I am slowly and steadily reading everything that Weitzman has ever written. Weitzman writes theory, not empirical papers. I use satellite imagery and other sources of big data to examine the spatiotemporal development of resource clusters. Specifically, and most recently, I estimate the impact of the 2011 Indonesian moratorium on new forest concessions for agriculture production. The intended impact was to reduce the rate of deforestation, but instead I find that the aggregate rate of deforestation increased as a direct result of the policy.
The market for online content is highly competitive. Consumers now move freely across websites with virtually zero transaction costs. And, in fact, Internet users have come to expect zero-cost interactions. Recent studies have shown that even nominal load times, only 400 milliseconds, are enough to drive users away from websites. Load time alone is a substantial cost to users, but it is just one component of multifaceted hassle costs which also include search costs and the mental overhead to engage with new content. What is the impact on website usage when marginal barriers are introduced before the content? Is the time to engage the only barrier, or are other factors just as important?
I work with George Judge to apply his application of permutation entropy to economic time series. Below, we've calculated the normalized permutation entropy for each 365 day block, offset by one day, for the Dow Jones Industrial Average between May 26, 1896 and March 4, 2013 for a total of 29,253 observations. Historical shifts in financial policy are clear, such as turmoil during the Great Depression and the Reagan deregulation.
Traditional remote sensing methods tend to focus on spectral characteristics of individual pixels. These methods are well-suited for moderate resolution imagery. However, now, commercial satellite imagery providers are able to serve out imagery on a regular basis at 25cm resolution. The higher resolution allows for more advanced methods in computer vision to be applied in full force to satellite imagery. These methods leverage information on the spatial structure of nearby pixels, not just the coloration within a pixel. Applications range from counting cars in parking lots to monitoring the growth of refugee camps.
I was the Graduate Student Instructor for Max Auffhammer's graduate-level econometrics class in Spring 2013. I wrote an open source repository for the section notes.
A critical component of the forest monitoring algorithm is identifying breaks in time series data. We extend this variety of break detection for spatiotemporal data. This presentation, given at Clojure/West, describes the algorithm for mapping a function across partitions to identify structural shifts in time series.