RDP 2015-05: The Social Costs of Currency Counterfeiting Appendix B: Alternative Model

This appendix contains the results of a structural vector error correction model (VECM) using the same endogenous, exogenous and dummy variables as the baseline model in the main text. The results suggest that the estimates of the baseline model are conservative. In contrast to the baseline model, we find that the response of bank deposits to the counterfeiting shock is statistically significant and much larger. The response of banknotes in circulation and credit card debt are not statistically different to the baseline estimates.

The structural VECM estimates both the short-run and the long-run dynamics of the relationship between the endogenous variables:

Yt is the vector of endogenous variables and Xt is the vector of exogenous variables. Aj are matrices of coefficients for each lag, j, of the endogenous variables governing short-run dynamics. γ0 captures the contemporaneous effect of the exogenous variables in each equation in the system. β is a single cointegrating vector governing the long-run equilibrium in the endogenous variables. α is a vector of speed of adjustment parameters that measure the speed with which endogenous variables return to the long-run equilibrium. The matrix A0 contains information governing the contemporaneous relationships between the endogenous variables. The vector of equation residuals, εt, are the structural shocks to the system.

The four endogenous variables are normalised by nominal GDP and enter the model in logs. The cash rate enters the model in its level as the single exogenous variable. Dummy variables account for seasonality effects and the effects of the global financial crisis. We identify counterfeiting shocks using the same recursive-ordering identification assumption as in the baseline model. We also assume that the counterfeiting shock does not have long-run effects on the other endogenous variables using restrictions on the α vector.

In contrast to the baseline model, the impulse response estimate for bank deposits is statistically significant and much larger (Figure B1). The response of banknotes in circulation to the counterfeiting shock is smaller in magnitude compared to the baseline, and the response of credit card debt is around the same magnitude, but neither are statistically significant. These results suggest that the estimates of the baseline model are conservative.

Figure B1: VECM Impulse Response Functions