Uskali Maki read three books by McCloskey on the "rhetoric of economics" with sympathy. But he wants McCloskey to choose between a coherence and a correspondence theory of truth. McCloskey notes in reply that modern epistemology - by contrast with the analytic philosophy circa 1955 that many philosophers of economics espouse - rejects the choice. Modern epistemology would say that economic scientists argue in many legitimate ways, governed by ethics. In brief, as Maki agrees, economics has a rhetoric. Rhetoric is a better guide than 1955-style analytic philosophy.

This paper considers an alternative asymptotic framework to standard sequential asymptotics for nonlinear models with deterministically trending variables. The asymptotic distributions of generalized method of moments estimators and corresponding test statistics are derived using this framework. The asymptotic distributions are shown to be the same with deterministically trending variables as with non-trending variables. That is, the distributions are normal and chi-squared respectively. The asymptotic covariance matrices of the estimators, however, are found to depend on the form of the trends. These findings provide a justification for the use of standard asymptotic approximations in nonlinear models even when the variables have deterministic trends.

This paper develops a framework for a general equilibrium analysis of asset markets when the number of assets is infinite. Such markets have been studied in the context of asset pricing theories. Our main results concern the existence of an equilibrium. We show that an equilibrium exists if there is a price system under which no investor has an arbitrage opportunity. A similar result has been previously known to hold in finite asset markets. Our extension to infinite assets involves a concept of an arbitrage opportunity which is different from the one used in finite markets. An arbitrage opportunity in finite asset markets is a portfolio that guarantees non-negative payoff in every event, positive payoff in some event, and has zero price. For the case of infinite asset markets, we introduce a concept of sequential arbitrage opportunity which is a sequence of portfolios which increases an investor's utility indefinitely and has zero price in the limit. We show that a sequential arbitrage opportunity and an arbitrage portfolio are equivalent concepts in finite markets but not in their infinite counterpart.

This paper examines the behavior of institutional traders. We use unique data on the equity transactions of 21 institutions of differing investment styles which provide a detailed account of the anatomy of the trading process. The data include information on the number of days needed to fill an order and types of order placement strategies employed. We analyze the motivations for trade, the determinants of trade duration, and the choice of order type. The analysis provides some support for the predictions made by theoretical models, but suggests that these models fail to capture important dimensions of trading behavior.

Résumé. Les travaux antérieurs qui ont porté sur la formulation d'hypothèses démontrent que les vérificateurs ont tendance, dans la formulation de leurs hypothèses initiales, à attribuer les fluctuations imprévues aux erreurs qui se produisent fréquemment dans les états financiers. L'on ne s'interroge cependant pas, dans ces travaux, sur la nature de l'incidence des hypothèses initiales sur la performance subséquente dans le diagnostic de l'erreur véritable. Les auteurs avancent que l'hypothèse formulée au départ et la capacité des vérificateurs de passer par la suite à une hypothèse différente interfèrent. Si, par exemple, l'hypothèse initiale était inexacte, il serait difficile pour les vérificateurs de changer d'hypothèse dans le diagnostic de l'erreur véritable. Plus encore, le fait d'invoquer initialement une erreur fréquente exacerberait cette difficulté. L'on a demandé aux vérificateurs‐sujets de produire une hypothèse initiale relative à l'erreur après avoir pris connaissance d'un modèle de fluctuations dans lequel les ventes et les comptes clients étaient surévalués. Une fois formulée leur hypothèse initiale, la moitié des sujets recevaient de l'information supplémentaire révélant l'existence d'une erreur très fréquente (dans la démarcation des ventes) et l'autre moitié recevaient de l'information révélant l'existence d'une erreur peu fréquente (celle de la double comptabilisation d'une même vente). Conformément à leurs prévisions, les auteurs constatent que le fait, pour les vérificateurs, d'invoquer initialement l'erreur très fréquente (soit la démarcation des ventes) par rapport à une autre erreur moins fréquente avait par la suite une incidence sur leur performance dans le diagnostic des erreurs véritables. Plus précisément, les vérificateurs qui invoquaient l'erreur très fréquente dans leur hypothèse initiale obtenaient les meilleurs résultats lorsque cette erreur était l'erreur véritable, mais ils obtenaient les résultats les moins bons lorsque l'erreur peu fréquente se trouvait être l'erreur véritable. Les vérificateurs qui invoquaient une erreur moins fréquente dans leur hypothèse initiale affichaient, pour leur part, une performance relativement bonne (c'est‐à‐dire se situant entre la performance la meilleure et la performance la moins bonne), aussi bien lorsque l'erreur véritable était fréquente que lorsqu'elle était peu fréquente. Les auteurs analysent les conséquences de ces résultats sur l'efficience et l'efficacité de la vérification.

Abstract. Previous work on hypothesis generation demonstrates that auditors tend to generate frequently occurring financial statement errors as their initial hypotheses to explain unexpected fluctuations. However, such work does not examine how the initially generated hypothesis affects subsequent performance at identifying an actual error. We hypothesized that the initially generated hypothesis would interfere with an auditor's ability to subsequently switch to a different hypothesis. Thus, if the initial hypothesis were incorrect, auditors would find it difficult to switch hypotheses in order to identify an actual error. Moreover, initially generating a frequent error would exacerbate this difficulty. Auditor‐subjects were asked to generate an initial error hypothesis after seeing a pattern of fluctuations in which sales and accounts receivable were overstated. After they generated their initial hypothesis, half of the subjects were provided with additional information that was consistent with a very frequent error (sales cutoff) and the other half were provided with information consistent with an infrequent error (sales recorded twice). As expected, we found that initially generating the very frequent error (i.e., sales cutoff) versus some other less frequent error affected auditors' subsequent performance at identifying actual errors. Specifically, auditors who generated the very frequent error as their initial hypothesis performed best when it was the actual error, but performed worst when the infrequent error was the actual error. In contrast, auditors who generated a less frequent error as their initial hypothesis performed moderately well (i.e., between best and worst) both when the actual error was frequent and when it was infrequent. The implications of these results for audit efficiency and effectiveness are discussed.