Political Connections Reduce Job Creation: Firm-level Evidence from Lebanon

Abstract

Using firm-level data, we document that politically connected firms (PCFs) create more jobs than unconnected firms in Lebanon. We observe, however, that the presence of PCFs in a sector is correlated with lower job creation. Although causality is difficult to establish due to endogeneity issues, we find that PCFs expand, and non-PCFs retract, more around elections. Our findings are consistent with the hypothesis that unfair competition by PCFs hurts unconnected competitors so much that aggregate employment growth in the sector is affected negatively.

Acknowledgments

An earlier version of this paper has been circulated under the title “Do Political Connections Reduce Job Creation? Evidence from Lebanon.” We acknowledge research funding from the Economic Research Forum. We thank Khalid Abu-Ismail, Mohammad Amin, Jihad Azour, Jose de Sousa, Shantayanan Devarajan, Sergei Guriev, Philip Keefer, Adeel Malik, Bob Rijkers, Charbel Nahhas, Nisreen Salti, as well as seminar participants at Harvard University, Université Paris-Dauphine, University of Massachusetts (Boston), Pomona College, University of Oxford, University of Massachusetts (Amherst), The American University of Beirut, The American University in Cairo, The Doha Institute for Graduate Studies, World Bank MENA Chief Economist Office, UN-ESCWA, and EBRD Chief Economist Office for useful comments, insights, and suggestions. Haya Anis provided effective research assistance. Jawad Ahmad Hassan provided inspiration. We thank the Directorate of Revenues, Lebanese Ministry of Finance (MoF) for granting us access to proprietary data subject to complying with the confidentiality requirements set by Lebanese Law. We are solely responsible for the conclusions and inferences drawn from these data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary materials

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2020.1849622.

Notes

1. It is important to recognise that there are payoffs beyond job creation that politicians may enjoy. For instance, political support by politically connected media is more likely to be in influencing public opinion rather than in creating jobs.

2. International Labour Organisation estimates show that total youth unemployment ranged between 17–23 per cent between 2005 and 2010 in Lebanon.

3. The restricted coalitions of the 1960s allowed for a large degree of macro stability that oversaw some of the fastest economic growth rates in the world. However, the narrow socio-economic coalition could not hinder the rise of groups that fell excluded, and the country burst increasingly into cycles of grievances and violence that ultimately led to the civil war of 1975–1989.

4. The Ta’if Agreement, also known as the National Reconciliation Accord, was reached in October 1989 to provide the basis for ending the Lebanese civil war. It was negotiated and signed in Ta’if, Saudi Arabia.

5. Different types of rents can differ in the extent to which they support economic growth (rents from cooperation) versus the extent to which they tax growth (external rents that generate political divisions, regulatory rents). The first type relates to the conditions of security that allowed Lebanon, at a time, to become a touristic entrepot and banking centre in a troubled region. The second type relates to the use of state provided patrimonial instruments (such as state employment, controls over various funds, choice of infrastructure investments) as well as economic distortions (such as access to public procurement, and the way in which the policy framework benefits particular groups).

6. The distribution of state favours in the post-war period neglected redistribution to the poorest in favour of equal shares to the different communities. Group composition and share of public spending were strikingly equal, based on distribution of public capital expenditure (1996–2005) and distribution of registered voters (Salti & Chaaban, 2010).

7. For example, The Council of Development and Reconstruction (CDR), the Council of the South, the Ministry of Energy (and its lucrative import of oil), and the Ministries of the Displaced, Public Works, and Health were under the suzerainty of different political groups (Leenders, 2012).

8. Public sector recruitment also rose from 3300 positions in 2008 to 5941 in 2009 but then dropped to 762 in 2010 (see Abou Jaoude, 2015).

9. These firms are legally bound to be also registered in the Commercial Register and at the National Social Security Fund (NSSF).

10. To check the accuracy with which the MoF data captures new firms, we compared the sector-capital-year-date-of-birth-data in the MoF dataset with the CR registry data. We found that the MoF data identified firm start dates accurately.

11. We defined ‘abnormally high’ volatility as a change that was equivalent to more than 100 per cent between t and t+1 followed by a change that took the output per worker level to less than its initial (t) value at t+2. An example of reporting error is when a firm is reported to pay taxes before it is established.

12. Ayyagari et al. (2014), Aga et al. (2015), and World Bank (2014).

13. In Turkey, Tunisia, and Jordan only 20 to 30 per cent of labour work in firms with more than 10 workers, while in Egypt and the West Bank and Gaza this figure is below 10 per cent. The share of employment in firms with less than five employees is much larger in Egypt and the West Bank (about 60%), Jordan (40%), Tunisia (37%), and even in Turkey (34%) (Figure 1.5, page 19, World Bank 2014).

14. This contradicts the result in World Bank (2014), which, using the same dataset, claims that most of the new jobs in Lebanon during the period were created by micro-firms. Upon further inspection, it turns out that this study mistakenly coded all employment in micro-firms as new jobs.

15. Haltiwanger, Jarmin, and Miranda (2013) found similar trend for United States as well. See Evans (1987), Klapper and Richmond (2011), Neumark, Wall, and Zhang (2011), and Van Biesebroeck (2005) for related evidence.

16. Using the framework of Macneil (1978), we considered that firm-level political connections in Lebanon are relational, as there is anecdotal evidence that firms in Lebanon maintain long-term benefits from political connections, instead of transactional, and directed towards short-term economic benefits.

17. We checked whether the effect of connections is different for those who were in office in 2005–2010 compared to ones who were in office before then, and did not find a statistically significant difference.

18. The process of matching (Arabic) names creates possibilities of errors when different database use a different spelling, or when different individuals have the same name. We tried to minimise this error by allowing for common spelling variants, and matching first, middle, and last names before classifying a firm as politically connected.

19. Admittedly, this procedure may exclude PCF firms owning other firms, if no politically connected individual is involved publicly in that lower tier firm. There is also a risk that our measure is correlated with a firm’s attributes, such as its number of owners, since having more owners may increase the chances of matching.

20. All the sectors identified by Leenders (2012) as politically connected are captured by our methodology.

21. T-tests have also been calculated and show the mean differences are statistically significant. T-test results are not reported for space reasons.

22. Bertrand et al. (2018) provide support for this hypothesis using data from France.

23. This supposes a long-term relation of trust, which is typically provided by repeated games that end up constituting a relational contract.

24. We are not in a position to determine the strength of political connections of firms as that involves making subjective judgements about how much influence each politician possesses. In addition, given that we consider firm-level political connections to be relational instead of transactional, and that we did not find statistically significant difference when we tested whether the effect of connections is different for those who were in office in 2005–2010 compared to those who were in office before then (see section 3.3), we did not differentiate between different types of connections based on alignment with the ruling coalition.

25. The 2009 elections resulted in winners and losers. One would expect positive impacts amongst firms aligned with the winners, and negative impacts among firms that are not. But our dataset ends in 2010 and thus does not allow further analysis of effect of election results on corporate behaviour.

26. HHI is defined as the sum of the squares of the market output shares of the firms within the sector, where the market shares are expressed as fractions. The result is proportional to the average market share, weighted by market share. As such, it can range from 0 to 1, moving from a large number of very small firms to a single monopolistic firm. To facilitate empirical interpretation, we multiplied HHI figures by 100; so, in the above estimations HHI figures range from 0 to 100.

27. For robustness checks, i.e., to reduce selection bias and avoid generating a spurious correlation between political connectedness and employment growth at firm level, in a separate exercise, we also limited the analysis to firms with more than 50 employees. Results hold but are not presented for space purpose.

28. Along the lines of the firm growth rate that was introduced by Davis, Haltiwanger, and Schuh (1998), NJC-f in column 3 of Table 2 refers to the log of net employment growth, using our measure of firm-level employment growth which is equal to the log of (change in employment from year t−1 to year t, divided by employment in year t−1).

29. After controlling for firm size and age, column 3 of Table 2 shows that annual growth rate of employment in PCFs is, on average, 24.23 per cent [100*(EXP(0.185 + 0.047 + 0.023 + 0.015−0.053)−1)] higher than annual growth rate of employment in NPCFs that operate within their same sector. Note that in computing these effects, we replace non-significant estimates by zero.

30. Moreover, Chaaban (2019) shows that banks in Lebanon are overwhelmingly owned by politically connected individuals, and we can thus expect them to lend disproportionally to firms that belong to their corporate networks.

31. Compared to NPCFs, PCFs pay 21.65 per cent [100*(EXP(0.151 + 0.037 + 0.028–0.02)−1)] higher wage per employee and enjoy 30.08 per cent [100*(EXP(0.238 + 0.033 + 0.016–0.024)−1)] more output per firm while output per employee is 26.94 per cent [100*(EXP(−0.228−0.051−0.019−0.026+0.01)−1)] lower.

32. Annual growth rate of employment in a PCF is, on average, 28.78 per cent [100*(EXP(0.186 + 0.038 + 0.018 + 0.011)−1)] higher than annual growth rate of employment in non-PCF that operate within their same sector.

33. We also checked whether political connections correlate with market concentration. We found on average significantly higher market concentration in sectors where a higher number of PCFs exist – although some highly concentrated sectors are not dominated by PCFs, and some sectors dominated by PCFs are not highly concentrated.

34. For robustness checks for all estimations in Table 2, i.e., to reduce selection bias, in separate exercises, we also limited the analysis to firms with more than 50 employees. Results hold but are not presented for space purpose. And, all regressions include time dummies for all years, not just 2009. The Year and PCF*Year interactive terms had coefficients that are much smaller while still positive in terms of magnitude in years other than 2009 but that are not significant at conventional statistical levels.

35. Also, Aghion (2009) reported empirical tests of predictions of the model with respect to the effects of product market competition and entry deregulation on growth.

36. A related argument is that rent-filled sectors may be naturally less competitive, for example because of high entry costs, and thus have lower entry and exit rates.

37. There are 289 sectors disaggregated at the 4-digit level in Lebanon. 29 of these sectors include PCFs.

38. All regressions in Table 3 include time dummies for all years, not just 2009. The Year and PCF*Year interactive terms had coefficients that are much smaller while still negative in terms of magnitude in years other than 2009 but that are not significant at conventional statistical levels.