Knowledge that Transforms

Jerry G. Thursby, A Test Strategy for Discriminating Between Autocorrelation and Misspecification in Regression Analysis, The Review of Economics and Statistics, Vol. 63, No. 1 (Feb., 1981), pp. 117-123

ECONOMISTS are just beginning to develop empirical knowledge about the basic relationships that determine the distribution of incomes and allocation of resources in the poor rural regions of the world. Understanding interactions between agriculture and the rest of the economy is particularly important in this context. This paper empirically investigates the allocation of human resources between farm work and other jobs and the compensation received by individuals for their work in agriculture and elsewhere. Specifically, two related tasks are undertaken in this paper: (1) the estimation of wage functions; and (2) the estimation of functions determining the allocation of time among occupations. There have been a series of papers which estimate the earnings function using data from poor countries. (For examples, see Blaug (1974); Chiswick (1977); and Valdes (1971).) However, this study focuses not on total earnings but on the wage rates a person may earn in labor market work, because the wage rate is the relevant price of time for many allocative decisions. For example, in dividing the work year between self-employed farming and work as an employee off the farm, a key price is the potential wage rate the person can earn in the labor market. (For discussion and evidence on this issue, see Sumner (1978).) Recently, a few studies have appeared which illustrate the importance of understanding potential wage variations across individuals in less developed countries. Rosenzweig (1978) uses wage estimates from aggregate Indian data in a model determining the impact of land reform. Bardhan (1979) develops wage functions from a sample survey in rural India as a step in estimating labor supply for rural households. Kusnic and DaVanzo (1980) use wage functions to impute potential incomes and to measure the full income distribution. These previous studies illustrate the importance of understanding the determinants of market wages. The present paper develops wage functions in more detail, focusing especially on industrial differentials. In many rural areas employment opportunities may be usefully categorized into (1) self-employment (mainly on small farms); (2) work for wages on large farms; and (3) work for wages in non-farm industries. The combination of selfemployment and work as an employee is also important in many rural areas in both rich and poor countries. This paper examines the impact of personal characteristics on the choice to participate in work for wages and on the choice of industry. These estimates also form a basis for studying industrial wage differentials. Tentative estimates of the dependence of participation on expected wages are also presented below. The empirical analysis in this paper is based on a 1974 sample of five villages in central Guatemala (see Corona (1978) and Stein (1978)). One village, Petapa, is larger, richer and more urban than the others. It lies south of Guatemala City within commuting distance of the capital. The other villages are outside the capital city region. The population sampled is almost exclusively ladino (Spanish culture as opposed to Indian). Data are available on the earnings, hours of work, industry, age, schooling, literacy, and village of 1,005 male heads of households. Approximately one-third of the sample work as both self-employed farmers and hired workers, and most of these men work on fincas (large plantations). Another quarter of the sample work as self-employed farmers and do no work for wages. The suburban village has mostly nonagricultural workers, whereas the four other villages have predominantly farmers and farm employees. In Received for publication June 21, 1979. Revision accepted for publication December 31, 1980. * North Carolina State University. Research on this paper began while I was a Rockefeller Foundation supported Post-Doctoral Fellow in the Labor and Population Group of the Economics Department of the Rand Corporation and continued at North Carolina State University supported by the Agricultural Research Service. At Rand, Robert Newman served as my research assistant and at N.C. State, James Cochell provided able assistance. I have benefited from the comments of colleagues at both institutions and from members of the Labor Economics Workshop at The University of North Carolina at Chapel Hill. I also thank two anonymous referees for useful suggestions. This is paper number 6558 of the Journal Series of the North Carolina Agricultural Research Service, Raleigh, N.C.

This paper contributes new evidence on the wealth distribution in the rural North on the eve of the Civil War,' evidence which shows a greater degree of egalitarianism among a large segment of the population than have other studies. We argue that the wealth distribution in this particular historical economy may be used as a yardstick against which contemporary distributions might be judged. We also present evidence on the demographic and social characteristics of various groups of wealthholders and posit a causal chain from these characteristics to the concentration of wealth. To the extent that contemporaries asserted that the rural, free northern states were an egalitarian society, a study of the nature, extent and sources of economic inequality in that environment might demonstrate what is realizable as a policy objective in contrast to the Lorenzian goals or even the new standard recently proposed by Morton Paglin (1975 and 1977). Traditionally, historical analyses of egalitarianism in the nineteenth-century United States have focused on the distribution of wealth in two areas where it was least likely to be manifest, the slave-South and the urban centers.2 Our focus, however, is upon that segment of the population among which wealth, historically, was felt to have been most equally distributed: the nineteenth century rural North, an area that has received little attention despite its importance. Even where the rural North has been studied, as by Main (1976) or Newell (1977), the focus has been on narrowly defined geographic areas. The figures reported in table 1 present measures of the distribution of wealth for that group.

It is the purpose of this paper to describe one method of extending the empirical applicability of the theory of the farm-household to multi-crop economies by integrating the econometric and linear programming models already available in the literature. Section I introduces the subject. In Section II the theoretical model is presented. In Section III Korean data are used to assess the quantitative significance of the approach by calculating and comparing household response to changes in input and output prices for two different specifications of the models. First, the household is analyzed from the consumption side on the assumption that farm profits are exogenous; this corresponds to the standard approach in consumer demand theory. And second, results are presented for a model in which consumption and production responses are integrated in a theoretically consistent fashion, and farm profits are allowed to reflect production responses to price changes. The results from this and other studies that use an integrated approach to production and consumption decisions highlight the need to change perceptions concerning agricultural household response to economic incentives in developing countries and to revise the design of economic projects and policies accordingly.

THE inconsistency the labor supply estimates using nonexperimental data has led to the costly income maintenance experiments that have taken place in various parts the country. In addition to being costly, the analyses experimental data also have their share econometric problems. 1 Both experimental and nonexperimental studies face the problems sample selection, model specification and variable measurement. In order to determine what effects the variables of interest have on labor supply, all relevant factors must be included. One these variables, namely health, is often included in a form which results in biased estimates. In this paper, we hope to show that the analysis experimental and nonexperimental labor supply data must carefully address the health variable problem before meaningful estimates can be obtained. In general, three treatments emerge in the literature: no control for health is used, a binary variable which we shall call a variable is included in the model as an exogenous variable, or the disability variable is used as a partitioning criterion to obtain a sample ,able-bodied men or women. The disability variable is often the response to the question, Does your health prevent you from working altogether? or Does your health limit the kind or amount work you could do? This study will examine the bias created by each these approaches and will suggest a method measurement for health which does not produce biased estimates.2 Health and Labor Supply

A CCORDING to the standard job search model,I a job searcher knows the distribution function of wage offers available to him in the labor market and the fixed search cost of an offer per period. Given this knowledge, the searcher draws an offer and then weighs it against the possibility of better future offers. If the expected gain from searching for the next (future) offer exceeds the search cost, the current offer is rejected; and the individual remains unemployed. The entire search process, i.e., the duration of unemployment, ends when the searcher obtains an acceptable offer. Thus, the standard search model focuses on the unemployment duration as a result of the searcher's rejection of job offers. Consequently, lower search cost is regarded as a major deterrent to shortening unemployment duration since it encourages the searcher to reject offers. Another important factor affecting unemployment duration, which has not been given adequate attention in the standard model, is the time elapsed in searching (search time) for an offer. Most determinants of the search time for an offer, e.g., labor market conditions and socio-economic characteristics of the searcher, are beyond his control over the duration of unemployment. However, the search time for an offer is not entirely exogenous to the searcher. It can be shortened as he intensifies search effort. Raising search intensity, however, is costly; and it increases the search cost for an offer. This contrasts with the fixed search cost assumption built into the standard model. Thus, the searcher should choose an optimal level of search intensity to maximize the net return from job search. This paper extends the standard search model in order to investigate the role of search time for an offer in the determination of unemployment duration. An econometric model of unemployment duration with variable search intensity is formulated in section II. The model distinguishes between two components of unemployment duration which are related to each other: (i) the search time for an offer and (ii) the unemployment duration due to rejection of offers. It also decomposes the search cost of an offer into a fixed cost and a variable cost, which depends on the level of search intensity. Under the assumption that the elasticity of the mean search time for an offer with respect to search intensity is greater than 1 but less than 0, several results are shown: First, higher fixed search cost per unit time lowers the mean duration of unemployment as usual. Second, the unit cost of search intensity is positively related to unemployment duration, contrary to the implication of the standard model. That is, as the unit cost of search intensity rises, the search cost of an offer increases; but the duration of unemployment does not decrease. This is because the resulting decrease in search intensity increases the first component of unemployment duration (search time for an offer), which more than compensates for the decrease in the second component due to the higher search cost of an offer. Section III presents empirical investigations on the duration of unemployment, which are shown to be consistent with the theoretical contentions of section II. Finally, conclusions are summarized in section IV.

We estimate substitution possibilities among a set of age-race-sex groups in the labor force. The estimates are based on cross-section data from SMSAs in 1969, and they allow us to consider how substitutable adult women are for young women or young men. The estimates are used, along with assumptions about the extent of wage rigidity and elasticities of labor supply, to simulate the direct and indirect effects of the growth of the female labor force on job opportunities for youth, assuming rigid wages for young workers, and on the wage rates of adult males, assuming these wages are flexible.

IN March of 1972, the Canadian federal government and the Ontario provincial government jointly announced that a second international airport for Toronto would be built near the town of Pickering, Ontario. Considerable controversy followed this decision in spite of the great number of governmental studies on which the decision had presumably been based. Among other criticisms it was alleged that while the government's research effort had been extensive, it had not necessarily been balanced, and the case for simply expanding the existing international airport at Toronto might have been understated. The problem faced by the Canadian government in this instance illustrates a frequent problem in the evaluation of public policy alternatives: how to provide balance in the development of information so that both the pro and con positions can be said to be adequately represented. Airport location studies provide particularly good illustrations of the problem of balance because it is impossible for the data and arguments ever to be considered complete. No matter what the expenditure of time and money, there is always room for further research. Even after the million-pound Rosskill cost-benefit study regarding the third London airport, Mishan (1970) and Flowerdew (1972) describe how differences of opinion continued to persist among economists on the commission itself. Similarly, for the Pickering Airport question, no amount of study could be expected to be definitive. In such cases the problem of how to assure that resources and talent are not overconcentrated on one side or the other is a problem of importance. The response of the Canadian government to the Pickering controversy was to appoint a three-man review board, headed by a federal judge, to serve as an Airport Inquiry Commission. The commission was assigned full responsibility for reviewing existing airport studies, holding new hearings, and presenting new summaries and recommendations regarding whether or not the new airport should be built. The procedure chosen in this instance illustrates a common approach to organizing information for decisions. A group such as an executive or legislative committee wishes to receive both balanced information and recommendations regarding some proposal, but allows the same people who are asked to make the recommendations to also summarize the information. When the group making the recommendations also summarizes the information, however, the balance in this information is subject to question. The information may be seen in part as a rationale for the recommendations, and both the information and recommendations may be considered too dependent upon the personal views of a small group. Consequently there have been suggestions for more formal information procedures for certain executive and legislative decisions. More formal information procedures, for example, have been proposed for government technology assessment. As one example, Uman (1975) outlines procedures for using science advisory panels as checks on internal government technology assessments. Other technology assessment procedures are surveyed in articles by Jones (1973) and Coates (1974). An even more formal approach to securing balance in information is the Science Court proposed by Kantrowitz and discussed in reports such as that by the Task Force of the Presidential Advisory Group in Anticipated Advances in Science and Technology (1976). The Science Court and many of the other formal procedures for balance, howReceived for publication March 20, 1978. Revision accepted for publication August 8, 1980. * York University. This study was supported by the University of TorontoYork University Joint Program in Transportation and Canada Council Grant No. 770121. The author would like to thank R. F. Boruch, H. 0. Hartley, B. L. Raktoe, and also the participants at the 1978 National Bureau of Economic Research-National Science Foundation Conference on Decision Rules and Uncertainty at Carnegie-Mellon University. The author is particularly indebted to the economists who formed the research teams for the experiment, John Evans with advisor George Hilton, and Sanford Borins with advisor Glen Jenkins. Finally the author would like to acknowledge useful suggestions from the referees.