Job Market Paper:
Abstract: Estimates of peer effects may suffer from bias if the network data has missing links. Moreover, if links are not missing at random, estimates of parameters in network formation models may also be biased. I contribute to the literature on identification of peer effects and network-formation models with partially sampled network data in three ways. My first contribution is to develop a consistent peer-effects estimator that uses link-formation probabilities from the true network-formation model. My second contribution is to develop two estimators of a network formation model that are robust to the missingness of links correlating with unobservable link-specific shocks. These are an inverse-probability-weighted likelihood estimator that uses the probabilities of observing links as weights, and a semi-parametric estimator. The first estimator requires the researcher to estimate the probability of a link being observed, while the second does not at the cost of a slower convergence rate. My third contribution is to show sharp partial identification of endogenous peer effects when there is no information on how the network is formed. The bounds from this exercise will be more informative if the researcher has information about the unobserved network. I apply my peer-effect estimator to a new dataset from two Norwegian schools that I merge with administrative data. In this data, I observe the complete network as well as a partial sample of links constructed by restricting students to only list some of their friends. Using the complete and partial sample of network links from the data, I find that my method reduces the bias in the peer-effect estimator by 65%. Finally, I demonstrate that naive estimators can lead to misleading results about household behavior in microfinance take-up by applying my peer-effect estimator to the dataset of Banerjee et al. (2013).
"Residential Investment and Recession Predictability"
International Journal of Forecasting (2019) (With André K. Anundsen and Knut Are Aastveit)
Abstract: We assess the importance of residential investment for the prediction of economic recessions for an unbalanced panel of 12 OECD countries over the period 1960Q1–2014Q4. Our approach is to estimate various probit models with different leading indicators and evaluate their relative prediction accuracies using the area under the receiver operating characteristic curve as our forecasting performance metric. We document that residential investment contains information that is useful for predicting recessions both in-sample and out-of-sample. This result is robust to adding typical leading indicators, such as the term spread, stock prices, consumer confidence surveys and oil prices. It is shown that residential investment is particularly useful for the prediction of recessions for countries with high home-ownership rates. Finally, in a separate exercise for the US, we show that the predictive ability of residential investment is — in a broad sense — robust to employing real-time data.
Work in Progress:
Abstract: Using blind and non-blind evaluations, we show that the common identification strategy of gender bias in schools cannot identify a weighted average of gender bias for different ability levels due to the discrete nature of grades. We propose a new method, similar in spirit to matching, that enables the identification of heterogeneity in gender bias along both observable and unobservable dimensions. Utilizing our estimator, we quantify the difference-in-difference estimator to have a bias of approximately 30 percent of the true effect in our Norwegian high school data, and we find that gender bias can explain most of the average grade gap between male and female students. Low-ability students are exposed to the highest levels of bias in a classroom setting, and a one-standard-deviation-higher expected bias leads to students, on average, losing $750–1,000 in labor income each of the first five years after high school graduation
An Empirical evaluation of trade-offs in hiring: Education vs. Experience (With Ragnar Alne and Christine Blandhol)
Abstract: We construct firms stated preference rankings across education and experience levels using a survey sent to 23200 Norwegian firms. Through the survey we get the firms to state their willingness to pay, both in time and money, for the different alternatives. Finally, we compare these stated preferences to the revealed preferences of the firms through their hiring decisions, allowing us to measure the cost of firms not having access to their ideal candidates. Data gathering is currently in process, due to finish early 2023.
Inference for the solutions to Linear Programs (With Jonas Lieber)