Publications

There are many economic environments in which an object is offered sequentially to prospective buyers. It is often observed that once the object for sale is turned down by one or more agents, those that follow do the same. One explanation that has been proposed for this phenomenon is that agents making choices further down the line rationally ignore their own assessment of the object’s quality and herd behind their predecessors. Our research adds a new dimension to the canonical herding model by allowing agents to di er in their ability to assess the quality of the offered object. We develop novel tests of herding based on this ability heterogeneity and also examine its efficiency consequences, applied to organ transplantation in the U.K. We nd that herding is common but that the information lost due to herding does not substantially increase false discards of good organs or false acceptances of bad organs. Our counter-factual analysis indicates that this is due (in part) to the high degree of heterogeneity in ability across transplant centers. In other settings, such as the U.S., where organ transplantation is organized very differently and the ability distribution will not be the same, the inefficiencies due to herding might well be substantial.

Functional coefficient (FC) regressions allow for systematic flexibility in the responsiveness of a dependent variable to movements in the regressors, making them attractive in applications where marginal effects may depend on covariates. Such models are commonly estimated by local kernel regression methods. This paper explores situations where responsiveness to covariates is locally flat or fixed. In such cases, the limit theory of FC kernel regression is shown to depend intimately on functional shape in ways that affect rates of convergence, optimal bandwidth selection, estimation, and inference. The paper develops new asymptotics that take account of shape characteristics of the function in the locality of the point of estimation. Both stationary and integrated regressor cases are examined. Locally flat behavior in the coefficient function has, as expected, a major effect on bias and thereby on the trade-off between bias and variance, and on optimal bandwidth choice. In FC cointegrating regression, flat behavior materially changes the limit distribution by introducing the shape characteristics of the function into the limiting distribution through variance as well as centering. Both bias and variance depend on the number of zero derivatives in the coefficient function. In the boundary case where the number of zero derivatives tends to infinity, near parametric rates of convergence apply for both stationary and nonstationary cases. Implications for inference are discussed and simulations characterizing finite sample behavior are reported. 

A semiparametric triangular systems approach shows how multicointegration can occur naturally in an I(1) cointegrated regression model. The framework reveals the source of multicointegration as singularity of the long run error covariance matrix in an I(1) system, a feature noted but little explored in earlier work. Under such singularity, cointegrated I(1) systems embody a multicointegrated structure and may be analyzed and estimated without appealing to the associated I(2) system but with consequential asymptotic properties that can introduce asymptotic bias into conventional methods of cointegrating regression. The present paper shows how estimation of such systems may be accomplished under multicointegration without losing the nice properties that hold under simple cointegration, including mixed normality and pivotal inference. The approach uses an extended version of high-dimensional trend IV estimation with deterministic orthonormal instruments that leads to mixed normal limit theory and pivotal inference in singular multicointegrated systems in addition to standard cointegrated I(1) systems. Wald tests of general linear restrictions are constructed using a fixed-b long run variance estimator that leads to robust pivotal HAR inference in both cointegrated and multicointegrated cases. Simulations show the properties of the estimation and inferential procedures in finite samples, contrasting the cointegration and multicointegration cases. An empirical illustration to housing stocks, starts and completions is provided.

This paper explores predictive regression models with stochastic unit root (STUR) components and robust inference procedures that encompass a wide class of persistent and time-varying stochastically nonstationary regressors. The paper extends the mechanism of endogenously generated instrumentation known as IVX, showing that these methods remain valid for short and long-horizon predictive regressions in which the predictors have STUR and local STUR (LSTUR) generating mechanisms. Both mean regression and quantile regression methods are considered. The asymptotic distributions of the IVX estimators are new and require some new methods in their derivation. The distributions are compared to previous results and, as in earlier work, lead to pivotal limit distributions for Wald testing procedures that remain robust for both single and multiple regressors with various degrees of persistence and stochastic and fixed local departures from unit roots. Numerical experiments corroborate the asymptotic theory, and IVX testing shows good power and size control. The IVX methods are illustrated in an empirical application to evaluate the predictive capability of economic fundamentals in forecasting S\&P 500 excess returns.

Lookalike Targeting is a widely used model-based ad targeting approach that uses a seed database of individuals to identify matching “lookalikes” for targeted customer acquisition. An advertiser has to make two key choices: (1) who to seed on and (2) seed-match rank range. First, we assess if and how seeding by others’ journey stages impact clickthrough (upstream behavior desirable for brand marketing) and donation (downstream behavior desirable in performance marketing). Overall, we find that lookalike targeting using other’s journeys can be effective-third parties can indeed identify factors unobserved to the advertiser merely from others’ journey stage to improve targeting. Further, while it is sufficient to seed on upstream journey stages for brand marketing, seeding on more downstream stages improves performance marketing outcomes. Second, we assess the effectiveness of expanding the target audience with lower match ranks between seed and lookalikes. The drop in effectiveness with lower match rank range is much greater for performance marketing (donation) than for brand marketing (click-through). However, performance marketers can alleviate the reduction in ad effectiveness for low match ranks by making targeting more salient; but increasing salience has little impact for high match rank. Overall, by increasing salience, performance marketers can make acquisition cost comparable for high and low match ranks.

Lookalike targeting is a widely used model-based ad targeting approach that uses a seed database of individuals to identify matching “lookalikes” for targeted customer acquisition. An advertiser has to make two key choices: (1) who to seed on and (2) seed-match rank range. First, we find that seeding on others’ journey stage can be effective in new customer acquisition; despite the cold start nature of customer acquisition using Lookalike audiences, third parties can indeed identify factors unobserved to the advertiser that move individuals along the journey and can be correlated with the lookalikes. Further, while journey-based seeding adds no incremental value for brand marketing (click-through), seeding on more downstream stages improves performance marketing (donation) outcomes. Second, we evaluate audience expansion strategies by lowering match ranks between the seed and lookalikes to increase acquisition reach. The drop in effectiveness with lower match rank range is much greater for performance marketing than for brand marketing. Performance marketers can alleviate the problem by making the ad targeting explicit, and thus increase perceived relevance; however, it has no incremental impact for higher match lookalikes. Increasing perceived targeting relevance makes acquisition cost comparable for both high and low match ranks.

We analyze nonlinear pricing with finite information. We consider a multi-product environment where each buyer has preferences over a d-dimensional variety of goods. The seller is limited to offering a finite number n of d-dimensional choices. The limited menu reflects a finite communication capacity between the buyer and seller.

We identify necessary conditions that the optimal finite menu must satisfy, for either the socially efficient or the revenue-maximizing mechanism. These conditions require that information be bundled, or “quantized,” optimally.

We introduce vector quantization and establish that the losses due to finite menus converge to zero at a rate of 1/n2/d. In the canonical model with one-dimensional products and preferences, this establishes that the loss resulting from using the n-item menu converges to zero at a rate proportional to 1/n2.