A comment on the empirical distribution of squared unexpected returns

Article Abstract:

Lobo and Mahmoud's (1989) study on the distribution of the normal theory test statistic obtained from two-day squared standardized unexpected returns is replicated with regard to firm size. Their conclusion that the test statistic distribution means are lower than normal theory suggests is not borne out, nor is their finding that large values for such statistics may result from smaller firms conclusively supported. The results suggest that controlling for firm size is essential in using squared returns in comparative studies, and that theoretically-derived means are inappropriate in such examinations. Lobo and Mahmoud's findings may have resulted from two errors in the programs used in their study which may have caused a FORTRAN-compiler-specific error.

User Contributions:

Comment about this article or add new information about this topic:

Comment:  (50-4000 characters)

Name:

E-mail:

Security Code:

Noncontrollable costs and optimal performance measurement

Article Abstract:

It is suggested that non-controllable cost allocation in a sequential department setting could be explained by intermediate division heads' private data or by a 'noisy' quality-generating process. A proposition is presented to demonstrate that under a linear technology in combination with a bivariate standard normal distribution, it is optimal for companies to allocate non-controllable intermediate expenses so that the compensation of each final product division manager depends on that division's revenue.

User Contributions:

Comment about this article or add new information about this topic:

Comment:  (50-4000 characters)

Name:

E-mail:

Security Code:

An empirical comparison of probit and OLS regression hypothesis tests

Article Abstract:

Rejection regions for probit test data are poorly specified for small samples, and ordinary least squares (OLS) appear to work at least as well as probit for the cases examined here. OLS has been used in past research, but its underlying assumption in regression significance tests was violated when a dichotomous dependent variable was used. Probit and logit demonstrate theoretical appeal, with probit used more often than logit in recent classificatory studies in accounting.

User Contributions:

Comment about this article or add new information about this topic:

Comment:  (50-4000 characters)

Name:

E-mail:

Security Code: