Why is it apparently so difficult to forecast exchange rate movements?
Introduction
The forecasting of exchange rate movements is one of the most intensively studied topics in the area of economics. The importance of exchange rates is visible in the most recent time with the floating of the Egyptian Pound, after it was previously pegged to the US-Dollar – at a 32.5 % discount rate to the previously official price. Nowadays it is down further 40 % to 19 Pound on the Dollar. Being able to predict this is helping companies with their hedging needs or for example Asset Managers when they are considering investment decisions. There are several models predicting exchange rates, mainly focussing on quantitative variables. In the following paper I am planning to identify the shortcomings of the major exchange rate forecast models which there are Monetary Model, Dornbusch-Model and Mundell-Fleming Model. I will start with a discussion of the crucial assumptions and specifications within these models, I will continue with an investigation of the problems arisen through the misspecification of the models followed by a chapter over the variables which are not included in the models. I will finish this essay with a conclusion.
Problems with assumptions and specifications of Exchange Rate Models
All three basis models make use of assumptions which cannot be justified in the long-run. While the Dornbusch-Model is a mix of the Mundell-Fleming Model (long-term) and the Monetary Model (short-term), based on the expectations of labour supply there are essentially three assumptions which prove to be essential for the models.
Perfect Capital Markets
Although no model requires perfect capital markets in the assumptions directly, the assumptions merely expect that there are no market imperfections. Indeed there is the assumption that capital mobility is less than perfect in the Mundell-Fleming model, however there are other factors to regard as well. Here I will mainly regard UIRP – with interest rates as a main factor influencing the movement of exchange rates and the existence of heterogenous expectations. However PPP shortcomings, related to exchange rate forecasts can be investigated in short as well. PPP is known to have a half-life time of 3 to 5 years, which is consistent with short-term exchange rate movements which seem to be free from influence through fundamentals. At the same time the exchange rate seems to reflect long-term changes in Purchasing Power. Thus models in the long run are to a certain extend stabilised by the PPP assumption.
UIRP holds at all times
UIRP basically states that interest rates should be relatively higher for countries where the exchange rate is expected to depreciate over the upcoming period of time. However, there is evidence, that this is not the case. It rather is vice versa. If a country’s interest rates are expected to increase over a period of time, so is the currency expected to appreciate over the same time horizon. This is rejecting not only that the interest rates are working as compensation for a possibly depreciating currency but that it is actually the exchange rate which is following the expected relative movement of interest rates. With this broken assumption the Dornbusch model will not work and thus won’t be effective in predicting exchange rates.
Interest rate forecasts seem to be a major influencer of currency forecasts – in a study it was found that good exchange rate forecasters are as well good interest rate forecasters.
Market Participants have rational expectations
All three models expect that currency prices are based on rational expectations, as is the case with many macroeconomic models nowadays. Although information is available to the majority of the market participants at the same time the traders are using the fundamentals in heterogenous ways which is even changing over time. Furthermore participants are using as well non-fundamental data for their decision making. This will change the overall expectation of the exchange rate. At the same time it is leading to the conclusion, which is proven by economical research, that exchange-rate forecasts are not formed rationally. This is due to the inclusion of predictable errors in forecasts. These forecast errors are created through specific behaviour among market participants, as they have biased expectations.
Problems created by misspecification of the models
Expectations and usage of past events
All models are using data of past events, such as money supply or real income (monetary model), which are collected from the past. Furthermore the Monetary Model is assuming PPP to hold at all times, this is essentially using a specific price level, usually indexed. However financial prices usually follow a Markov stochastic process and thus are independent of there past movements. Thus they are not necessarily related to economic variables of the past.
Linear models relevance in a complex world
The models used to determine possible exchange rates are not very complex, they try to focus on a small set of variables. However this is largely inappropriate in a complex, globalised world. Thus models “cannot account well enough for existing complex relations”.
Variables not mentioned in the Models
Models only use Macroeconomic Variables
The usage of pure macroeconomic, quantitative factors is more easy to use however there is evidence that there are further factors influencing the exchange rate directly or indirectly. The two major factors are political of nature and of social nature, such as unemployment. Shortly mentioned is also the self-fulfilling predictions occurrence in exchange rate markets as an influencing factor on exchange rate movements.
Politics, Military and Psychological Reasons need to be integrated
Factors which are influencing the exchange rate are not only of numerical nature, as they are shown in the three models. There is also influence on the exchange rate through political decisions, military or even market psychology. Further it needs to be mentioned that there are opportunities to profit from exchange rate movements through technical analysis. This however is strictly against the rational market expectation, as a rational market cannot be expected to trade without using any fundamentals purely based on charts. “90 % of participants […] rely on technical strategies.”
Unemployment as a Social Factor indirectly influencing the exchange rates
Central Banks are one of the major players in foreign exchange markets. Thus they are influencing to an extend the exchange rate movements. However they also have as one of there main targets is to keep unemployment at low levels. Thus unemployment is influencing the monetary politic, such as interest rate politic, of the central bank, indirectly affecting the exchange rate. Furthermore the effect of a depreciating currency, due to lower, not higher, interest rates, paired with unemployment is strengthened with weakening foreign direct investment, which is again happening due to lower growth perspectives and the high unemployment rate.
Further influencing factors of currency usage
All models are expecting currencies to be held purely by domestic investors. This is not the case. First bigger currencies, such as the USD, EUR, GBP are often used as reserve currencies, e.g. to strengthen the own currency – or are bought in order to devalue the currency. This is not included in the models, however with over 100 countries possessing Dollar reserves this will be quite a big amount of the actual cash flow. Furthermore the models are not taking into account private reasons for currency movements, such as studies or holiday trips. Although they will only account for a minor part, still they are providing long-term benefits to the economy, with money long-term spent within the domestic economy.
Conclusion
The exchange rate models are failing to forecast due to a variety of reasons. These mistakes can be found in the model assumptions with neither PPP nor UIRP holding to the extend needed by the models. Also it can be seen that the real world is much more complex as it could be modelled accurately. However most severe are the missing external variables from politics and unemployment as well as the wrong assumption about the connection between interest rate expectations and the exchange rate expectations.
References:
Afat, Dincer; Gómez-Puig, Marta; Sosvilla-Rivero, Simón, 2015: “The failure of the monetary model of exchange rate determination” (in: Applied Economics, Vol. 47)
Chang, P. H. Kevin; Osler, Carol L., 1999: “Methodical Madness: Technical Analysis and the Irrationality of Exchange-Rate Forecasts” (in: The Economic Journal, 109 (October))
Copeland, Laurence, 2014: “Exchange Rates and International Finance” – Sixth Edition
Dick, Christian D.; MacDonald, Ronald; Menkhoff, Lukas, 2015: “Exchange rate forecasts and expected fundamentals” (in: Journal of International Money and Finance 53 (2015)
Hull, John C. 2012: “Options, Futures And Other Derivatives” – Eighth Edition
Reuters, Nov. 4th 2016: “Egyptian Central Bank Floats Pound” http://www.arabnews.com/node/1006056/business-economy (Date Last Opened: March 15th 2017)
The Economist, October 30th, 1999: “Better than guesswork?”
Wang, Zhi-Bin; Hao, Hong-Wei; Yin, Xu-Cheng; Liu, Qian; Huang, Kaizhu, 2010: “Exchange rate prediction with non-numerical information” (in: Neural Computing and Applications, 2011)