Firm dynamics in the Cambodian garment industry: firm turnover, productivity growth and wage profile under trade liberalization

Abstract

The international garment trade was liberalized in 2005 following the termination of the Multifibre Arrangement (MFA) and ever since then, price competition has intensified. Employing a unique firm dataset collected by the authors, this paper examines the changes in the performance of Cambodian garment firms between 2002/03 and 2008/09. During the period concerned, frequent firm turnover led to a growth of the industry's productivity, and the study found that the average total-factor productivity (TFP) of new entrants was substantially higher than that of exiting firms. Furthermore, we observed that, thanks to productivity growth, an improvement in workers’ welfare, including a rise in the relative wages of the low-skilled, was taking place. These industrial dynamics differ considerably from those indicated by the ‘race to the bottom’ argument as applied to labor-intensive industrialization in low-income countries.

Keywords:

• Cambodia
• firm turnover
• garments
• Multifibre Arrangement
• race to the bottom
• total-factor productivity

Acknowledgements

We thank two anonymous referees for helpful comments. We also express our special thanks to all the Cambodian garment firm managers who kindly participated in our surveys in 2003 and 2009. We are grateful to the substantial assistance from the Council for the Development of Cambodia (CDC), the Cambodia Investment Board (CIB) and the GMAC. Our field surveys would not have been possible without the hard work of staff belonging to LIDEE Khmer in 2003 and the Economic Institute of Cambodia (EIC) in 2009. We thank all the seminar participants at the EIC-IDE joint seminar held in Phnom Penh on 5 August 2010, and those at the 12th International Convention of the East Asian Economic Association. The research for this study was financially supported by the Institute of Developing Economies, Japan External Trade Organization (IDE-JETRO). An earlier version of this paper was released as an IDE discussion paper. All remaining errors are our own.

Notes

1. Calculated from Cambodia Ministry of Commerce data and from ADB (2009).

2. According to Chan and Sok (2007, 19), in 2005, 270,000 workers sent home around US$50 million, which was equivalent to 3% of the total income generated from the whole of the agriculture sector.

3. Although sample sizes of some firm surveys such as ILO-BFC (various years), Oka (2010) and Kang and Dannet (2010) are relatively large, their focus is on labor-related issues and they did not collect firm-performance data.

4. Some 88% of Cambodian garment exports to the USA were also in direct competition with exports from Vietnam (EIC 2008). Buyers frequently claim that Vietnam has a more attractive investment climate in terms of investment security, lower energy costs and what is perceived as a skilled labor force (Chan and Sok 2007, 23–35). So far as Cambodian garment exports are concerned, Vietnam has emerged as a much stronger competitor especially since 2007, when Vietnam became a member of WTO.

5. According to data of the Cambodia Ministry of Commerce, there were 192 and 251 garment firms as of September 2003 and 2009, respectively. As we missed the firms not responding to the questionnaire, our sample is not randomly sampled from the population. However, as explained in the next paragraph, share of exiting and entering firms are very close to that in the population. Also, the average gross product and the number of workers are comparable with those taken from the industry-level statistics calculated based on Ministry of Economy and Finance and Ministry of Commerce. (The results will be provided on request.) Those evidences show that our sample does not suffer from significant sample-selection bias.

6. Note that our survey identified firm entry and exit at the time of survey, 2003 and 2009.

7. Refer to Appendix 1 for selection of the sample used for productivity analysis. Despite significant reduction in the number of samples particularly in 2008, the average gross product and the number of firms are similar to those taken from the industry-level analysis as explained in the Note 5.

8. These changes reflect effect of exit and entry of firms as well as change of the firms in operation throughout the period. The latter firms grew in gross product and value-added, but labor and capital value decreased on average.

9. Our estimate of profits includes tax. Note that part of the capital cost, such as rent and dividends, may not be subtracted in cases where a firm owner has used his or her own assets for the purchase of land, premises and equipment, unless the details are shown in an accounting book. Hence, profits may be overestimated, but we think that change of profit is more reliable assuming that ignorance of the owner's contribution occurs at roughly same frequency in both surveys.

10. It is noted that a subcontractor does not purchase material, and thus, their cost structure significantly differs. The figures in Figure 2 are based on the firms with non-zero material costs. If the sample is restricted to the firms whose share of material cost is greater than 20%, the share of material cost becomes 54.5% in 2002 (N = 68) and 47.4% in 2008 (N = 13), and the share of profits increases from 15.3% in 2002, to 18.8% in 2008.

11. If a firm manager can predict productivity in the next period, he or she can adjust the labor input. Thus, labor input may be correlated with the residual through correlation with productivity (Marschak and Andrews 1944).

12. High-skilled labor includes managers/executives, other officers, engineers, supervisors and quality controllers. Low-skilled labor includes operators, helpers and holders of other miscellaneous jobs. Constant returns to scale (CRS) are assumed based on the fact that the hypothesis test of CRS for estimated parameters by OLS could not be rejected. Refer to Appendix 1 for detail of input and output data.

13. Absolute value of covariance tends to be greater for the group of firms with greater average output. Division by average output controls difference of output size across the firm groups.

14. It is reported that some garment firms close operations when incentives granted for exporters, i.e. tax waiver, expire, and set up a new firm with a different name. By tracking change of company's name using questionnaire and the GMAC member list, we identify nine ‘re-entrants’. However, since we may have missed some re-entrants, good performance of the entrants may be partly attributed to them. We thank the referee for this point.

15. The model indicates that the exit threshold clearly separates exited firms from continuing firms, but our results do not exhibit such a threshold and productivity distribution of exited and continuing firms partly overlaps. This can be explained by heterogeneous factor prices and transaction costs as well as a lag between the timing of the productivity measurement and the actual exit in our dataset.

16. Sample size (number of firms) was 164 in 2002 and 121 in 2008, respectively.

17. The figures are for the sewing section. The average working hours per month were estimated by multiplying the average working hours per shift per day (9.2 hours in 2002 and 9.3 hours in 2008) by the average annual working days divided by 12 months (23.6 days in 2002 and 24.4 days in 2008).

18. Since we do not have working hour data for highly skilled workers, we examined monthly wages.

19. The decrease in wage premium for working experience is bad news for experienced operators. This might be related to changes in labor contracts: for instance, employers might have increased the use of short-term contracts and may have decided to renew them in order to avoid having to award a seniority bonus for the experienced workers. Unfortunately, our surveys did not ask about the duration of labor contracts. This aspect could be included in future research.

20. The monthly minimum wage in the Cambodian garment industry was raised further to US$61 (including US$6 COLA) in October 2010.

21. We asked firm managers about average education level by job categories. Based on these information, estimated average years of education have been computed by assigning the following years of education to the original education categories in the questionnaire: Below primary (0 years); Primary (6 years); Lower secondary (9 years); Higher secondary (12 years); Bachelor's degree or higher (16 years).

22. Our analysis mostly applied to the period between 2002 and 2008. After late 2008, the Cambodian garment industry was hit hard by the global financial crisis. Firm exits increased and both garment exports and employment declined. However, since 2010, garment exports and employment have started to increase again and new factories have opened (Cambodia Daily 2010a, 2010b). Such a rapid recovery, even under conditions of increasingly tough price competition, seems consistent with our findings on the productive features of Cambodian garment firms.