Mechanical Mean Reversion of Leverage Ratios: Analysis of South Korean Firms

Abstract

The traditional view that firms have target leverage ratios has been challenged given that a firm’s leverage ratio could mechanically mean revert whether or not a target leverage level actually exists. Evidence of leverage ratio mechanical mean reversion for U.S. firms has been well documented by Chang and Dasgupta (2009). A replication of the analysis using data on South Korean firms highlight the potential that the mechanical mean reversion of the leverage ratio also exists for South Korean firms, challenging the interpretation of prior research that infers target behavior of firms based on tests using target adjustment models. Although one might be tempted to deduce that South Korean firms follow target behavior based on the fact that the mean reversion parameter obtained from the regression analysis yields statistically significant parameter values, our analysis shows that similar statistically significant values for the mean reversion parameter can also be obtained when simulation samples with non-target random financing is used. The stock and financial accounting data for South Korean firms was obtained from FnGuide’s Data Guide database for all firms listed on the Korea Composite Stock Price Index (KOSPI) and the Korean Securities Dealers Automated Quotations (KOSDAQ) over a period of 16 years starting from 2000 to 2015. All values were converted into 2010 constant values in order to eliminate the effects of inflation with industry effects being controlled for within the regression analysis using the KSIC (Korea Standard Industrial Classification) codes at the twenty-one single alphabet level. Following the exclusion of non-financial firms, as well as firms with missing book value of assets, an additional requirement that firms have at least five years of continuous and non-missing accounting and stock price information for inclusion was imposed. This particular requirement was imposed to ensure that the leverage ratios in the simulation samples are given an acceptable length of time to evolve differently from the actual leverage ratio when random financing is assumed. Finally, firm-year observations with negative or greater than one book leverage or that have incomplete data were dropped to produce a final unbalanced panel data set that has 20,102 firm-year observations for 1,689 firms. Using the obtained panel data, three different simulation samples were generated by assuming different financing behavior and using data from the actual sample. The first simulation sample, denoted by S(p= 0.5, actual deficit), assumes an equal probability of debt issuance and repurchase, that is p is assumed to be 0.50. The second simulation sample obtains p using the actual financing deficit and newly retained earnings in the actual data and is denoted by S(p=empirical frequency, actual deficit). The third simulation sample, denoted by S(p=0.5, random deficit), was generated to remove any potential effects of endogeneity in the actual financing deficit as endogeneity could cause the simulated sample to produce results that are consistent with target behavior. Panel regression of these simulation samples, along with the actual sample, was conducted using a modified version of the leverage model that has leverage ratio as the dependent variable and the lagged leverage ratio as well as various firm-specific variables as the independent variables. The parameter coefficient estimates for the simulation samples were obtained by taking the average of 500 replications of the particular simulation. The signs for the coefficients on the lagged leverage ratio and firm-specific variables obtained from the leverage regressions were in the expected directions with respect to the leverage ratio. The significance of the coefficients can be viewed as verification that the leverage ratio for South Korean firms is related to the firm-specific variables selected for the leverage regression analysis.