The current study incorporates a modified version of Kouri's (1983) portfolio balance rational expectations model to determine and forecast the bilateral exchange rate between India and the US for the period 1996:Q2-2019:Q3. The assumption of rational expectations enables us to analyse the factors representing current and capital account as macroeconomic determinants of exchange rates. The most significant contribution of the current study is however the inclusion of the microstructure theory within Kouri's (1983) framework, which permits us to determine the role of micro factors and macro factors, in influencing exchange rate. The use of the novel econometric tool, the Nonlinear Auto-Regressive Distributed Lag (NARDL) model, combined with various post-estimation tests, allows us to conclude in favour of asymmetric relationship between some of the exogenous variables and the exchange rate, with both micro and macro factors being important determinants of the exchange rate in the short-run. While evaluating the forecasting accuracy of the modified Kouri's (1983) model, we compare it to the Random Walk Model (RWM) across three forecast horizons: 6-months, 1-year, and 2-years. The results show that the modified Kouri (1983) model outperforms the RWM for all the forecast horizons considered.