Endnotes

¹Sri Lanka’s 1977 economic reform program had measures to create a liberal FDI regime: liberalization of foreign investment laws, tax holidays on profits and salaries, duty-free access to imported inputs and the setting up the Katunayake Investment Processing Zone (Ganeshamoorthy, 2002). The clothing industry saw a twenty-two fold increase in annual FDI inflows from $12 million in 1987-89 to over $270 million in 2000-2005.

²Empirical studies on clothing, largely at the macro and sectoral levels, have focused on the trends and determinants of FDI inflows. For a selection see Lakshman (1989), FIAS (1993), Kelegama and Foley (1999), Athukorala and Rajapathirana (2000), and UNCTAD (2004). The entry of export-oriented FDI into Sri Lankan clothing is attributed to: a strategic geographical location, access to multi-fibre agreement (MFA) quotas, attractive investment incentives, and ample supplies of low cost, trainable labor. Locational disadvantages are said to include: political instability, poor quality infrastructure, and weak sub-contractors/suppliers to MNCs.

³The major trade theories (the Heckscher-Ohlin model, theories of economies of scale and oligopolistic competition, neotechnology theories, and theories of economic geography) attribute the export performance of a small open developing economy (e.g., Sri Lanka) to its comparative advantage over another in terms of access to certain factor inputs—capital, labor, economies of scale, technology, and geography (for surveys see Wakelin, 1997; Deardorff, 2005). Empirical applications to developing countries have sought to explain the export performance of each industry/product in terms of their various characteristics.

⁴The share of skilled production workers and the average wage, however, do not show up as significantly different between the two groups.

⁵Of the 205 clothing firms in the survey, 86 have zero export values. One of the problems in the estimation of the determinants of the export ratio is that there may be selectivity bias if we were to include only firms with positive exports. The Tobit model, however, includes all firms, i.e. also those with zero exports. See Maddala (1983) for a discussion of Tobit models.

⁶Owing to data constraints, many studies of firm-level exporting only include factory floor skills, proxied by the average wage. In this study we use a more appropriate measure that takes into account different levels of skills—the skill-adjusted wage rate in relation to productivity. This is defined as follows: W/S = (W/L) / (S/L) where W/S is the share of wages to sales, W/L is the skill-adjusted average wage rate, and S/L is the sales per worker.

⁷This was measured discretely by a scoring system of seven categories ranging from below secondary schooling (1) to post-graduate degree (7). While this variable can be treated as a dummy variable with six values, we treated it as a single variable using it as a scoring system.

⁸Some correlation (0.36) between employment and foreign equity (see Table A1 [ PDF 67KB | 1 pages ]) indicates the possibility of multicollinearity. Hence, firm size was captured by a dummy variable that takes a value of 1 if a firm is large (more than 100 employees) and 0 if a firm is small (equal to or more than 100 employees).

⁹Drawing on the Lall (1992) taxonomy of technological capabilities, the ranking procedure integrates objective and subjective information into measures of a firm’s capacity to set up, operate, and transfer technology. Five technical functions performed by the Sri Lankan clothing firms were highlighted (including search for technology, inventory control, process adaptation, minor adaptation of products, and new product introduction) and a score for each function was awarded based on the assessed level of competence in that function. A firm is ranked out of a total score of 5 and the result is normalized to give a value between 0 and 1. This figure can be interpreted as the overall capability score for a firm.

¹⁰A correlation test was used to detect multicollinearity. The correlation matrix of variables in Table A1 [ PDF 67KB | 1 pages ] shows that no large correlations were noted between any of the independent variables, thus indicating that multicollinearity was not a problem. The Goldfeld-Quandt test revealed an F-statistic of 1.006 with 1% level of significance indicating mild heteroskedasticity in equation (2). To correct for heteroskedasticity for the Tobit model, equation (2) was re-estimated using interval regression, which allows the use of robust option to obtain the Huber-White robust standard errors. The estimated coefficients and their levels of significance using this method turned out to be almost the same as those of the Tobit estimates. Hence, the Tobit estimates were retained for analysis.

¹¹The variance inflation factors for the independent variables specified in the fitted model were considerably low, suggesting that multicollinearity was not a problem in equation (2). Some heteroskedasticity was expected to be present in a sample of varied firm sizes and was confirmed using the Breusch-Pagan/Cook-Weisberg test for heteroskedasticity for linear regression models. Accordingly, equation (2) was re-estimated using the robust variance estimator, which corrects for heteroskedasticity, and the results are presented in Table 3.

Download this Discussion Paper [ PDF 248KB| 18 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.