MSE-improvement of the least squares estimator by dropping variables

It is well known that dropping variables in regression analysis decreases the variance of the least squares (LS) estimator of the remaining parameters. However, after elimination estimates of these parameters are biased, if the full model is correct. In his recent paper, Boscher (1991) showed that the LS-estimator in the special case of a mean shift model (cf. Cook and Weisberg, 1982) which assumes no “outliers” can be considered in the framework of a linear regression model where some variables are deleted. He derived conditions under which this estimator outperforms the LS-estimator of the full model in terms of the mean squared error (MSE)-matrix criterion. We demonstrate that this approach can be extended to the general set-up of dropping variables. Necessary and sufficient conditions for the MSE-matrix superiority of the LS-estimator in the reduced model over that in the full model are derived. We also provide a uniformly most powerful F-statistic for testing the MSE-improvement.     

Einzelbesprechungen

Ohne ZusammenfassungAus dem Italienischen übersetzt von M. St. Braun, WienAus dem Russischen übersetzt von A. GerschenkronAus dem Italienischen übersetzt von Erich Allina     

Severance Pay and Corporate Finance: Empirical Evidence from a Panel of Austrian and Italian Firms

This paper examines the empirical link between severance pay and corporate finance. Severance pay is an economic debt of the employer and hence should be taken into account by the market in its assessments of risk. Using a hand-collected dataset of accounting data from Italy and Austria we find there is only a limited relationship between severance pay and market risk indicators. This suggests that arguments that severance pay systems destroy corporate value may need to be reassessed.     

Self-selection and variations in the laboratory measurement of other-regarding preferences across subject pools: evidence from one college student and two adult samples

We measure the other-regarding behavior in samples from three related populations in the upper Midwest of the United States: college students, non-student adults from the community surrounding the college, and adult trainee truckers in a residential training program. The use of typical experimental economics recruitment procedures made the first two groups substantially self-selected. Because the context reduced the opportunity cost of participating dramatically, 91 % of the adult trainees solicited participated, leaving little scope for self-selection in this sample. We find no differences in the elicited other-regarding preferences between the self-selected adults and the adult trainees, suggesting that selection is unlikely to bias inferences about the prevalence of other-regarding preferences among non-student adult subjects. Our data also reject the more specific hypothesis that approval-seeking subjects are the ones most likely to select into experiments. Finally, we observe a large difference between self-selected college students and self-selected adults: the students appear considerably less pro-social.     

Competitive signalling

With imperfect information about product quality there are incentives for buyers to make use of proxy variables as “signals”, and hence for sellers to invest in the activity of signalling. Received theory suggests that there are plausible circumstances under which there exist a whole family of potential “signalling equilibria” each of which successfully distinguish quality differences.In this paper it is shown that from the family of “equilibria” only one, the Pareto-dominant member, survives plausible experimentation by buyers. With moderately more sophisticated experimentation it is further shown that there is no competitive equilibrium.     

Estimating the covariance matrix for regression models with ar(1) errors and lagged dependent variables

We propose a simple procedure, based on an artificial linear regression, for consistently estimating the covariance matrix of the parameter estimates for linear regression models with serially correlated errors and lagged dependent variables.     

On the human aspects in structural modeling

This article deals with two facets of structural modeling often ignored: a) the several human roles that effective participative modeling must encompass, and b) the group procedures developed to assist modelers in defining the elements of the system to be modeled. The roles are 1) the method technician, 2) the facilitator, and 3) the participant. Their recognition on the part of the technology assessment leader is vital to the successful conduct of participative modeling. The element-generating group procedures surveyed here are organized into two categories: those which emphasize an atmosphere for free-wheeling thinking, and those which emphasize structured guidance (either through use of words or geometric and analytic techniques).