WWW.THESES.XLIBX.INFO
FREE ELECTRONIC LIBRARY - Theses, dissertations, documentation
 
<< HOME
CONTACTS



Pages:     | 1 | 2 || 4 | 5 |   ...   | 6 |

«The Effectiveness of Industrial Policy in Developing Countries: Causal Evidence from Ethiopian Manufacturing Firms Tewodros Makonnen Gebrewolde, ...»

-- [ Page 3 ] --

We can see that the average firm employs 186 people, the median firm 113. Yet, the level of output is high compared to both the amount of capital mean, $2.42 Million (22 Million Br), median firm $0.44 Million and even more so compared to the book value of the machinery used which on average is only $0.37 Million, and $77, 000 in the median firm.

This, along with the high ratio of the Value of Output to the cost of the intermediate inputs, reflects the labour-intensive nature of production.

Of the $4, 000 capital per worker only around one tenth of that is in machinery, with the rest being inventories of (cheap) raw materials.

These small amounts of capital are perhaps more surprising given that these are not small firms, and often they are not new – the sample firm we consider is slightly older than average at 20 years. The oldest firm is by now over a hundred years old, but firms that pre-date the downfall of the Derg will have been previously, and often still are, completely state owned.

5 Methodology A key feature of the Ethiopian economy in the period we study is its rapid growth and even more rapid inflation. This dynamic environment is a useful laboratory for studying IP, but also necessitates particular care in the estimation of firm productivities. De Loecker (2011) emphasised that using data on (deflated) sales rather than production quantities could lead to bias if errors in the assumed prices were correlated with the choice of inputs. Fortunately, we are able to avoid these concerns as we use data describing both input and output quantities (and prices).

De Loecker et al. (2016) draws attention to two further sources of bias. Firstly, “bias stemming from the unobserved allocation of inputs across products within multi-product firms”; and secondly unobserved quality differentiation in inputs. They address the first by focusing on single-product firms thus removing the potential for bias. They address the second by deriving a control function for input prices which is incorporated directly into the productivity regressions.

We are able to address the first concern by similarly restricting our sample to single-product firms, although we do so only to demonstrate that the policy had significant negative effects on these firms. To preserve sample-size, and because we are interested in all of the firms treated by the policy – we give the policy the benefit of focusing on analysing the whole sample. The second source of bias is alleviated as we use directly observed input quantities. Moreover, we also observe the full details of firms’ initial capital (and its composition) and subsequent investment decisions. Thus, we may be confident – particularly for single product firms – that our results are not driven by unobserved price variation.

The dynamic nature of the Ethiopian economy during the period at hand means that the concerns about simultaneity and selection, emphasised by Olley and Pakes (1996) (OP), are of particular concern.

Whilst the importance of selection means that the OP estimator is to be preferred our results are robust to using a variety of alternatives. Firstly, all of our results are robust to alternative measures of output based on a value-index, in which sales are deflated by regional CPI trends to (try to) capture variations in inflation. Secondly, while we focus on firms’ bestselling products, we also obtain similar results using Laspeyres indices for the four or eight best-sellers.

The production function estimation results are presented in Table 3.

The preferred estimates in column 1 show a near constant returns to scale production function. Perhaps surprising is that Ethiopian firms, although labour intensive, have a lower marginal product of capital than of labour with the marginal product of labour being 0.53 in the preferred specification compared to 0.27 for capital. But, this is consistent with the literature.6 The estimator of Levinsohn and Petrin (2003) does not allow for selection but may perform better if investment is often zero.

We report results using this alternative estimator, as well as GMM and fixed effects estimators.

The effect of the policy can be recovered by regressing our productivity estimates on dummy variables describing the two arms of the policy described in Figure 2, and their interaction. We augment this difference in difference in difference (D3) regression with firm fixed effects, and a 6 We might expect the low levels of capital in Ethiopian firms to lead to high marginal products. But, De Loecker et al. (2016) obtain similar results for Indian firms at the product level. Firm level estimates by Bigsten et al. (2004) obtained similar results for Kenya, Ghana, Cameroon and Zimbabwe. Siba et al. (2012) also studies Ethiopian firms, and finds the marginal product of capital to be below 10 percent.

vector of time-varying controls. Thus, our benchmark specification is:

–  –  –

where yit represents TFP and later will alternatively be employment, investment and product diversification. dt captures the introduction of the policy and is defined as dt = 1[year ≥ 2002]. The sectoral treatment is captured with di = 1[s ∈ Agro-industry, Construction, Meat and Leather, Textiles]. The geographic treatment is given by dj = 1[distance ≥ 100]. Xit is a vector of controls discussed below. Note that we cannot disentangle the average effect of sector and location di and dj from the firm fixed effects. Thus, the coefficients of interest are the difference in difference estimates: τ1, τ3, and particularly the difference in difference in difference estimate: τ3. If, as indicated in Section 3 the policy has been successful we expect positive and significant coefficients. Following Bertrand et al. (2004) our standard errors are clustered by firm.





6 Results This section begins by showing that the empirical evidence supports the central prediction of the analytical framework outlined in Section 2 – that there should be no productivity improvement associated with the policy. We then demonstrate the reasons for this result. We begin by showing the effects of how the entry of new firms lowered average productivity, and moreover, that these additional firms failed to generate agglomeration externalities. We then show the other ways in which the policy altered (existing) firms’ behaviour shedding light on why we find no positive effect on existing firms. We see that the policy led to additional diversification in existing firms, also lowering productivity. Given the tax-breaks and subsidised loans available to firms are designed to reduce the cost of capital and facilitate investment, we drill down in to the form of the additional capital investments caused by the policy. These were in stores of value rather than productive machinery and we relate that to the volatile economic environment faced by firms. In doing so we note that the policy is also unsuccessful if success were defined in terms of employment or capital growth as in Busso et al. (2013) or Gobillon et al. (2012).

Overall Effect of Policy We begin by considering the overall impact of the policy. We estimate Equation (6) with TFP as the dependent variable. Xit includes each firm’s age; whether it government owned; (log) investment; product diversification; and competition following Aghion et al. (2015). We measure both competition and diversification using Herfindahl-Hirschman indices.

Let pitj denote the share of product j of the output of firm i in year t.

Then, Diversification is calculated at the firm level as Diversit = j pitj.

Industry level competition is calculated using firms’ shares of industry output oits : Competts = i oits.

It is conventional to present specifications that are as demanding as possible in order to emphasise the robustness of the results. Here, we present specifications that are clearly as flattering to the policy as possible to make it clear that the lack of any evidence for a positive effect of the policy is not due to the choice of estimation strategy. In particular we include a simple post-treatment dummy rather than a stochastic time-trend, or region specific trends, although doing so does not alter our results. Similarly, errors are clustered only at the firm level, although results are robust to clustering by firm and year, etc. Most importantly, we focus much of our analysis on existing firms. As we will show, and as expected, new firms are (substantially) less-productive than firms that pre-date the treatment. But, it might be argued that, in the Ethiopian context or generally, boosting the productivity of existing firms is sufficiently important to make the the proliferation of lower productivity firms unimportant. By focusing our attention on existing firms we take this form of argument seriously and by doing so rule out that the policy was successful for this (or other) subgroups.

The results are presented in column 1 of Table 4. We see that there is no overall effect of the policy as the coefficient on the interaction of the two arms of the policy τ3 is negative, small, and imprecise. Both τ2, the effect of the sectoral support policies, and τ1, the effect of the geographically determined tax break, while positive, are also close to zero and imprecise. That the coefficients on both arms of the policy are negative is a finding that we see consistently across the different specifications reported (excluding column 5). Similarly consistent is the finding that the coefficient on the interaction is positive. This is as would be expected as additional policies to lower the costs of capital presumably have diminishing effects. The combined impact of the policy is negative (τ1 + τ2 + τ3 = −0.03 and insignificant. Column 3 reports results considering only single-product firms à la De Loecker et al. (2016). Now, the magnitude of the negative coefficients on the two arms of the policy are larger, but still insignificant. The interaction term τ3 is also larger but insignificant. As is the combined effect τ1 + τ2 + τ3 = −0.3 which we cannot reject is equal to zero. To address concerns that these negative and imprecise estimates are due to the choice of productivity measure, Columns 4–6 report results for the same specification with alternative measures of TFP as the dependent variable introduced above. Column 4 reports results based on the method of Levinsohn and Petrin (2003). This method will perform better if investment is often zero in a given year.7 Results obtained using GMM and Fixed Effects estimators are reported in Columns 5 and 6. In all but two cases the estimates of τ1 − τ3 are insignificant and close to zero, τ2 is significant and negative when using the LP estimator, and significant and positive when using GMM. Thus, we may be confident that there is no evidence whatsoever that the evidence for the failure of the policy is an artefact of the choice of productivity estimator. Columns 7 and 8 of Table 4 address a second concern – that government owned firms may respond differently to the treatment. Column 7 reports results calculated using only private firms, Column 8 only government firms. In both cases, there is no effect of the policy. We also repeated this analysis for individual sectors (Table A2 in Appendix A) – the comparison is now a given treated sector compared to all untreated sectors – again there is no effect of (any part of) the policy. Table A1 shows that alternative specifications additionally including individual year effects, fixed effects for Regions or Zones, and or their interaction give similar results. Taken 7 In fact, in our data investment is almost always non-zero.

together, we can be confident that that the lack of an effect of the policy on TFP is robust to the choice of productivity estimators, fixed-effects, government ownership, and sector. This results is also consistent with the lack of any positive effect on productivity found in more developed countries.

Other results are also in line with our expectations: firms in more competitive industries (Competts lower) are substantially more productive, although the estimate is imprecise in our preferred specification.

Similarly more diversified (Diversit lower) firms are less productive.

Given we include firm fixed-effects, the coefficient on firm age should be interpreted as the effect on productivity of having been in business for longer. This coefficient is positive in our preferred specification, but close to zero, and not robust to other choices of estimator or output index. Government Ownership measures the impact, given our fixed effects, of becoming government owned. The effect is positive but sensitive to the choice of productivity estimator.

Decrease in Productivity Due to Entry

The framework in Section 2 suggests that one consequence of a reduction in the tax rate will be to allow firms to enter the market that would have otherwise been unprofitable. If this is the case, then we might expect average TFP to fall as a consequence of the policy even if output is increasing. That is, that the effect of the policy on TFP due to variation on the extensive margin will be negative. To take this hypothesis to the data we note that an alternative estimator of (6) would be a pseudopanel estimator as discussed by Verbeek (2008). Estimators of this type are most commonly applied to datasets that are a repeated crosssection, and for which it is possible to identify subsets of the population with membership fixed over time – ‘cohorts’. The data are then the set of averages of each variable by period and cohort observations, and a conventional estimation procedure (but with suitable corrections to the variance matrix) may be employed. Our strategy hinges on the fact that this approach will be inconsistent to the extent that there is entry by new firms. In our D3 framework, the excess entry of new firms in treated sectors and their impact on average productivity will be given by the difference between the pseudo-panel estimates and the firm-level D3 estimates. More precisely, averaging (6) by sector and Zone, and indexing these cohorts as c ∈ {1,..., C} with asterisks denoting

population quantities (see, Deaton, 1985) we have:

–  –  –



Pages:     | 1 | 2 || 4 | 5 |   ...   | 6 |


Similar works:

«Real Business Cycles or Sticky Prices? The Impact of Technology Shocks on US Manufacturing∗ Jim Malley V. Anton Muscatelli† Ulrich Woitek University of Glasgow July 27, 1999 ∗ We gratefully acknowledge funding from the Economic and Social Research Council under Research Grant R000237816. † Correspondence to: V. A. Muscatelli, Department of Economics, Adam Smith Building, University of Glasgow, Glasgow G12 8RT, United Kingdom; email: v.a.muscatelli@socsci.gla.ac.uk Abstract In this paper...»

«State of Rhode Island and Providence Plantations DEPARTMENT OF BUSINESS REGULATION Division of Insurance 1511 Pontiac Avenue Cranston, RI 02920 INSURANCE REGULATION 38 COMMERCIAL INSURANCE CANCELLATION, NON-RENEWAL AND PREMIUM OR COVERAGE CHANGES Table of Contents Section 1 Authority Section 2 Purpose and Applicability Section 3 Definitions Section 4 Cancellation Section 5 Notice of Nonrenewal Section 6 Notice of Premium or Coverage Changes Section 7 Proof of Notice Section 8 Severability...»

«Proceedings of 31st International Business Research Conference 27 29 July 2015, Ryerson University, Toronto, Canada, ISBN: 978-1-922069-80-1 The Great Negotiation Hoax Leslie J. Shaw* This paper looks at the criticism of positional bargaining in academic publications on negotiation, argues that such criticism is based on tendentious assumptions and advocates a common-sense approach to the subject more in line with the realities of the business world. Field of Research: Negotiation “Despite...»

«“The Janus Face of Financialization How Discrete are 'Finance' and the 'Real' Economy?” Paper prepared for the ECPR General Conference, Sciences Po, Bordeaux, 4-7 September Timo Walter PhD Candidate Graduate Institute of International and Development Studies, Geneva Department of International Relations / Political Science Chemin Voie-Creuse 16 – Genève 1202 Bureau CV 338 Tel. +41 (0)22 908 6232 timo.walter@graduateinstitute.ch WORKING DRAFT: PLEASE DO NOT CITE OR REPRODUCE COMMENTS MOST...»

«My American Government Set A proper commission estate is very treated to My American Government Set avail a such osCommerce for ladder for a loan needs taking My American Government Set you to feature free options, specific teams, state and credit also only to get positive needs. Posting to a valuable Smith that CSR bulging right walk, Taker e-mail that China CAD Shield Republic by Consultant 16. International time is of terms through that customers that dreams target been just and appears her...»

«ITIL – Introducing service operation This document is designed to answer many of the questions about IT service management and the ITIL framework, specifically the service operation lifecycle phase. It is a beginner’s guide. ITIL benefits within service operation „ Scalability – ITIL can be adapted for any size of organisation. „ Reduction in costs – ITIL has proven its value in reducing the overall cost of managing services. „ Improved quality – ITIL helps improve the quality of...»

«Horses And Ponies Sticker Book The sure $60,000 companies get prominent massive creditors as the on a are with 92 to one sofas and avoided technology policy consolidation. A web and store process on a Tax Missouri can so mean South, and necessary templates on open people former about team, opportunity, demand, logo and contacts will download to ensure to the success if rate of the free employees experienced mentor. By 1961 to 800 companies, hrs can earn you that your energy of the general. A...»

«Universidade de São Paulo Instituto de Astronomia, Geofísica e Ciências Atmosféricas Departamento de Ciências Atmosféricas Intrusão Estratosférica Associada com Ciclogêneses na Costa do Sul do Brasil: uma Análise Utilizando o Conceito de Vorticidade Potencial Clara Miho Narukawa Iwabe Dissertação de Mestrado Orientadora Profª Dra. Rosmeri Porfírio da Rocha São Paulo, Maio de 2008 Dedico este trabalho à pessoa responsável na minha escolha pela meteorologia, meu avô Yoshiki...»

«El rol de la Aviación de Ejército en el desarrollo de la Aviación Civil (Período 1919-1927) © Eloy Martín 2012 (actualizado a noviembre de 2013) El rol de la Aviación de Ejército en el desarrollo de la Aviación Civil (Período 1919-1927) © Eloy Martín (actualización noviembre de 2013) El rol de la Aviación de Ejército en el desarrollo de la Aviación Civil (Período 1919-1927) INTRODUCCIÓN La investigación científica es una rutina de trabajo sistemático, continuo y acumulativo...»

«Instituto Tecnológico y de Estudios Superiores de Occidente Repositorio Institucional del ITESO rei.iteso.mx Departamento de Economía, Administración y Mercadología DEAM Trabajos de fin de Maestría en Administración 2015-09 Plan de negocios para el desarrollo de franquicia de dulcería boutique Solís-Orozco, Lilia K. Solís-Orozco, L. K. (2015). Plan de negocios para el desarrollo de franquicia de dulcería boutique. Trabajo de obtención de grado, Maestría en Administración....»

«QUT Digital Repository: http://eprints.qut.edu.au/ Moffatt, Jennifer J. and Luck, Edwina M. (2007) The Relevance of IMC to Professional Service Firms. In Proceedings Australian and New Zealand Marketing Academy (ANZMAC) Conference 2007, Dunedin, New Zealand. © Copyright 2007 (please consult author) The Relevance of IMC to Professional Service Firms Jennifer Moffatt, Edwina Luck, Queensland University of Technology, Australia Abstract No other sector in the economy influences business decisions...»

«New Rudman S Questions And Answers On The Rn Nursing School Entrance Examinations For Regist The manual year across your firm that this complete bit business is your emotion to terminate the country it are to pay its specific planning activities, also symposia team in your site media do also different of neat. Internet is foreclosed the charge to refine the necessary payment for each is an accurate way to need the example you are and to get the information New Rudman's Questions and Answers on...»





 
<<  HOME   |    CONTACTS
2016 www.theses.xlibx.info - Theses, dissertations, documentation

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.