| Popis: |
In empirical financial accounting research, ratios or first differences are usually calculated based on the financial and accounting data provided by the CRSP, Compustat and I/B/E/S datasets. However, in the construction of longer time series, the variables in these datasets present a challenge to researchers. To be able to make valid empirical estimations, variables have to be stationary. If they are not, spurious results may be yielded. When creating separate sets of accounting, finance and analyst variables for the NYSE/AMEX and NASDAQ stock exchanges, we concluded that many popular variables are non-stationary: they have partial unit-roots and systematic structural breaks, mostly around important economic events, such as the 1991-92 recession, the dot.com bubble and the 2008 mortgage-related financial crisis. In this study we present several approaches to locating structural breaks in data. Our objective is twofold: 1) to facilitate researchers in improving the quality of their estimations and inferences, and 2) to better enable them to identify models for handling endogeneity issues. |
