Abstract: We suggest the use of outdegrees from graph theory to rank locations in terms of their contagiousness. We show that outdegrees are equal to the column sums of spatial autoregressive matrices, which may be estimated using econometric methods for spatial panel data. In contrast to outdegree, R is invalid for 'traffic light' shading because it fails to distinguish between the export and import of contagion between sub-national locations. Simulation methods are used to illustrate the concept of outdegrees and its structural determinants in terms of centrality, indigenous contagion and spatial contagion. An empirical illustration is presented for Israel. A secondary criterion for traffic light shading involves the stochastic structure of morbidity shocks, which induce 'spiking' through their autoregressive persistence, conditional heteroscedasticity and diffusion jump parameters.
Abstract: This study shows that when uncertainty is higher, developers are willing to forego more revenue to sell early. The results are economically large, highly significant, and robust to various specifications. Using a new model of sale timing under uncertainty, I show that these results are consistent with developers using presale to accumulate precautionary savings to hedge against the possibility of demand being low when the building is completed. This precautionary motive for presale is consistent with industry participants reporting that a “failed” project results in reputational damage that is far greater than any increase in reputation from “successful” projects. Thus, industry participants’ statements and empirical evidence both point towards a negatively skewed outcome structure for developers. Developers having such an outcome structure has important implications for housing policy and affordability as well as for our understanding of macroeconomic fluctuations.
Abstract: I study the effects of improved public information on equilibrium welfare and price dispersion, providing sufficient conditions for negative and positive effects. Public information affects welfare by reducing excessive (though rational) pessimism induced by sequential learning. Reduced pessimism results in fewer agents withdrawing from the market prematurely (compared with the full information benchmark), which increases own ex-post utility in expectation but may also cause a congestion externality. I show that either effect can dominate. Observed search duration on the short side of the market is an indicator of the welfare effects of public information. The context is a search, matching and bargaining market with uncertainty about the meeting probability on both sides. Fully rational, ex-ante identical participants gather private information endogenously through costly search. Full trade is the unique perfect Bayesian equilibrium under general conditions. Full trade implies learning terminates following a positive but not following a negative signal, which results in a declining belief path and in declining reservation prices during search. Public information is modelled as the precision of a public signal about the true state. The results hold for any prior distribution where a more precise public signal slows learning.
Abstract: This work investigates how the amount of information available for a given transaction affects the price distribution and seller time on the market in that transaction. Novel measures of transaction level information are constructed based on random omission in records of previous transactions in similar apartments. Some of the ex ante more reliable information measures yield strong evidence that more information can lead to shorter seller time on the market. Other information measures either have no effect, have an unstable effect or are not credible. The evidence regarding the effect on the price distribution (price levels and variance) is inconclusive. The evidence regarding time on the market is consistent with previous evidence and with theoretical predictions.
The Causal Effect of a Housing Transaction
Abstract: What is the causal effect of a housing transaction on subsequent transactions? This question is at the heart of recent studies of housing market phenomena such as time series anomalies (Glaeser and Nathanson JFinEcon 2017) and foreclosure contagion (Gupta JFin 2019). I propose a novel method for answering this question by separating the real effects of the transaction (subtraction of a potential buyer and a listed house from the market) from its informational effect (adding a transaction to the list of comparables for future similar transactions). To do so, I leverage exogenously missing data about apartment characteristics in the published deeds records between 2010-2015 to study the effects of comparable transactions on future transactions. Missing characteristics perturb the informational effect of a given transaction. At the same time, I am holding fixed the real effect by using a retroactively corrected database. This allows me to separately identify the two effects.
Spatio-Temporal Epidemiological Modeling of COVID-19 Transmission in Israel (Joint with David Genesove, Michael Beenstock and Daniel Felsenstein)
Abstract: Transmission of Covid-19 requires physical proximity between people, and people make decisions about where to go that partially take this into account. We build on these two ideas, as well as on advances in network theory, spatial econometrics and economic decision models to model Covid-19 transmission. Tools from network theory allow us to identify areas that are responsible for a disproportionate share of overall infections (super-spreaders), while tools from spatial econometrics allow us to identify the rate of asymptomatic contagion. Economic decision models allow us to account for people’s endogenous choice of how much and where to go as a response to fears of contagion and government policy. We incorporate these insights into a unified econometric framework. We then use administrative data on infections, tests and vaccination per statistical areas in Israel, as well as detailed demographic information and cell-phone based mobility patterns, to provide new insights about optimal policy.
Belief Updating During Unemployment (Joint with Nofar Duani)
Abstract: We design an experiment where we repeatedly elicit the full distribution of job seekers’ beliefs about their future length of unemployment and provide them with information about the true distribution of comparable job seekers. This allows us to study how job seekers update their beliefs based on official data. By eliciting the beliefs of the same job seekers again after several weeks, as well as their search outcomes during these weeks, we can further study how beliefs are updated based on search outcomes. Eliciting the full prior of beliefs is a novel feature of our design which allows us to separately estimate changes in the mean as well as in the dispersion of beliefs. The dynamics of belief dispersion are crucial for understanding how confidence and learning interact to shape the path of reservation wages and the timing of quitting from search.