Decision biases in commercial loan risk estimation

Journal of Bank Cost & Management Accounting, The, 1999 by Krause, James D, Joseph, Gilbert W

ABSTRACT

According to an accepted psychological theory, a long cue series in an estimation task may bias belief adjustment toward a primacy effect (first cues have more influence). Lending institutions often help loan officers to estimate the magnitude of credit risk by using commercial loan aids (long lists of financial ratios). Credit risk bias may result if more important financial ratios are not placed first in the series. Five "very important" financial ratios (determined by a pretest of highly experienced subjects) were placed differently in the cue series in an experiment involving bank loan officers. The primacy effect predominated even over higher taskspecific experience. Subjects with "very important" financial ratios early in the series performed the same as subjects shown only those ratios. Average credit risk for these two groups was most consistent with actual borrower risk.

INTRODUCTION

Hogarth and Einhorn (1992) formalized a theory of a "belief adjustment model" to explain how the order of information presented to a decision maker can influence a final decision. Building on years of research observations, this "belief adjustment model" demonstrated how the following factors can bias a decision maker's final belief:

Type of task performed-whether the task is to form a "true/false" impression (an evaluation task) or a magnitude of belief (an estimation task).

Length of the cue series-whether the list of cues is short or long ("long" being any list with more than 17 cues).

How the cues are processed-either the final belief is formed at the end-of-sequence (EoS) or in an incremental manner called step-bystep (SbS).

The initial anchor strength-whether the decision maker possesses preconceptions or the initial belief is based solely upon the cues observed.

Cue sophistication-whether each new piece of information is simple or complex.

Type of evidence presented by the cues-whether the cues consistently confirm or consistently disconfirm a prior belief, or whether the evidence is mixed (i.e., some cues confirm and other cues disconfirm prior beliefs).

Although never previously tested, Hogarth's and Einhorn's model could be applied to loan risk estimation by bank loan officers. A bank loan officer estimates the amount of credit risk associated with a loan application in order to determine if the loan should be given and to determine the interest rate to impose on the loan. Simply stated, such a task contains all the characteristics defined by the "belief adjustment model" and could result in a decision bias. Specifically a "primacy effect" should be observed. That is, cues evaluated early in the series should have a greater impact on the bank loan officer's decision than cues that appear later in the series. Hogarth's and Einhorn's theory suggests that a primacy effect results from tasks similar to credit risk estimation because they exhibit the following characteristics:

A loan risk assessment is an "estimation" type task (how much loan risk is involved?).

The series of cues is often long (many loan aid materials use lists of financial ratios comprised of 30 or more ratios, such as that developed by Robert Morris Associates).

When the series of cues is long, human cognitive limitations prohibit reviewing all the cues and then arriving at a belief (EoS mode). Instead the decision maker is forced to employ a step-by-step (SbS) mode whereby one or several cues are evaluated and the belief is readjusted repeatedly until all cues have been evaluated.

The initial belief anchor is often weak and is established by evaluating the cues (loan officers may know little about the borrower beyond financial information shown them).

According to the theory, whether the cues are simple or complex (some financial ratios can individually appear to be very complex, some are very simple) is less important when the list of cues is long. Similarly, whether the cues are consistent or mixed is less important than the length of the cue series. Given all these characteristics together, the "belief adjustment model" predicts that credit risk decisions should be affected by primacy. That is, cues evaluated first should affect the final judgment more than cues evaluated later in the series. If this is true, bank loan officers could be making decisions that are biased. They could be misjudging the actual credit risk. There is research evidence that a similar decision bias (i.e., recency) is robust and cannot be overcome. To test this premise, the researchers conducted an experiment using a commercially available loan aid (a list of 36 distinct financial ratios presented in a fixed order).

THE EXPERIMENT

The subjects were 216 trained commercial loan officers from the majority of banks in the local area (local, regional, and superregional banks). During a pretest, the ten most experienced loan officers were involved in a pretest and did not participate in the remainder of the experiment. These ten loan officers were shown the list of 36 financial ratios (with no values) and were asked to place each ratio into one of three categories reflecting the ratio's importance for assessing the type of loan tested"very important," "important," and "less important." All ten subjects selected five ratios as being "very important:" current ratio;