Essays on partial identification in econometrics and finance

This dissertation collects three essays on the issue of partial identification in Econometrics and Finance. The first two deal with the issue of partial identification in Econometrics, and the last one with a problem of partial identification in Finance. The first essay offers a conceptual framework to think about partial identification in link with the theory of Dempster-Shafer belief functions. It introduces an original testing methodology by the mean of a Kolmogorov-Smirnov test for Choquet capacities. The second essay propose an alternative testing methodology with favorable computational properties, the "Dilation Bootstrap," a testing methodology based on probabilistic coupling representations of the empirical distribution. The last essay takes a slightly different angle to relate the issue of partial identification with the Law of Absence of Arbitrage in finance, in the baseline case of an option market on two correlated assets.