Homepage of Eric R. Nielsen

Academic History

A pdf of my CV can be found here.

• A.B. Economics and Mathematics, Harvard University (2007)
• M.A. Economics, University of Chicago (2010)
• Ph.D. in progress, University of Chicago.

Research

My research fields are labor, industrial organization, and econometrics. My current research focuses on econometric theory, fertility dynamics, and the design of financial markets.

Working Papers

1. A Destructive Role for Speed Traders
2. Bounding Missing Observations using the Wasserstein Metric
   Econometricians frequently need to make inferences about a population distribution using data in which outcomes are missing for some observations. Inference for the population may be biased if these data are not missing at random. This paper investigates a non-parametric method, proposed by Santos and Kline, in which quantile bounds are constructed using only a restriction that the observed and unobserved distributions are "close" in the space of probability measures according to some metric on that space. Santos and Kline use the Kolmogorov-Smirnoff metric in their analysis. Here I argue that this measure does a poor job capturing the economically relevant cases of selection and that these cases are better examined using the Wasserstein metric. Using this alternate measure, I construct quantile bounds and prove they are tight. I next demonstrate how symmetry properties can be used to simplify the computation of the bounds and show that these symmetry properties hold in certain classes of threshold-crossing models common in the empirical literature. Finally, I show a tight connection between selection in classic parametric selection models and the Wasserstein distance between the observed and population distributions.
3. Testing the IID Assumption with Order Statistics
   Applied work with auction data must condition on observable differences across auctions in order to invoke techniques for recovering the distribution of private values from bidding behaviour. If this conditioning is not adequate, the implied private values from different auctions will not be order statistics from iid samples of a common value distribution and subsequent identification will be inconsistent. This paper develops two distinct non-parametric tests of the iid valuation assumption underlying much econometric inference on auction data. The first test uses the Fischer-Tipett-Gnedenko Theorem to compare the normalized maxima of independent samples to a known limiting distribution. The second builds on work in Zeithammer and Adams [2009] to construct a tail comparison test statistic for situations in which at least two order statistics are available per auction. This statistic is shown to be asymptotically normal when values are iid draws of a common distribution. Monte Carlo evidence shows that the theoretical results derived hold true in simulation. When there is random variation in the number of bidders the Zeithammer and Adams statistic loses some of its symmetry properties while tests based on the Fischer-Tipett-Gnedenko Theorem continue to perform reasonably well. Both tests show evidence of power against heterogeneity in the form of iid auction-level demand shocks.
4. The Geometry of Learning
   This paper examines several different models of costly experimentation in a network setting. Building on Bala and Goyal as well as on other results from the literature on Bayesian play on networks, I examine the likelihood of convergence and the speed of convergence when agents are myopic and the type space is non-trivial. I show that convergence need not occur when there are multiple payoff-relevant types of people. Following Bolton and Harris, I then model experimentation as a two armed bandit problem to explore aspects of strategic play in learning games. This model becomes intractable in the general network setting. I then investigate how the structure of the network affects its convergence properties by examining the differences and similarities between random graphs and power graphs. Power graph models more accurately mimic the structure of real-world social networks, and I find that such graphs are comparatively unlikely to be display convergence in a general setting.
5. The Effect of College Ranking on Alumni Giving
   This paper investigates the extent to which U.S. News and World Report (USNWR) college rankings influence giving decisions by college alumni. Using an instrumented variable technique to look only at ranking changes attributable to exogenous ranking formula changes, I find evidence that an increase in college ranking raises average alumni giving and participation.

Research Projects

1. Demand estimation in auctions with unobserved heterogeneity.
2. Marriage surplus under different legal regimes.
3. Information and trading frequency.
4. Testing deviations from Missing-At-Random in labor data sets.

Teaching Material

I have taught introductory graduate-level microeconometrics and price theory. Below are some lecture notes and example problems I have written. Feel free to use with caution but please give proper attribution when necessary.

Econometrics

1. Notes on non-parametric Manski Bounds.
2. Notes on BLP Demand estimation. These are based on a TA session I gave and enough students asked for notes afterwards that I decided to post these.
3. Notes on Completeness and its relationship to nonparametric identification.
4. Notes on dynamic discrete choice models and identification. Mostly based on the Rust (1994) handbook chapter.

Price Theory

1. Price Theory Encylopedia with answers by me, several classmates, and some students from prior years. Use with caution as many answers may be wrong or incomplete. I plan to improve on this and post more complete solutions over time.
2. Short paper on adverse selection in health insurance markets. Written for non-economists.