Parallel Currency Markets and the Monetary Exchange Rate Model: A VECM Application to Turkey Over 1987–1998

Abstract

This study takes a novel approach to testing the monetary model of exchange rate determination by allowing for distortions from the underground economy to be reflected via inclusion of the parallel exchange rate. Using a VECM methodology, an augmented model is tested and compared to a traditional flexible-price monetary model for Turkey over 1987–1998. While in-sample results are supportive of both models, out-of-sample analysis favors inclusion of the parallel exchange rate. Furthermore, the augmented model beats a random walk with and without drift over to mid-range forecast horizons. These results highlight the need to consider potential market distortions of exchange rate movements.

Keywords:

• emerging markets
• forecasting
• parallel exchange rates
• underground economy

JEL Classification:

• F31
• F33
• F37

ACKNOWLEDGMENTS

Special thanks to Merih Uctum, Wim Vijverberg, and Amanda Stype. I am especially grateful for the anonymous reviewers’ comments which greatly improved this research.

FUNDING

This work was supported by a grant from the Graduate Center, City University of New York DSRG award program.

Notes

1. Mirus and Smith (1997) provide a formal definition of underground economic activity, broken down by type of transaction (monetary vs. nonmonetary, legal vs. illegal.)

2. Following Reinhart and Rogoff (2004), define $S_t^P$ and S_t as the parallel and official nominal spot exchange rates in levels. The premium $\frac{S_t^P-S_t}{S_t}$ is the spread between the two rates relative to the official exchange rate.

3. See Schneider and Enste (2000) and Schneider and Buehn (2013) for further information related to the techniques involved in estimating the size of shadow (underground) economies.

4. One point of distinction is that Van Den Berg and Jayanetti (1993) model the parallel exchange rate as a function of the official rate and cross-country fundamentals. Given the system’s endogeneity and the likelihood that the parallel exchange rate is more forward looking, I normalize the official exchange rate.

5. Both tests are conducted including twelve lags and a constant term. The test on $(m_t-m_t^{\ast })$ also includes a trend term.

6. Forecast evaluation results for 1998:8–1998:12 are available upon request. As previously noted, the results are potentially contaminated by spillover from the Russian financial crisis of 1998.

7. This result is as expected. The official lira-US$ nominal exchange rate more closely resembles a random walk with drift, given the near-persistent devaluation over 1987–1998. An application of the simulation-based approach confirms that a random walk with drift yields a better fit to the data. Results are available upon request.

8. For brevity, only a random walk with drift is presented as the alternative for out-of-sample comparisons. With nearly all M ≈ 1, the driftless random walk simulation results are consistent with Table 5. Full results are available upon request.

Additional information

Notes on contributors

Evan Warshaw

Evan Warshaw is a doctoral candidate at the Graduate Center, City University of New York.