Presentation is loading. Please wait.

Presentation is loading. Please wait.

Variable Selection: A Case of Bank Capital Structure Determinants

Similar presentations


Presentation on theme: "Variable Selection: A Case of Bank Capital Structure Determinants"— Presentation transcript:

1 Variable Selection: A Case of Bank Capital Structure Determinants
Nonna Sorokina David Booth Wake Forest University Kent State University

2 OVERVIEW Bank capital serves as a source of stability and provides a safety net. Thus it is heavily regulated. For economic reasons banks often hold capital beyond regulatory requirements. Understanding these reasons is important for: Efficiency of bank regulations For risk management of banks For investors and customers assessment of a bank’s soundness

3 GOALS To determine the important determinants (predictors) of bank capital structure with lasso Lasso not so robust to outliers thus we divide our data into two groups a) with no outliers b) only outliers

4 GOALS (continued) We find differences in the two groups in terms of the predictors We find evidence of moral hazard in Systematically Important Financial Institutions (SIFIs), i.e. their capital structure is independent of risk and collateral

5 LITERATURE Grop and Heider (2010) studied100 largest U.S. banks and 100 largest E.U. banks. They found: a) substantial variation in equity capital ratios b) most reliable factor of nonfinancial firms leverage (Frank and Goyal (2009)) are also significant for the bank’s leverage in their sample – used lasso

6 METHODS Increase sample size (leverage ratios vary significantly implying potential presence of discretionary capital). Extend period – use 1973 to 2012 to look at times a) without uniform capital requirements b) with uniform capital requirements but no risk-weighting of assets (Pre- Basel)

7 METHODS (continued) Test all banks and SIFIs separately during the three regimes Original results by OLS and Huber M- estimator (robust to outliers) look at a) outliers b) VIFs

8 METHODS (continued) Variable selection by adaptive lasso (Zou (2006)) produces Properly estimated coefficients Adjusts for multicollinearity and provides Best Predictive Variable Selection (oracle property) Look at individual banks only outliers and non-outliers (most banks) separately

9 HYPOTHESES 1. Market to Book, Profit, Size Tangibility and Risk are significant determinants of leverage in broad sample of U.S. banks 2. Size and sign of the corresponding coefficients are similar to Gropp and Heider (2010) 3. Risk is not a significant determinant of leverage for SIFIs due to implied guarantee(e.g.too big to fail); other coefficients may be different when compared to all banks.

10 HYPOTHESES (continued)
Leverage = β0+ β1MTB + β2Profit + β3Size + β4Collateral + β5Dividend + β6Risk + u

11 DATA From: COMPUSTAT North America COMPUSTAT Bank CRSP Missing data implies observation not included RESULTS Full paper with Tables available on request

12 RESULTS For a typical bank, leverage is a function of collateral, size and risk, after Pre-Basel (uniform capital requirements). After Basel (risk weighted capital requirements) more freedom is seen in the variables predicting leverage.

13 RESULTS (continued) The SIFIs are different. These results differ significantly across the regulatory environments. Lasso results show that typical SIFI leverage decisions DO NOT depend on risk or collateral in an regulatory period. Lasso works well in this type of study. It shows outliers are LESS risky for SIFIs.

14 RESULTS (continued) Hmmmmmm!


Download ppt "Variable Selection: A Case of Bank Capital Structure Determinants"

Similar presentations


Ads by Google