremely rich, containing detailed information on both the establishment and ownership details of each ﬁrm. We make use of much of this information, and
summarise the information we use below:
• Ownership: Gender of the proprietors, and the proportion of a ﬁrm’s capital in public, private, or foreign ownership.
5 Indeed, our results are robust to the use of an 80km or 100km threshold.
• Establishment: Detailed information on the month and year of establishment as well as a ﬁrm’s initial capital are available.
• Employment: Classiﬁed by gender, salary group and occupation on a quarterly basis. Information on wages and other beneﬁts for workers is also included.
• Products: Data are on up to 12 products. This includes the unit price, beginning stock, production quantity and production value, and we use these data to construct our output index and productivity measures. Data on sales and exports are also available at the product level.
• Investments: A ﬁrm’s assets are aggregated into diﬀerent categories such as ﬁxed assets, furniture, machinery and vehicles.
The levels of each are detailed with the beginning stock, annual changes and ending stock.
• Intermediate inputs: These are at the level of the ﬁrm rather than the product. They include unit price, quantity, value, source (local versus imported) of the input.
• Expenses: Production expenses, such as utilities, energy, and tax, are available at the ﬁrm level.
Importantly, as discussed below, these data contain detailed quantity information about both quantities of products produces and the quantities of the inputs used to do so. This, unusual level of detail allows us to understand precisely how the policy aﬀected treated ﬁrms. Table 2 provides the usual summary of our key variables. In the rightmost column, to provide additional intuition about the complexity, scale, and nature of the manufacturing ﬁrms we study, we also describe a particular ﬁrm chosen to be representative of the median Ethiopian manufacturing ﬁrm.