Adam conducts research in the field of econometrics. Much of his recent work has focused on hypothesis testing problems when size, the probability of concluding a true hypothesis is false, is difficult to control. Adam has developed new techniques that both control size and "reject" false hypotheses with high probability. Adam also does research in time series econometrics, developing tools to accurately analyze the persistence properties of economic variables recorded over time.
Adam's research lies broadly within the field of econometric theory. Much of his recent work has focused upon hypothesis testing problems in environments under which size, the probability of concluding a true hypothesis is false, is difficult to control. In order for a hypothesis test to provide informative results when taken to the data, an econometrician must be able to limit the probability he incorrectly concludes a hypothesis is false. That is, must be able to control the size of the test. There are many examples of tests used by applied economists that exhibit size-distortions, or lack of size control, when standard methods are used. These include testing after selecting a regression model, testing after a pretest, testing in autoregressive models that may contain a unit root, testing when a parameter may be on a boundary, testing after using an instrumental variable, testing when identification may be weak and testing after taking averages across different models. Most existing methods that successfully control size in these environments can be very conservative in the sense that they rarely tell the applied economist that a false hypothesis is indeed false. Adam has developed new econometric techniques that both control size and "reject" false hypotheses with high probability. He has also developed a new mathematical framework for analyzing some of these problems. Adam continues to work on improving size-controlled tests in order to make them more informative to the user. He is especially interested in addressing unresolved issues arising from commonly used tests after model selection and pretests. Model selection and pretests are procedures that help the empirical economist decide how to model the data before subsequently testing a hypothesis.
Adam also has conducted and continues to conduct research in time series econometrics. His research as a graduate student focused on developing robust estimation methods for time series models frequently employed in finance and macroeconomics such as GARCH, ARMA, long-memory and stochastic volatility models. These models are often applied to market volatility and inflation data, for example. When the mean of a time series changes within an observed time span, standard estimation methods for popular models are largely biased, making the time series look more persistent than it is in reality. In a variety of contexts, Adam has developed new methods that overcome this problem, allowing one to precisely obtain the persistence properties of a time series without needing to specify if and/or how the mean of the time series is changing. When these procedures are taken to economic data, the results are often striking, decreasing classic persistence measures by 80-90% when compared with standard estimation methods. Using methods they developed in one study, Adam and Pierre Perron (of Boston University) provide evidence that the common finding of "long-memory" in certain stock market volatility series is, at least in part, spurious.