Valutakursprognostisering med multivariata felkorrigeringsmodeller: En prognosstudie av nio valutapar
(2020) STAK11 20172Department of Statistics
- Abstract
- Ever since the formation of the modern global foreign exchange market in the early 1970s, scholars have sought to develop models that can explain movements in foreign exchange rates. Arguably one of the most popular approaches to exchange rate modelling is the monetary approach, where movements in the exchange rate are modelled through changes in macroeconomic variables affecting the relative money demand for a currency. Over the years, authors have tried to validate the monetary models on real data by using different econometric methods. The varying results from these studies have divided scholars into two groups – those who believe that the monetary approach to exchange rate modelling is flawed and those who believe that more dynamic... (More)
- Ever since the formation of the modern global foreign exchange market in the early 1970s, scholars have sought to develop models that can explain movements in foreign exchange rates. Arguably one of the most popular approaches to exchange rate modelling is the monetary approach, where movements in the exchange rate are modelled through changes in macroeconomic variables affecting the relative money demand for a currency. Over the years, authors have tried to validate the monetary models on real data by using different econometric methods. The varying results from these studies have divided scholars into two groups – those who believe that the monetary approach to exchange rate modelling is flawed and those who believe that more dynamic econometric models can validate the monetary approach. In this thesis, I have sought to investigate the latter by forecasting the exchange rate using econometric models that capture both the long- and short-run dynamics of the monetary models. To be able to draw more general conclusions, I have included four monetary models and nine of the most liquid currency pairs in my study. The forecasts will be evaluated against the performance of both a random walk and a random walk with drift on two measures: forecast accuracy and directional accuracy. Many previous studies do not statistically prove their results, the use of statistical significance tests is, therefore, an important part of this thesis. In earlier studies, the authors have either used the monthly exchange rate defined as the end-of-month closing price or the arithmetic average of daily closing prices over the month. To investigate if the measurement of the exchange rate affects the results, I have used both definitions of the exchange rate as the dependent variable in my econometric models. When evaluating the results in the out-of-sample period, forecasts from the monetary models based on the monthly average show better forecast accuracy and directional accuracy than the random walk models. Whilst forecasts based on the end-of-month value of the exchange rate show worse forecast accuracy and only equal directional accuracy compared to the random walk models. However, when subjected to statistical significance tests, most of the performance differences between the monetary models and the random walk models cannot be sustained, regardless of the choice of exchange rate measurement. In conclusion, my findings support the notion that the monetary approach to exchange rate modelling cannot be validated using econometric models that capture both the long-run and short-run dynamics of the monetary models. The choice of foreign exchange rate measurement does seem to affect the results, albeit only marginally when statistical significance tests are employed. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9017438
- author
- Wikrén, Björn LU
- supervisor
- organization
- alternative title
- Exchange Rate Forecasting with Vector Error Correction Models: A Forecast Study of Nine Major Currency Pairs.
- course
- STAK11 20172
- year
- 2020
- type
- M2 - Bachelor Degree
- subject
- keywords
- valutakurser, valutakursmått, monetära modeller, prognostisering, prediktion, kointegration, multivariata felkorrigeringsmodeller, prognosprecision, rörelseriktningsprecision
- language
- Swedish
- id
- 9017438
- date added to LUP
- 2020-06-22 11:08:01
- date last changed
- 2020-06-22 11:08:01
@misc{9017438, abstract = {{Ever since the formation of the modern global foreign exchange market in the early 1970s, scholars have sought to develop models that can explain movements in foreign exchange rates. Arguably one of the most popular approaches to exchange rate modelling is the monetary approach, where movements in the exchange rate are modelled through changes in macroeconomic variables affecting the relative money demand for a currency. Over the years, authors have tried to validate the monetary models on real data by using different econometric methods. The varying results from these studies have divided scholars into two groups – those who believe that the monetary approach to exchange rate modelling is flawed and those who believe that more dynamic econometric models can validate the monetary approach. In this thesis, I have sought to investigate the latter by forecasting the exchange rate using econometric models that capture both the long- and short-run dynamics of the monetary models. To be able to draw more general conclusions, I have included four monetary models and nine of the most liquid currency pairs in my study. The forecasts will be evaluated against the performance of both a random walk and a random walk with drift on two measures: forecast accuracy and directional accuracy. Many previous studies do not statistically prove their results, the use of statistical significance tests is, therefore, an important part of this thesis. In earlier studies, the authors have either used the monthly exchange rate defined as the end-of-month closing price or the arithmetic average of daily closing prices over the month. To investigate if the measurement of the exchange rate affects the results, I have used both definitions of the exchange rate as the dependent variable in my econometric models. When evaluating the results in the out-of-sample period, forecasts from the monetary models based on the monthly average show better forecast accuracy and directional accuracy than the random walk models. Whilst forecasts based on the end-of-month value of the exchange rate show worse forecast accuracy and only equal directional accuracy compared to the random walk models. However, when subjected to statistical significance tests, most of the performance differences between the monetary models and the random walk models cannot be sustained, regardless of the choice of exchange rate measurement. In conclusion, my findings support the notion that the monetary approach to exchange rate modelling cannot be validated using econometric models that capture both the long-run and short-run dynamics of the monetary models. The choice of foreign exchange rate measurement does seem to affect the results, albeit only marginally when statistical significance tests are employed.}}, author = {{Wikrén, Björn}}, language = {{swe}}, note = {{Student Paper}}, title = {{Valutakursprognostisering med multivariata felkorrigeringsmodeller: En prognosstudie av nio valutapar}}, year = {{2020}}, }