Organization of Innovation in a Multi-Unit Firm: Coordinating Adaptive Search on Multiple Rugged Landscapes

In Chang and Harrington (2000a), a computational model of a multi-unit firm is developed in which unit managers continually search for better practices. Search takes place over a rugged landscape defined over the space of unit practices. There it is shown that a more centralized organization is optimal when markets are not too different and the horizon is not too long. The robustness of those results are explored here with respect to the shape of the landscape. In particular, we find that centralization does better when the search space is larger and there is a stronger correlation in a consumer's preferences across different dimensions. A richer description of comparative dynamics is also provided.

On the Non-Existence of Reputation Effects in Two-Person Infinitely-Repeated Games

Consider a two-person infinitely-repeated game in which one player is either a normal "rational" type or a "commitment" type that authomatically plays a fixed repeated-game strategy. When her true type is private information, a rational type may want to develop a reputation as a commitment type by mimicking the commitment type's actions. But, the uninformed player, anticipating the behavior of the rational type, may try to "screen out" the rational type by choosing an action which gives the rational type a low payoff when she mimics the commitment type. My main result shows that for "comparably" patient players, if the prior probability that the player is a commitment type is sufficiently small, the "screening" process may take so long that the rational player does not benefit from developing a reputation. In the case of equally patient players, I show that the folk theorem holds even when both players possess a small amount of private information. Schmidt (1994) and Cripps, Schmidt and Thomas (1993) argue that reputation effects can rule out outcomes permitted by the folk theorem, regardless of how small the prior probability that the player is a commitment type. My results show that this argument only applies when one player is "infinitely" more patient than the other.

Does Banning Affirmative Action Harm College Student Quality?

Banning affirmative action from college admissions decisions cannot prevent an admissions office that cares about diversity from achieving it through channels other than the explicit consideration of race. We construct a model of college admissions where candidates from two groups with different average qualifications compete for a fixed number of seats. When an admissions office that cares both about the quality and diversity of its entering class can use group identity as a criterion for admissions, its preferred admissions rule selects the best-qualified candidates from each group. When it cannot use affirmative action, the admissions office's preferred rule generally does not select the best-qualified candidates from either group: it randomizes over candidates to achieve diversity, at the expense of within-group selection. A ban always reduces diversity, and may also lower average quality. Moreover, even when a total ban on affirmative action raises average quality, a partial ban may raise average quality even more.

The Diffusion of Innovations in Social Networks

We consider processes in which new technologies and forms of behavior are transmitted through social and geographic networks. Agents adopt behaviors based on a combination of their inherent payoff and their local popularity (the number of neighbors who have adopted them) subject to some random error. We characterize the long-run dynamics of such processes in terms of the geometry of the network, but without placing a priori restrictions on the network structure. When agents interact in sufficiently small, close-knit groups, the expected waiting time until almost everyone is playing the stochastically stable equilibrium is bounded above independently of the number of agents and independently of the initial state.

Rates of Information Aggregation in Common Value Auctions

Near-Rationality and Inflation in Two Monetary Regimes

Sticky-price models with rational expectations fail to capture the inertia in U.S. inflation. Models with backward-looking expectations capture current inflation behavior, but are unlikely to fit other monetary regimes. This paper seeks to overcome these problems with a near-rational model of expectations. In the model, agents make univariate forecasts of inflation: they use information on past inflation optimally, but they ignore other variables. The paper tests sticky-price models with near-rational expectations for two periods in U.S. history, the post-1960 period of persistent inflation and the period from 1879 to 1914, when inflation was not persistent. The models fit the data for both periods; in contrast, both rational-expectations and backward-looking models fail for at least one period.

Welfare Benefits and Female Headship in U.S. Time Series

There has been a considerable amount of work on the relationship between AFDC benefits and family structure in the U.S. The evidence to date which uses cross-state variation in welfare benefits and family structure, often with state fixed effects, indicates that there is some nonzero effect of those benefits on marriage and fertility, although there is disagreement on the magnitude of the effect. However, it is undisputed that time series trends in family structure are not correlated in the direction that the cross-state evidence would suggest, for real benefits have been falling, even relative to wages, in aggregate time series. This paper reexamines the time series evidence with particular attention to the role of wages in explaining trends in headship, and notes that the correct specification includes both male as well as female wages. When both are controlled, welfare benefits have a slight positive impact on female headship even in time series. The results demonstrate the importance of labor market factors in explaining trends in female headship.

International Comovement Revisited: The Role of Non-Traded Goods and Price Stickiness

No abstract available.

Policy Interventions, Low-Level Equilibria and Social Interactions

Interest in social interactions, neighborhood effects, and social dynamics in the last several years has seen a revival. Unfortunately, little progress has been made on empirical estimation of such interactions and testing for their presence, on the development of policy interventions which work through social interactions, or on the evaluation of such interventions because several basic identification and estimation problems have not been seriously confronted. Nevertheless, most of these problems are in principle solvable and methods for identifying social interactions and estimating their magnitudes are available and are outlined in this paper. These methods address simultaneity, correlated unobservables, errors-in-variables, and endogenous group membership problems. Moreover, while policy interventions with presumed effects on social interactions have not been well-designed thus far, at least to measure social interactions per se, this problem is not inherent and several policy interventions are suggested which could work primarily through social interactions and whose evaluation could establish their magnitudes.

Measuring the Equilibrium Effects of Unemployment Benefits Dispersion

We analyze the impact of unemployment benefits and minimum wages using an equilibrium search model which allows for dispersion of benefits and productivity levels, job-to-job transitions, and structural and frictional unemployment. The estimation method uses readily available aggregate data on marginal distributions of unemployment durations as well as wages and benefit levels. Different causes of structural and frictional unemployment are investigated. We investigate the efficiency of the imposition of a single benefit level for all household types and the introduction of an Earned Income Tax Credit.

Portfolios of the Rich

Recent research has shown that 'rich' households save at much higher rates than others (see Carroll (2000); Dynan, Skinner, and Zeldes (1996); Gentry and Hubbard (1998); Huggett (1996); Quadrini (1999)). This paper documents another large difference between the rich and the rest of the population: portfolios of the rich are heavily skewed toward risky assets, particularly investments in their own privately held businesses. The paper explores three possible explanations of these facts. First, perhaps there is exogenous variation in risk tolerance, so that highly risk tolerant households engage in high-risk, high-return activities, and the risk-lovers who are lucky constitute the rich. A second possibility is that capital market imperfections a la Gentry and Hubbard (1998) and Quadrini (1999) require entrepreneurial activities to be largely self-financed, and these same imperfections imply that entreprenurial investment will yield high average returns. The final possibility is that wealth enters households' utility functions directly as a luxury good as in Carroll (2000) (one interpretation is that this reflects the utility of anticipated bequests), implying that risk aversion declines as wealth rises. The paper concludes that the overall pattern of facts suggests both Carroll-style utility and Gentry/Hubbard-Quadrini style capital market imperfections are important.

'Risky Habits' and the Marginal Propensity to Consume Out of Permanent Income, or, How Much Would a Permanent Tax Cut Boost Japanese Consumption?

Papers in a variety of disparate literatures have recently suggested that habit formation in consumption may explain several empirical puzzles, ranging from the level and cyclical variability of the equity premium (Abel (1990,1999); Constantinides (1990); Jermann (1998); Campbell and Cochrane (1999)) to the 'excess smoothness' of aggregate consumption (Fuhrer (2000)) to the apparent fact that increases in economic growth cause subsequent increases in aggregate saving rates (Carroll and Weil (1994); Bosworth (1993); Attanasio, Picci, and Scorcu (2000); Rodrik (1999); Loayza, Schmidt-Hebbel, and Serven (2000)). This paper examines an implication of these models that has mostly been overlooked: Habits strong enough to solve these puzzles imply an immediate marginal propensity to consume out of permanent shocks of much less than one. When the model is calibrated to roughly match the rise in the Japanese saving rate over the postwar period, it implies that the immediate MPC out of permanent tax cuts may be as low as 30 percent, suggesting that calls for a permanent income tax cut as a quick means of stimulating aggregate demand in Japan may be misguided.

A Theory of Quantifiable Beliefs

Building upon the works of Anscombe and Aumann (1963) and Karni and Schmeidler (1981), we develop a general axiomatic theory of quantifiable beliefs - a form of probabilistic sophistication that does not preclude state-dependent preferences and does not require the reduction of compound lotteries. The theory includes the state-dependent expected utility model of Karni and Schmeidler (1981) and the state-independent non-expected utility model of Machina and Schmeidler (1995) as special cases. The theory is flexible enough to admit recursivity in the decision-making process. One specific example of this recursive class is shown to be compatible with a quantifiable beliefs version of Schmeidler's (1989) Choquet expected utility maximizing model and thus capable of rationalizing Ellsberg-paradox type behavior.

Input and Output Inventories

This paper presents a new stage-of-fabrication inventory model with ordering, usage, and stocking of input materials that distinguishes between gross production and value added. It extends the traditional linear-quadratic model of output (finished goods) inventories by adding joint determination of input inventories, which largely have been ignored. Empirically, input inventories are more important than output inventories, especially in business cycle fluctuations. Maximum likelihood estimation of the decision rules yields relatively strong support for the model using data for nondurable and durable good industries. The value added specification outperforms gross production because adjustment costs on the change in materials usage are critical to fitting the data.

Social Employment Of Welfare Recipients In Belgium: An Evaluation

In Belgium, welfare agencies receive a subsidy to employ welfare recipients for a period sufficiently long to entitle them to unemployment benefits. This work experience program is called Social Employment (SE). We investigate the effect of SE on the exit rate from welfare. We propose a grouping/IV estimator of the SE effect that eliminates selection bias. The estimator is consistent, even if the selection into SE depends on the average unobserved characteristics of welfare recipients in a region and in a welfare duration interval. The empirical analysis suggests that there is creaming in the selection process. Without correction for selectivity we find that SE reduces welfare dependence, but after correction this conclusion is reversed. These results are consistent with the adverse incentives faced by the welfare agencies.

Semi-Nonparametric Estimation of an Equilibrium Search Model

We specify and estimate an equilibrium job search model with productivity differences across labor market segments. The model allows for two types of unemployment: frictional unemployment due to search frictions and structural unemployment due to wage floors. Wage floors exist because of high unemployment benefits or binding minimum wages. The productivity distribution is estimated semi-nonparametrically along the lines of Gallant-Nychka, using Hermite series approximation. We decompose the total unemployment rate and we examine the effect of changes in the minimum wage.

Efficient Estimation of Average Treatment Effects Using the Estimated Propensity Score

We are interested in estimating the average effect of a binary treatment on a scalar outcome. If assignment to the treatment is independent of the potential outcomes given pre-treatment variables, biases associated with simple treatment-control average comparisons can be removed by adjusting for differences in the re-treatment variables. Rosenbaum and Rubin (1983, 1984) show that adjusting solely for differences between treated and control units in a scalar function of the pre-treatment variables, the propensity score, also removes the entire bias associated with differences in pre-treatment variables. Thus it is possible to obtain unbiased estimates of the treatment effect without conditioning on a possibly high-dimensional vector of pre-treatment variables. Although adjusting for the propensity score removes all the bias, this can come at the expense of efficiency. We show that weighting with the inverse of a nonparametric estimate of the propensity score, rather than the true propensity score, leads to efficient estimates of the various average treatment effects. This result holds whether the pre-treatment variables have discrete or continuous distributions. We provide intuition for this result in a number of ways. First we show that with discrete covariates, exact adjustment for the estimated propensity score is identical to adjustment for the pre-treatment variables. Second, we show that weighting by the inverse of the estimated propensity score can be interpreted as an empirical likelihood estimator that efficiently incorporates the information about the propensity score. Finally, we make a connection to results to other results on efficient estimation through weighting in the context of variable probability sampling.

Solving Consumption Models with Multiplicative Habits

This paper provides derivations necessary for solving an optimal consumption problem with multiplicative habits and a CRRA 'outer' utility function, either for a microeconomic problem with both labor income risk and rate-of-return risk, or for a macroeoconomic representative agent model.