First difference maximum likelihood and dynamic panel estimation

First difference maximum likelihood (FDML) seems an attractive estimation methodology in dynamic panel data modeling because differencing eliminates fixed effects and, in the case of a unit root, differencing transforms the data to stationarity, thereby addressing both incidental parameter problems and the possible effects of nonstationarity. This paper draws attention to certain pathologies that arise in the use of FDML that have gone unnoticed in the literature and that affect both finite sample performance and asymptotics. FDML uses the Gaussian likelihood function for first differenced data and parameter estimation is based on the whole domain over which the log-likelihood is defined. However, extending the domain of the likelihood beyond the stationary region has certain consequences that have a major effect on finite sample and asymptotic performance. First, the extended likelihood is not the true likelihood even in the Gaussian case and it has a finite upper bound of definition. Second, it is often bimodal, and one of its peaks can be so peculiar that numerical maximization of the extended likelihood frequently fails to locate the global maximum. As a result of these pathologies, the FDML estimator is a restricted estimator, numerical implementation is not straightforward and asymptotics are hard to derive in cases where the peculiarity occurs with non-negligible probabilities. The peculiarities in the likelihood are found to be particularly marked in time series with a unit root. In this case, the asymptotic distribution of the FDMLE has bounded support and its density is infinite at the upper bound when the time series sample size T→∞$T\to \infty$. As the panel width n→∞$n\to \infty$ the pathology is removed and the limit theory is normal. This result applies even for T$T$ fixed and we present an expression for the asymptotic distribution which does not depend on the time dimension. We also show how this limit theory depends on the form of the extended likelihood.     

Female Earnings and Divorce Rates: Some Australian Evidence

The purpose of this article is to examine whether female earnings have influenced divorce rates in Australia, using state‐level data for the past four decades. Following a recent study by Ressler and Waters (2000), which concludes from comparable US data that female earnings and divorce rates may be jointly endogenous, initial testing is performed to identify whether female earnings can be treated as exogenous. A Hausman specification error test finds no evidence of a simultaneous relationship in the Australian data, in contrast to the findings of Ressler and Waters. The test result supports the hypothesis that other underlying factors affect female earnings, of which higher divorce rates are merely another symptom. A divorce rate equation is estimated. In accordance with much of the literature, the rise in female earnings over the past four decades is found to have increased Australian divorce rates.     

Economics Education for all Schoolchildren

In the first of a series by teachers on the role of economics in school curricula and beyond, Fred Hankins, Secretary of the Economics Association, identifies the essentials of economics analysis — not least opportunity cost - desirable to equip pupils for a career.     

A heuristic programming approach to the employee scheduling problem and some thoughts on “managerial robots”

We describe a system for the automatic scheduling of employees in the particular setting in which: the number of employees wanted on duty throughout the week fluctuates; the availabilities of the employees varies and changes from week to week; and a new schedule must be produced each week, by virtue of the changing demand for service.The problem which we address appears in a variety of settings, including: airline reservation offices; telephone offices; supermarkets; fast food restaurants; banks and hotels.Previous approaches to the problem have relied chiefly on formal methods, generally involving one or another variation of linear or integer, mathematical programming. We suggest that except in cases involving very small problems (only a handful of employees) that those approaches have not proven promising, especially where union rules and management requirements impose complex constraints on the problem, and that a heuristic approach has proven to be substantially superior.We set forth the general features of our heuristic approach, which we see as an application of artificial intelligence; we show how, in contrast to other approaches, which design shifts as if employees were always available and try to fit those shifts to employees who are not always available, our system design shifts with deference to the employees' limited availabilities; we suggest that, for a given service level, our system produces schedules with a better “fit”—number of employees actually on duty comparing more favorably with the number wanted; and we state that while, for a given service level, a ‘manual scheduler’ may take up to 8 hours each week to prepare a good schedule, our system, on most micro computers, routinely produces better schedules involving up to 100 employees in about 20 minutes.The scheduling of employees is generally considered to be a managerial function, in the setting of the problem we address. When a craft employee is replaced on an assembly line by a machine which performs the same function, we speak of the replacing mechanism as an industrial robot.We suggest that systems like that which we describe deserve a name, to distinguish them from comparable, computer based systems which do not replace, but rather supplement a manager, and we suggest the name ‘managerial robot’ for such systems.We set forth the characteristics which we feel would justify applying the term ‘managerial robot’ to a computer based system, and suggest that classification is basic to understanding and communication and that just as terms such as decision support systems and expert systems prove useful in our increasingly advanced, technological society, so also the term managerial robot has a place in our scheme of things.Decision support systems do not qualify as managerial robots for the reason that managerial robots don't simply support the decision making process, but rather replace the manager in his performance of a function which, when performed by a human being, is considered a managerial function.Nor do we consider managerial robots to qualify as expert systems. While our scheduling system contains an inference mechanism, and could be enhanced to improve the quality of its schedules thru ‘experience’ (and thus to ‘learn’?), that—lacking a knowledge base in the sense of expert systems-and most of all in replacing rather than supporting the decision maker, the managerial robot needs a term of its own.We elaborate, in this paper, a specific application of our system, and show how the design of shifts, and the placement of breaks, serve to yield a fit whose quality no human scheduler can duplicate.     

New technology introduction in a marginal coffee area of Colombia

Income stagnation in the downward cycle of primary product prices is a problem for small farmers in developing countries. The recent abrupt price decline in coffee poses a large adjustment problem in marginal production areas. One method of farm income stabilization for marginal export crop production areas is diversification into food crops with improved technology. The potential for a cushioning of this farm income decline via the introduction of improved food crop technology was examined. A substantial moderating effect was obtained even without uprooting the improved coffee technology. Moreover, an analysis of the constraints to the earlier introduction of the Caturra coffee variety helped identify some of the components of farmer decision making relevant to adoption of new technology.