Understanding how the exchange rate will impact a foreign investment is a critical step prior to committing any monetary value. Financial forecasting is both an art and a science, as financial analysts utilize different models to predict how the exchange rate will fluctuate over a given period of time. The number of models for forecasting exchange rates or other financial predictors are numerous and can be divided into a variety of categories based on the variables they incorporate, the calculations they perform, and the outcomes they predict. Some models are useful for predicting the exchange rate, while others are more beneficial at predicting the differential rate of inflation or growth. Before investing in German markets, it is crucial to forecast the potential value of the euro against the U.S. dollar in order to lend investment portfolios credibility. In order to recommend investment, forecasts of the euro-dollar exchange rate must be reliable and demonstrate soundness. They should not indicate a significant decrease of either currency against the other.
For example, if the euro significantly increases in value, and the U.S. dollar begins to perform unfavorably, an investment in the euro would be encouraged as a sound financial decision. However, if the euro begins to perform unfavorable, an investment would not be considered profitable. While analysts might evaluate the euro-dollar exchange rate over the next year or the previous year using a variety of different calculations, there are three major categories of forecasting techniques: fundamental forecasting, technical forecasting, and market-based forecasting. Each forecasting technique incorporates different models using different variables for divergent purposes and resulting in a slightly different perspective on the international exchange rate. In order to most accurately predict how the euro will fare against the U.S. dollar, all three techniques can be examined to determine their outputs over a period of time, before a final investment recommendation is proposed.
As a composite currency, the euro was first introduced in 1999 as an attempt to increase the European Union’s economic competitiveness with the U.S (Jamaleh, 2002). As a result, it was highly unexpected when the euro began to depreciate rapidly against the U.S. dollar shortly after its launch (Jamaleh, 2002). Specifically, the euro dropped 27 percent in the year and nine months after it was introduced (Jamaleh, 2002, p. 422). Many analysts began to wonder whether or not the euro was destined to plummet so sharply in comparison to the dollar because of fundamental differences in gross domestic product (GDP) growth and short-term interest rate differentials (Jamaleh, 2002). Others wondered if the euro was not being deliberately undervalued, so as to further increase the U.S. economic dominance (Jamaleh, 2002).
As a result, there are numerous models, based on the different types of forecasting. For example, long-run equilibrium models are based on fundamental forecasting (Jamaleh, 2002). The unobserved components model is another way to forecast the exchange rate based on a permanent equilibrium exchange rate (Chen & MacDonald, 2015). Market-based models include the purchasing power parity-based model, which incorporates U.S. export and import price indexes to determine the dollar’s value in comparison to the euro, which is market-based forecasting (Grossman & Simpson, 2011). An example of technical forecasting is the varying parameters model using technical features of a time series (Osinska & Matuszewska, 2006). Another example is the alternative novel neural network architectures that include three different model designs explored with a moving average convergence/divergence model (Duns, Laws & Sermpinis, 2008). While numerous models abound, all with substantial predictive abilities, Jamaleh’s (2002) models provide the most conclusive results with the simplest variables and terms, economic fundamentals that can be easily interpreted and applied to the market.
Specifically, Jamaleh (2002) sought to examine whether a linear error correction model might produce reliable forecasts of the euro-dollar exchange rate. They proposed that short interest rate differentials, GDP growth differentials, and inflation rate differentials were the primary drivers of the euro-dollar exchange rate (Jamaleh, 2002). They conceded that this particular model fails to account for outliers and abnormalities present in the euro-dollar exchange rate, and as a result it is expected that non-linear model would better incorporate these temporary deviations (Jamaleh, 2002). This non-linearity was assumed to be the source of a plausible alternative threshold regression model that would render better in-sample fitting and out-of-sample forecast performances than the linear model (Jamaleh, 2002). Their research also assumed that the deliberate undervaluation of the euro as a result of monetary policy intervention increased the risk of inflation stability conditions, as well as showed vulnerability to unfavorable GDP growth differentials (Jamaleh, 2002). However, despite this dynamic between the euro and the dollar, any positive stock market performance could have the effect of improving the euro-dollar exchange rate regardless of other macroeconomic features (Jamaleh, 2002).
In order to determine the potential gains from an international investment portfolio, an examination of three different forecasting techniques will be conducted to determine if the investment is financially profitable. It is proposed that the euro-dollar exchange rate will remain steady over the next year, regardless of forecasting technique, with the dollar making only slight increases less than a percentage point. To test this hypothesis, each of the forecasting techniques will be explained in detail and its calculations compared to illustrate its impact on predicting value.
In the simplest terms, fundamental forecasting is the process of using a prediction framework based on economic fundamentals to illustrate trend developments in the exchange rate between two currencies within a set macroeconomic scenario (Jamaleh, 2002). Fundamental forecasting is dependent on long-term equilibrium being a natural and normal feature of the exchange rate and its various determinants (Jamaleh, 2002). These determinants include interest rates, GDP growth, and the inflation rate, but other models might track additional variables used in the financial market (Jamaleh, 2002). In this way, fundamental forecasting is the product of a forward looking approach with expectations of performance, rather than actual performance (Jamaleh, 2002). As such, the following equation was proposed for calculating the euro-dollar exchange rate:
Figure 1. Table or image redacted in preview, but included in download.
“where EURDOLt is the euro/dollar exchange rate, SRt is the short interest rate differential, GDPt is the expected GDP growth differential and CPIt-1 is the inflation rate differential one-time lagged, and ut is a disturbance term. All differentials are between the Euro Area variable and the correspondent US variable, hence the short rate differential is the difference between the Euro Area interest rates and the US interest rates, and so on…According to economic theory, the coefficients a1 and a2 should have a positive sign, while a3 should be negative” (Jamaleh, 2002, p. 425).
From their first round of long-run estimates, the researcher was then able to make some statements about the relationship between the economic fundamentals and their deviations (Jamaleh, 2002, p. 425). Specifically, they proposed that:
" (i) higher short rates, as opposed to long rates, which incorporate an inflation premium, should favor a currency, since they make financial investments attractive in relative terms, through a liquidity effect… (ii) higher GDP growth, as far as it hints at solid economic conditions of the business cycle, should benefit a currency by attracting flows of business investments; while (iii) higher inflation should negatively affect a currency, by discouraging both financial and business investments, due to a lack of confidence with respect to the healthy state of economic fundamentals” (Jamaleh, 2002, p. 425; Johansen, 1991; Johansen, 1995; MacDonald, 1994).
Jamaleh (2002) performed the calculations using data from 1999 to 2000. Their calculations were compiled into a table as estimation results (Table 1).
Table 1. Table or image redacted in preview, but included in download.
Then those coefficients and t-values were used to conduct a second univariate ECM using the following equation:
Figure 2. Table or image redacted in preview, but included in download.
In their second round of calculations, they also compiled a second table of estimation results (Table 2).
Table 2. Table or image redacted in preview, but included in download.
They concluded that the best forecasters of exchange rate dynamics would need to take into consideration economic fundamentals and long-term deviations from the equilibrium level (p. 426). While their proposed model was able to illustrate whether the exchange rate increased or decreased, it could not account for large variations; despite this typical feature of linear models, it correctly captured the direction of the exchange rate (Jamaleh, 2002).
However, non-linearity becomes apparent through this inability to capture larger deviations along with a variable ECM term, which prompted a third equation that would formally test linearity in order to increase the reliability of their conclusions:
Figure 3. Table or image redacted in preview, but included in download.
“where wt-d is the threshold variable (which could be a lagged endogenous variable or an exogenous variable), d is the lag parameter, s1 is the threshold value. In this model [yt] is an observable output and [xt], [zt], . . . are observable inputs. [σjεt, j=1, 2] are sequences of heterogeneous strict white noises, with 0 mean and 1 variance, each independent from the input sequences [xt], [zt], . . .. These two error sequences are also independent from each other” (Jamaleh, 2002, p. 430).
The benefit of this third equation arises from its inherent non-linearity across time, and its local linearity across space through the addition of a threshold variable (Jamaleh, 2002). The threshold value makes it possible to track the breakpoints between the two regimes, or scenarios, and predicting within each moment in time what the euro-dollar exchange rate might be.
Going even further, the research questioned the influence the stock market exerted over the exchange rate, especially in light of international equity markets. For example, they discussed the correlation between Nasdaq’s monthly percentage variations and the euro-dollar exchange rate (Jamaleh, 2002). They found a much stronger correlation with Nasdaq than the S&P or the Dow Jones indexes, which made sense in light of rises in Nasdaq stocks coinciding with decreases in the euro’s value to investors seeking to generate capital flows (Jamaleh, 2002). Here, their research led them to pinpoint an average breakpoint around three percent, indicating both regimes were the location of losses and gains, or gains only, prompting further investigation into a final model.
The fourth model incorporated six non-linear threshold models that included the same explanatory variables already utilized in the linear model equations.
Figure 4. Table or image redacted in preview, but included in download.
For each model, they further specify that “the threshold variable wt-d is respectively (i) in model (1) ΔEURDOLt-1, (ii) in model (2) ECMt-1, (iii) in model (3) ΔSRt, (iv) in model (4) ΔGDPt, (v) in model (5) ΔCPIt, and (vi) in model (6) ΔNASt” (p. 433). They compiled these estimation results (Table 3 and Table 4).
Table 3. Table or image redacted in preview, but included in download.
Table 4. Table or image redacted in preview, but included in download.
Their estimates revealed that macroeconomic conditions in one regime or scenario might impact the other scenario in a completely different way (Jamaleh, 2002). One determinant for an exchange-rate deviation might have no significance in one regime, or be highly significant in the other, or it may be significant to a lesser or greater degree (Jamaleh, 2002).
After performing several rounds of calculations and estimations, their research delineated outcomes for each of their six proposed threshold models accounting for the differences in fundamental, technical, and market-based forecasting. With regard to the first model, they stated that “if at the previous time the exchange rate posted negative or small positive changes, the short rate differential has a relevant weight, which may reject the potentially active role of monetary policy in defending a currency or, which is more consistent with the euro case, it rejects the major role of policy rates when euro weakness poses some threat to price stability. Instead, when some currency strengthening occurred at t-1, the autoregressive component and GDP growth differentials become relevant. However, given the same magnitude of the AR (1) and AR (2) coefficients, in the presence of two similar positive variations in a row, the prevailing role is played by GDP differentials. If, at time t-2, the exchange rate depreciated, this should instead strengthen at time t, unless GDP spreads are unfavorable” (p. 434).
With regard to the second model, they suggested that “if the euro is highly undervalued with respect to its long-run fundamental equilibrium level (i.e., the ECM term is negative), only the short- rate differential may play a significant role. This conforms the importance of an active monetary policy when the euro is weaker with respect to some fundamental equilibria, therefore threatening price stability, not when it is weak per se… When the undervaluation degree is small, or the euro is correctly priced or overvalued, again a potentially re-equilibrating role is played by the autoregressive components and by the relative business cycle conditions” (pp. 434-435).
With regard to the third model, they concluded that “unfavorable interest rate developments are not relevant, unless the euro has a GDP growth disadvantage, because the only significant macro-variable is the GDP spread in this regime. When rates are favorable, these are highly important, hence potentially making a ‘strength position’ of the currency to persist if this was in place at t-1” (p. 435). In regard to the fourth model, they stated “when economic growth expectations worsen, the only significant variable is the AR (1) component, thus describing the possibility that, if some currency weakness is present, this might persist, unless expectations improve…. if…a depreciation occurred which made the euro undervalued, this disequilibrium may be corrected” (p. 435).
In regard to the fifth model, Jamaleh (2002) found “when inflation conditions are favorable, the relevance of the GDP term means that if this goes together with healthy business cycle conditions then the euro can appreciate; if, on the contrary, lower inflation is an indication of worse GDP expectations, the euro could instead weaken” (pp. 436-437). They further specified that while GDP is important, if there is also high inflation, then the euro-dollar exchange rate will not be favorable to the euro (Jamaleh, 2002). At the same time, the interest rate spread is also indicative of investment potential because of its connection with speculative financial opportunities, or investors borrowing at the lower interest rates and investing once the currency reaches a higher yield, which will only serve to further lower the currency, decrease GDP growth, and increase inflation (Jamaleh, 2002). This is why monetary policy may be another factor in that it works to keep inflation down and improve GDP growth potential, which may favor the euro (Jamaleh, 2002).
After conducting extensive research and modeling multiple equations, Jamaleh (2002) finalized that GDP growth was most correlated with euro overvaluation; if the euro appreciated, then decreasing GDP growth was strongly correlated. After GDP growth, inflation and short rate interest rates were found to correlate with a weak euro, contrary to intuitive expectations (Jamaleh, 2002). Additionally, non-linear models created the most fitted graphs when compared to the linear models (Jamaleh, 2002). When analyzing the predictive performance of the six models, it was clear that non-linear models illustrated direction-of-change with much greater precision, even after several performance tests (Jamaleh, 2002). Non-linear models were most reliant on fundamental forecasting, especially when predicting long-run equilibrium of the euro-dollar exchange rate (Jamaleh, 2002).
Fundamental, technical, and market-based forecasting attempt to predict the performance of particular currencies against each other to determine the soundness of international market investments. Seeking to invest in Germany, the euro has been compared to the dollar. The research found that when GDP growth or Nasdaq performance is positive or normal in the U.S., the euro will be devaluated in comparison to the dollar. On the other hand, when inflation is controlled with short interest rates, the euro’s performance against the dollar stabilizes. Despite these trends, there is also the potential for market disruptions to cause outliers and abnormal variations in the exchange rate across time. As a result, the models examined here work in conjunction to illustrate a linear model and a non-linear model with a threshold variable that captures the exchange rate dynamics in different regimes or scenarios. With that in mind, it is proposed that if GDP growth is high in the U.S., avoid investing in the euro; but if inflation is high and short interest rates are being manipulated by monetary policy, then investment in the euro is encouraged, but may lead to decreases in the near future.
Chen, X. & Macdonald, R. (2015). Measuring the dollar-euro permanent equilibrium exchange rate using the unobserved components model. Journal of International Money and Finance, 53: 20-25. DOI: http://dx.doi.org/10.1016/j.jimonfin.2014.12.008.
Dunis, C. L., Laws, J., & Sermpinis, G. (2009). The robustness of neural networks for modeling and trading the EUR/USD exchange rate at the ECB fixing. Journal of Derivatives & Hedge Funds, 15(3): 186-205.
Grossman, A. & Simpson, M. W. (2011). Predictability of the U.S. dollar index using a U.S. export and import price index-based relative PPP model. Journal of Economics and Finance, 35: 417-433. DOI 10.1007/s12197-009-9102-6.
Jamaleh, A. (2002). Explaining and forecasting the euro/dollar exchange rate through a non-linear threshold model. The European Journal of Finance, 8(4): 422-448. DOI: 10.1080/13518470210167301.
Johansen, S. (1991) Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models, Econometrica, 59, 1551–1580.
Johansen, S. (1995) Likelihood-based Inference in Cointegrated Vector Autoregressive Models. Oxford University Press.
MacDonald, R. and Taylor, M.P. (1994) The monetary model of the exchange rate: long-run relationships, short-run dynamics and how to beat a random walk, Journal of International Money and Finance, 13(3), 276–290.
Osinska, M. & Matuszewska, A. (2006). Detecting some dynamic properties of the euro/dollar exchange rate. International Advances in Economic Research 12: 327-341. DOI: 10.1007/s11294-006-9021-7.