A brief note on thrift failures: a more rigorous analysis of causal factors.

• Author: Belton, Willie J., Jr.

1. Introduction

   The Savings and Loan (hereafter S&L) industry has experienced widespread economic difficulties in recent years. These difficulties are reflected by, among other things, a failure rate of federally insured S&Ls in magnitudes unprecedented since the Depression years prior to 1934. From the beginning of 1934 through the end of 1979, there were comparatively few S&L failures. Indeed, from 1934 through 1979, the number of S&L failures reached double digits in only three years, 1941, 1966, and 1971. By contrast, the 1980s witnessed the failure of 525 federally insured S&Ls.

   In a well-known study, Barth [2] develops an in-depth framework to explain the causes of S&L failures. The objective of this note is to subject the model developed in Barth [2] and empirically tested using OLS by Cebula and Hung [10] to more rigorous testing by employing the method of cointegration. Each variable in the analysis is examined to determine its order of integration, i.e., to determine whether each is stationary in levels or in first differences. We then examine the variables for cointegration using the Johansen-Juselius [14] procedure.

   Section II of this paper presents the empirical model tested by Cebula and Hung [10], identifying specific factors that are hypothesized by Barth [3] as having had an impact on S&L failures. Section III describes the cointegration approach and also provides the results of the cointegration test. Finally, concluding remarks and observations are found in section IV.

2. The Framework

   The model below is developed in Barth [2] and is estimated by Cebula and Hung [10] using OLS:

   SLFR = f(INS, OILP, COST, MORT, TAP) (1)

   where SLFR is the S&L failure rate; INS is the ceiling level of federal deposit insurance per insured account, in constant dollars; OILP is the price per barrel of imported crude oil, in constant dollars; COST is the nominal annual average cost of funds to S&Ls; MORT is the nominal annual average S&L mortgage rate; and TAP is the ratio of tangible S&L capital to assets.

   Based on the arguments in Amos [1], Loucks [15; 16], and Cebula and Hung [10] we expect:

   [f.sub.1] [greater than] 0, [f.sub.2] [less than] 0, [f.sub.3] [greater than] 0, [f.sub.4] [less than] 0, [f.sub.5] [less than] 0. (2)

   The data used here to measure the variables listed in equations (1) and (2) are semi-annual data. Given restrictions on the availability of certain of the data [COST and MORT], the time period examined runs from 1963 through 1989. Therefore, with two data points for each year, the study deals with a total of 54 data points. The data for variable SLFR are the percentage of federally insured S&Ls that failed in either the first half or the second half of each year that was studied. The variable SLFR includes all of those S&Ls that were either liquidated or forced to merge with other (healthier) institutions with the assistance of the FSLIC. Those thrifts that voluntarily merged with other institutions without FSLIC assistance are not included in this variable. As defined, the variable SLFR precisely corresponds to those thrifts that Barth [2] classifies as "resolutions." As Barth [2, 31] observes, "Liquidation and assisted merger, generally referred to as resolutions, were meant to be final and to impose costs on the FSLIC." We also observe that the variable SLFR is to some extent a political variable such that the timing of the S&L failures was not entirely related to the economic variables identified in equation (1). We have undertaken two courses of action to allow for this political dimension of the thrift problem. The first is to include a variable to expressly allow for the "forbearance" phase of the thrift crisis. The second is to adopt a measure of S&L capital that expressly excludes "good will." Federal regulators had permitted S&Ls to include a value for good will in their computations of capital, presumably enabling the S&Ls to create the illusion of solvency and postpone resolution. Both of these courses of action are elaborated upon in the text that follows.

   The variable INS represents the ceiling level of federal deposit insurance per account, expressed in thousands of 1982 dollars, in either the...