This paper offers an original survey of the Malaysian crisis and the effects of the consequent imposition of capital controls by authorities in September 1998 and of their subsequent relaxation in February and September 1999. We identify Malaysia's unique strengths and weaknesses before the crisis, appreciate the differential timing and nature of the Malaysian crisis vis-à-vis the other neighbouring crisis countries, and distinguish carefully between the restrictive and incentive components of the imposed controls. Against this backdrop, we analyse both the "level" (first-order) effects and the "volatility" (second-order) effects of controls on key macroeconomic, banking and financial market variables. On the level effects, we found the Malaysian recovery (starting late 1998) to be at least as quick, strong and lasting as that of the other crisis countries, and discovered important channels of influence from controls to interest rates (which were lowered) and stock markets (which recovered dramatically). These results on the effectiveness of controls are consistent with earlier studies by Edison due to the longer time period used here, a stronger restatement of their conclusions is now possible. We study the secondorder effects of controls by introducing a Capital Asset Pricing Model (CAPM)-based portfolio choice model and show how controls, especially when they work as an asymmetric tax on short-term investment, reduce both the volume of speculative flows and the associated interest rate volatility. To test these theoretical results, we set up a standard Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model of interest rate and stock market volatility where capital control dummies are introduced in the variance equation. Our model is an improvement over earlier studies in two ways: a more sophisticated capital control dummy was used to take account of the relaxation of controls in February 1999; and we dealt with the problem of endogeneity in the mean equation by using regressors that are not Granger-caused by the regressand (Malaysian interest rate and stock returns). The model shows that controls did limit interest rate volatility in line with the theoretical prior, but worsened stock market volatility. The latter result lends credence to the view that controls shifted the burden of adjustment from quantity to prices.