In contrast, the common feature of all works on global identification is the reliance on some numerical algorithm to search over all observationally equivalent structural parameters. If the model is only locally non-identified, this should work well. However, if the model is locally but not globally identified, we cannot rule out a situation in which we failed to find some observationally equivalent points (even if such points exist) and falsely conclude that the model is globally identified. Thus, the fact that no analytical results have been offered to date is a serious methodological gap. The objective of this project is to fill this gap and propose an analytical framework which, when combined with recent developments in the field of symbolic (i.e. analytical) computation, will be able to effectively prove whether a given DFLE model is globally identified at a given point in the parameter space or not and, in the latter case, help understand the source of identification failure.
While designed to deal with global identification, our framework will also offer some additional insights into the local identification problem. In contrast to existing and well-established approaches, in the latter case, our identification analysis will explicitly indicate whether there is a possibility to achieve identification by fixing some particular parameters. This formal information concerning identifiability is not available using any alternative frameworks. When applied to important macroeconomic models existing in literature, our framework may either strengthen the confidence in their economic implications (if global identification is confirmed) or force economists to rethink them or at least treat them with caution (if global identification failure is detected). Needless to say, such findings have the potential to exert a significant influence on the design of macroeconomic models in the future.