Finance is a
branch where we have the most up-to-minute data from stock exchanges and the
field is characterized by strong empirical traditions. Still we conduct
experiments in finance and try to gather even more data, is it needed? In the words of Editor, Econometrica, “Finance is
in no need of experiments. We have lots of data.”. Also, the real world markets
are bigger and complex as compared to the laboratory so can an experimentally
proved model fit the real world market.
Experiments prove
their relevance as they facilitate the researchers in isolating and
manipulating a single variable at a time, thus, the causal effects are
illustrated better by filtering the effects of different variables associated
with the model. Another aspect of experiments is that it allows to observe the
otherwise unobservable, dependent and independent variables due to the
laboratory settings and eliminates the complications of self-selection by
assigning subjects randomly to the different treatments.
However, a key challenge is to construct an experiment that is purely
based on the assumptions of the model and in which the alternative hypotheses
are plausible such that the results of the experiment do not rely completely on
the forgone conclusions. It can be done by examining the settings which are too
complex to be definitively modeled or by relaxing the different assumptions
underlying the the model, namely, behavioral, structural and equilibrium
assumptions.
Archival data analysis faces major challenge in interpreting the results
as data that are considered for analysis are not meant for answering the
research questions. This may lead to problems such as omitted variables biases,
self-selection biases, unobservable independent variables, and unobservable
dependent variables.
