Skip to main content
main-content

Über dieses Buch

The authors present compelling evidence, supported by their own measure: the 'adjusted root mean square error', to finally solve the Meese-Rogoff puzzle and provide a new alternative.

Inhaltsverzeichnis

Frontmatter

1. The Meese-Rogoff Puzzle

The Meese-Rogoff puzzle refers to the proposition that exchange rate models cannot outperform the random walk in out-of-sample forecasting of exchange rates. This proposition is regarded as a puzzle because it does not make much sense for profit-maximising firms to pay for professional forecasts when the ‘better’ forecasts generated from the random walk are available for free. This proposition is valid only if forecasting accuracy is measured in terms of criteria that depend on the magnitude of the forecasting error only. The origin of the puzzle, the 1983 paper of Meese and Rogoff exhibits many flaws.

Imad A. Moosa, Kelly Burns

2. A Selective Survey of Subsequent Studies

The literature on the Meese-Rogoff puzzle deals with attempts to resolve the puzzle and overturn the Meese and Rogoff results. While various studies show that the results cannot be overturned, some economists have claimed victory over the random walk, but most of these claims are groundless. A conclusion like this is reached either without appropriate statistical testing to find out if the difference in the root mean square errors is statistically significant or through the use of dynamic models, which amounts to beating the random walk with a random walk. The literature also deals with various issues such as the effect of the forecasting horizon and whether the appropriate benchmark should be the random walk with or without drift.

Imad A. Moosa, Kelly Burns

3. Basic Methodology, Data and Results

Three models are used to generate forecasts for six exchange rates: the Frenkel-Bilson flexible-price monetary model, the Dornbusch-Frankel sticky-price monetary model and the Hooper-Morton monetary model with current account effects. The basic methodology resembles that used by Meese and Rogoff. The models are estimated over part of the sample period, and then forecasts are generated out of sample by using recursive estimation. For the basic results, the root mean square error is used as a measure of forecasting accuracy. As a benchmark, the choice between the random walk with and without drift depends on the statistical significance of the drift term.

Imad A. Moosa, Kelly Burns

4. Alternative Measures of Forecasting Accuracy

Alternative measures of forecasting accuracy include direction accuracy, the adjusted root mean square error, profitability and proximity to a perfect forecast. The results demonstrate that the random walk can be outperformed in exchange rate forecasting when forecasting accuracy is assessed in terms of measures that take into account more than just the magnitude of the forecasting error. Evaluating forecasting accuracy by using alternative measures leads to vastly different conclusions from those reached by using conventional measures such as the root mean square error. The three models produce better forecasts than the random walk when evaluated in terms of alternative criteria. This is a potential explanation for the Meese-Rogoff puzzle.

Imad A. Moosa, Kelly Burns

5. Stochastic Movements in the Underlying Parameters

Incorporating stochastic movements into the parameters of exchange rate models (by estimating the models in a time-varying parametric framework) leads to an improvement in forecasting accuracy in terms of the magnitude of error. Although Meese and Rogoff are correct in suggesting that the use of TVP enhances forecasting accuracy, the improvement is insufficient to outperform the random walk in terms of the magnitude of the error. However, the random walk is outperformed by exchange rate models estimated in a TVP framework when forecasting accuracy is assessed by alternative metrics. The Meese-Rogoff puzzle can be resolved using alternative measures of forecasting accuracy, but not by the mere use of TVP estimation while the RMSE is used as a criterion.

Imad A. Moosa, Kelly Burns

6. Model Misspecification

Model misspecification is put forward as a possible explanation for the Meese-Rogoff puzzle, in the sense that models fail to outperform the random walk because they are misspecified. However, three variations of exchange rate determination models (the post-Keynesian flow model, a version of the F-B model that distinguishes between traded and non-traded goods, and the F-B, D-F and H-M models without the proportionality and symmetry restrictions) cannot outperform the random walk in terms of the magnitude of the forecasting error. In contrast, when the forecasts of the three models are assessed by alternative measures of forecasting accuracy, a different conclusion is reached. The random walk can be outperformed by alternative model specifications only if forecasting accuracy is assessed by measures that do not rely exclusively on the magnitude of error.

Imad A. Moosa, Kelly Burns

7. The Effect of Non-linearities

It has been suggested that inability to outperform the random walk may result from the use of exchange rate models that are linear in parameters when the exchange rate is a non-linear function of macroeconomic variables. It is well documented that many functional relations in finance are intrinsically non-linear and that there may be non-linearities in exchange rate adjustment. We find that the forecasting performance of exchange rate models (in terms of the magnitude of error) improves substantially when specified as a non-linear error correction model. Despite this finding, the non-linear model cannot produce an RMSE that is numerically smaller and statistically different from that of the random walk.

Imad A. Moosa, Kelly Burns

8. Simultaneous Equation Bias

Simultaneous equation bias, which arises as a result of endogeneity of the explanatory variables, has been suggested as an explanation for the Meese-Rogoff puzzle. Endogeneity of macroeconomic variables may arise from the possibility of monetary policy feedback in the monetary model, which means that a model that incorporates endogeneity should produce better forecasts in terms of the magnitude of error. However, we find, with one exception, that the VAR models corresponding to the Frenkel-Bilson, Dornbusch-Frankel and Hooper-Morton static models do not produce significantly smaller RMSEs than that of the random walk.

Imad A. Moosa, Kelly Burns

9. Sampling Errors

Some economists suggest that sampling errors may resolve the Meese-Rogoff puzzle and that the forecasting power of exchange rate models is sensitive to sample selection. It is also suggested that the length of the estimation window is important in that using additional historical observations enhances forecasting performance. Furthermore, it is suggested that the selection of the forecasting window impacts forecasting performance. We find that a narrower estimation or forecasting window cannot explain the Meese-Rogoff puzzle. In general we find that changing the sample period and using different forecasting and estimation windows does not make much difference.

Imad A. Moosa, Kelly Burns

10. Modelling Expectations

Expectations play an important role in the monetary models of exchange rates because the long-run expected inflation rate is an explanatory variable in both the Dornbusch-Frankel and the Hooper-Morton models. This is why it has been suggested that improper modelling of inflationary expectations may explain the Meese-Rogoff puzzle. However, we find that regardless of how inflationary expectations are measured, the random walk cannot be outperformed in terms of the magnitude of the forecasting error. Very little happens to measures of forecasting accuracy as we change the specification of the expectation formation mechanism in either of the two models. The Meese-Rogoff puzzle cannot be resolved by changing the specification of the expectation formation mechanism as long as the RMSE is used to measure forecasting accuracy.

Imad A. Moosa, Kelly Burns

11. Concluding Remarks

The Meese-Rogoff puzzle should not be considered a puzzle, and exchange rate models should not be expected to outperform the random walk in terms of magnitude-only measures of forecasting accuracy. The superior performance of the random walk in terms of the RMSE and similar criteria should be expected because the forecasting error of the random walk is the period-to-period change in the exchange rate, which is typically small. The Meese and Rogoff findings cannot be overturned, but only in the narrow sense that exchange rate models cannot produce a numerically smaller and statistically different magnitude of error compared to the random walk. The only plausible explanation for the Meese-Rogoff puzzle is that forecasting accuracy is assessed exclusively by the magnitude of error.

Imad A. Moosa, Kelly Burns

Backmatter

Weitere Informationen

Premium Partner

micromStellmach & BröckersBBL | Bernsau BrockdorffMaturus Finance GmbHPlutahww hermann wienberg wilhelmAvaloq Evolution AG

BranchenIndex Online

Die B2B-Firmensuche für Industrie und Wirtschaft: Kostenfrei in Firmenprofilen nach Lieferanten, Herstellern, Dienstleistern und Händlern recherchieren.

Whitepaper

- ANZEIGE -

Blockchain-Effekte im Banking und im Wealth Management

Es steht fest, dass Blockchain-Technologie die Welt verändern wird. Weit weniger klar ist, wie genau dies passiert. Ein englischsprachiges Whitepaper des Fintech-Unternehmens Avaloq untersucht, welche Einsatzszenarien es im Banking und in der Vermögensverwaltung geben könnte – „Blockchain: Plausibility within Banking and Wealth Management“. Einige dieser plausiblen Einsatzszenarien haben sogar das Potenzial für eine massive Disruption. Ein bereits existierendes Beispiel liefert der Initial Coin Offering-Markt: ICO statt IPO.
Jetzt gratis downloaden!

Bildnachweise