The first and last chapter of this dissertation are devoted to the econometric theory of twounrelated topics. The second chapter covers an empirical study of the novel model in the first chapter. The first chapter studies novel estimation procedures with supporting econometric theory for a dynamic latent-factor model with high-dimensional asset characteristics, that is, the number of characteristics is on the order of the sample size. Utilizing the Double Selection Lasso estimator, our procedure employs regularization to eliminate characteristics with low signal-to-noise ratios yet maintains asymptotically valid inference for asset pricing tests. The second chapter studies the dynamics of crypto asset returns through the lens of factor models, and in particular compare the out of sample pricing ability of our novel factor model against relevant benchmarks. We were motivated to develop our new method given, in the setting of crypto asset returns, there are a limited number of tradable assets and years of data as well as a rich set of available asset characteristics. In an additionally novel empirical panel, we find the new estimator obtains comparable out-of-sample pricing ability and risk-adjusted returns to benchmark methods. We provide an inference procedure for measuring the risk premium of an observable nontradable factor, and employ this to find that the inflation-mimicking portfolio in the crypto asset class has positive risk compensation. Finally, specifying a factor model with nonparametric loadings and factors, we utilize recent methods in deep learning to maximize out-of-sample risk-adjusted returns in an hourly panel, which yields economically significant alphas even after a detailed accounting of transaction costs. The third chapter (coauthored with Manu Navjeevan) studies a novel estimator for the conditional average treatment effect (CATE) with a doubly-robust inference procedure. Plausible identification of CATEs can rely on controlling for a large number of variables to account for confounding factors. In these high-dimensional settings, estimation of the CATE requires estimating first-stage models whose consistency relies on correctly specifying their parametric forms. While doubly-robust estimators of the CATE exist, inference proce- dures based on the second-stage CATE estimator are not doubly-robust. Using the popular augmented inverse propensity weighting signal, we propose an estimator for the CATE whose resulting Wald-type confidence intervals are doubly-robust. We assume a logistic model for the propensity score and a linear model for the outcome regression, and estimate the param- eters of these models using an `1 (Lasso) penalty to address the high-dimensional covariates. Inference based on this estimator remains valid even if one of the logistic propensity score or linear outcome regression models are misspecified. To our knowledge, we are the first paper to develop doubly-robust pointwise and uniform inference on an infinite dimensional target parameter after high dimensional nuisance model estimation.