Voters penalize federal and state spending growth. This is the central result of my analysis of voting behavior in Presidential, Senatorial, and gubernatorial elections from 1950–1988. The composition of federal spending growth seems irrelevant. The vote loss to the President's party from an extra dollar of defense or nondefense spending is the same. However, in gubernatorial elections, expansion of state welfare spending exacts a disproportionate political price. Deficit financing of federal or state spending does not appear to matter politically. I conclude by discussing the obvious question of why government budgets have grown in the face of this voter hostility.

Gerald R. Jensen, Donald P. Solberg, Thomas S. Zorn, Simultaneous Determination of Insider Ownership, Debt, and Dividend Policies, The Journal of Financial and Quantitative Analysis, Vol. 27, No. 2 (Jun., 1992), pp. 247-263

The paper analyzes the contemporaneous association between market returns and earnings for long return intervals. The research design exploits two fundamental accounting attributes: (i) earnings aggregate over periods, and (ii) expanding the interval over which earnings are determined, is likely to reduce ‘measurement errors’ in (aggregate) earnings. These concepts lead to the level of (aggregate) earnings as a natural earnings variable for explaining security returns. We hypothesize that the longer the interval over which earnings are aggregated, the higher the cross-sectional correlation between earnings and returns. The empirical findings support this hypothesis.

This paper compares risk-taking of insured.and uninsured thrifts operating under strict and less-strict regulatory regimes during the 1930's. Analysis of balance-sheet data indicates that while newly insured thrifts undertook less risk than their uninsured counterparts, possibly because of screening by deposit-insurance authorities, moral hazard emerged gradually. Insured institutions operating under relatively permissive regulatory regimes were more prone to undertake risky lending activities than their more tightly regulated counterparts, possibly because of screening by deposit-insurance authorities, moral hazard emerged gradually. Insured institutions operating under relatively premissive regulatory regimes were more prone to undertake risky lending activities than their more tightly regulated counterparts. Given the current system of deposit insurance, the results suggest that effective regulation and supervision will play a key role in maintaining thrift stability in the 1990s. Copyright 1992 by American Economic Association.

This paper compares risk-taking of insured and uninsured thrifts operating under strict and less-strict regulatory regimes during the 1930's. Analysis of balance-sheet data indicates that while newly insured thrifts undertook less risk than their uninsured counterparts, possibly because of screening by deposit-insurance authorities, moral hazard emerged gradually. Insured institutions operating under relatively permissive regulatory regimes were more prone to undertake risky lending activities than their more tightly regulated counterparts. Given the current system of deposit insurance, the results suggest that effective regulation and supervision will play a key role in maintaining thrift stability in the 1990's.

[In the last two decades of accounting research, many studies have investigated the relation between accounting variables and risk-adjusted security returns. The earliest studies (e.g., Ball and Brown 1968) used simple, nonparametric methods and focused mainly on the question of whether accounting earnings are associated with residual equity returns. Subsequent studies made methodological refinements in both the measurement and the statistical techniques. One important statistical refinement was the "unexpected earnings response regression model" (UERRM), a linear statistical model that uses the unexpected earnings variable as a regressor to explain risk-adjusted returns. The UERRM has become well-known, and many recent studies (e.g., Cornell and Landsman 1989, 686; Daley et al. 1988, 580; Doran et al. 1988, 392; Landsman and Damodaran 1989, 107; McNichols 1989, 15) have used some form of it as a "benchmark" model, against which to compare more complicated models. In one paradigm (so popular that it has become almost standard practice), various accounting variables are added to the UERRM, and their "incremental information content" is assessed by testing the statistical significance of their coefficients. The inferences derived from this procedure are, of course, conditional on the degree to which the UERRM is correctly specified. A critical problem caused by using a misspecified UERRM is that its least squares estimator can lead to erroneous inferences in the research design. Despite the UERRM's popularity, researchers have expressed concerns regarding its specification. For example, Lev (1989) recently surveyed a large number of research papers that used some form of an UERRM and found that for the most part R2 s were low and often bordered on "the negligible." His table 1, which includes statistics from 19 studies, indicates that most reported R2 s are less than 10 percent. Although low R2 s are not proof of major specification problems, Lev does suggest that they are a cause for concern and may be the result of specification problems. Investigation of multiple specification problems is an important aspect of the present study because the various specification issues are interrelated. For example, nonnormality can be associated with nonlinearity, which, in turn, can be associated with heteroscedasticity or variation in the coefficients (Judge et al. 1985, 455, 814, 839). Thus, ad hoc tests for a single specification problem can be misleading and can fail to identify the fundamental problems. To our knowledge, no studies have comprehensively and formally evaluated the specification of the cross-sectional, ordinary least squares model that relates unexpected earnings to risk-adjusted security returns. Therefore, the purpose of this study is to test such a model systematically and empirically for specification problems. Specifically, this study tests for nonlinearity, heteroscedasticity, residual nonnormality, omitted variables, and interfirm systematic and random coefficient variation. Also, when appropriate, adjusted R2 -statistics are included to indicate the degree of misspecification-information that is not directly observable from the tests themselves. A high degree of generality is obtained by using three samples of earnings forecasts as proxies for expected earnings. These were obtained from IBES financial analyst consensus forecasts, Value Line financial analyst forecasts, and COMPUSTAT-based time-series forecasts. Daily and monthly security returns are considered for short and long event-windows, respectively. In addition, one study recently published in The Accounting Review (Cornell and Landsman 1989) is replicated. The findings from all of the samples and the replication indicate that the specification error is large enough to affect conclusions regarding economic relationships. For example, in the replication, the specification problems are shown to lead to substantial instability in inferences from the model.]

This paper studies the class of denumerable-armed (i.e. finite- or countably infinitearmed) bandit problems with independent arms and geometric discounting over an infinite horizon, in which each arm generates rewards according to one of a finite number of distributions, or "types." The number of types in the support of an arm, as also the types themselves, are allowed to vary across the arms. We derive certain continuity and curvature properties of the dynamic allocation (or Gittins) index of Gittins and Jones (1974), and provide necessary and sufficient conditions under which the Gittins-Jones result identifying all optimal strategies for finite-armed bandits may be extended to infinite-armed bandits. We then establish our central result: at each point in time, the arm selected by an optimal strategy will, with strictly positive probability, remain an optimal selection forever. More specifically, for every such arm, there exists (at least) one type of that arm such that, when conditioned on that type being the arm's "true" type, the arm will survive forever and continuously with nonzero probability. When the reward distributions of an arm satisfy the monotone likelihood ratio property (MLRP), the survival prospects of an arm improve when conditioned on types generating higher expected rewards; however, we show how this need not be the case in the absence of MLRP. Implications of these results are derived for the theories of job search and matching, as well as other applications of the bandit paradigm.

*University of Southern California; tUniversity of Colorado at Boulder. We would like to thank Gilbert Bloom of KPMG Peat Marwick, Bob Rosen of Ernst & Young, Wayne Gazur, Robert Jamison, Sally Jones, Stewart Karlinsky, and David Mason for their assistance in validating the instruments; Eugene Willis and the AICPA for allowing us to collect data at the National Tax Education Program; Stephen Conrad of Arthur Andersen, John Lanning of KPMG Peat Marwick, Jerry Marrs of Ernst & Young, and Randy Stein of Coopers & Lybrand for allowing us to collect data at their respective firms; Minou Bohlin, Linda Levy, David Mason, and Paul Walker for their research assistance; and Vairum Arunachalam for his assistance in collecting data. The authors also gratefully acknowledge the helpful comments of three anonymous referees, Alison Ashton, Robert Ashton, C. Brian Cloyd, David Frederick, Joan Luft, Robert Libby, Laureen Maines, Mark Nelson, Michael Roberts, Frank Selto, D. Shores, Ira Solomon, Rick Tubbs, S. Mark Young, and workshop participants at Arizona State University, Cornell University, Duke University, Indiana University, the University of Illinois Tax Symposium, the Journal of Accounting Research Conference, University of Texas at Arlington, University of Utah, and University of Wisconsin. Finally, the financial support of the KPMG Peat Marwick Foundation and the University of Colorado is gratefully acknowledged. 1 We infer expertise in this study from the level of performance in a specific task, here issue identification in tax planning. This inference is consistent with much of the literature on expertise in accounting and other disciplines (e.g., Bonner and Lewis [1990],