Modeling Short Term Debt: Panel Evidence from Turkish Manufacturing Firms

Using a mixture of macro variables such as growth and bank lending rates and firm-specific micro variables, this paper examines the determinants of short term debt using firm level panel data for Turkish manufacturing companies traded in Istanbul Stock Exchange. The original sample drawn from Finnet©’s Financial Analysis module included 183 firms with items extracted from the balance sheet of various manufacturing subsectors covering the 2002-2012 periods. The panel is unbalanced on account of some newly established firms entering the market within the period and after filtering some extreme observations. The choice of variables- with the exception of the unavailable risk proxy - follows the works of Bougheas et al. (2006) and Gonzales et al. (2007) in particular.Therefore, the ratio of short term debt to total debt is our dependent variable and the firm’s age, size, return on equity, gearing and collateral are firm-specific explanatory variables. The age variable is calculated by subtracting the date of foundation from the current year whereas the size can be calculated by deflating the firm’s total assets by producer’s price index. Both are expressed in natural logarithmic forms to account for the scaling issue. The collateral variable can be found by taking the ratio of tangible assets to total assets and since it is expected that the ratio of tangible assets in total enhances access to longer-term debt thus reducing the proportion of short-term to total debt, we expect an inverse relationship with short term debt. The ratio of total debt to shareholder funds as an indicator of indebtness of firms in relation to their equity is represented by the gearing variable. In addition to these micro level effects we also introduce the growth rate of real GDP to account for cyclical effects and the bank’s weighted average lending interest rate. The source for these two came from Central Bank of Turkey (CBT) dataserver.To capture the impact of the ripple effects of the global crisis in 2008 we also introduce a time dummy. The modeling framework is based on dynamic panel estimation techniques which would be natural candidates since they allow us to properly estimate models in which the endogenous variable-which is the firm debt variable-, is lagged one year. Such models include both lagged dependent variables and unobserved firm-specific effects in the specification and therefore these models are powerful because they allow for empirical modeling of dynamics while accounting for individual-specific dynamics. We use both Arellano-Bond and Arellano-Bover methods. The estimation results indicate that firm size and collateral variables are both significant and in line with our expectations. Moreover, the impact of the persistence of the debt structure and the global crisis can also be justified. On the other hand, gearing, age and return on equity variables don’t seem to be influential on firm’s short term debt structure. The two macroeconomic variables are not found to be significant either.