| The main contribution of the first essay is to provide a framework in which the notion of farsighted stability for games, introduced by Chwe (1994), can be applied to directed networks. In particular, we introduce the notion of a supernetwork. A supernetwork is made up of a collection of directed networks (the nodes) and uniquely represents (via the arcs connecting the nodes) agent preferences and the rules governing network formation. By reformulating Chwe's basic result on the nonemptiness of farsightedly stable sets, we show that for any supernetwork (i.e., for any collection of directed networks and any collection of rules governing network formation), there exists a farsightedly stable directed network. We also introduce the notion of a Nash network relative to a given supernetwork, as well as the notion of a noncooperative supernetwork. To illustrate the utility of our framework, we present several examples of supernetworks and we compute the farsightedly stable and Nash networks.; The second essay focuses on the topic of volatility forecasting. The quality of conventional volatility forecasting models can be improved through exploiting the information content of open, high, low, and close prices and by making the models less sensitive to the impact of outliers. Simple techniques that exploit distributional properties and readily available additional information may prove useful for improving volatility forecasts as measured as an ability to predict one-day ahead as well as longer-term volatility. Range based volatility estimators are viable alternatives to using high frequency data and to more complex models. The method of trimming also is introduced for volatility estimation and improves forecasting accuracy through making models less susceptible to the effects of outliers. Both inclusion of better distributional measures of volatility and trimming of outliers have an appreciable impact on forecasting accuracy for a selected sample of equities. |
