Date of Degree


Document Type


Degree Name





Merih Uctum

Committee Members

Thom Thurston

Temisan Agbeyegbe

Chun Wang

Subject Categories

Econometrics | Finance | International Economics | Macroeconomics


State Space Representation Model, Bayesian, Econometrics, Macroeconomics, International Finance


This dissertation develops three new econometric models using Bayesian state space representation model in order to apply to macroeconomics and international finance. It consists of 3 chapters. Chapter 1 develops a Markov mixture model of macroeconomic fundamentals to analyze the short-run dynamics of foreign exchange rates. In our model, foreign exchange rates are simultaneously determined by three parities: the interest rate parity, the inflation rate parity, and the equity return rate parity. Using four exchange rates: the U.S. dollar price of the British pound, the German mark, the Japanese yen, and the Canadian dollar, the findings are: (1) Our model better explains the short-run dynamics of the exchange rates than a random walk with drift does. (2) The equity return is an important factor in determining the exchange rates. In particular, a higher return in a home-country equity market leads to depreciation of the home currency. (3) Our model can explain the short-run movement of the exchange rates that may be caused by rare events, such as the Plaza Agreement, the Asian and Russian financial crises, and the global financial crisis of 2008-2010. (4) There is weak evidence that our model outperforms a driftless random walk in terms of out-of-sample forecast errors for each currency on average over a 144-month period. Chapter 2 explores an econometric model of cross-country monetary transmission mechanism. We particularly examine the effect of the U.S. monetary policy shock on the other Group of Seven (G-7) countries, and develop a panel version of the factor-augmented vector autoregressive (FAVAR) model, allowing both common and country-specific unobservable factors. We also allow interdependency among the country-specific factors to produce the comovement of business cycles in the G-7 countries. The findings are that: (1) the existence of such interdependency is statistically significant. (2) This interdependency reduces the magnitude of U.S. business cycle. (3) After the contractionary U.S. monetary shock, Japan's output is affected most, followed by the Anglo-Saxon countries, the UK and Canada. The continental European countries, Germany, France, and Italy, are the least affected among them. (4) The results are consistent with international consumption risk sharing. (5) Our 2-country version of dynamic stochastic general equilibrium (DSGE) model shows the evidence of such risk sharing, and also suggests that the comovement of business cycles might be as a result of foreign central bank reactions to the change in home monetary policy and the exchange rate. Chapter 3 proposes to examine a fiscal policy reaction function by remedying 2 issues in existing literature. We allow the fiscal policy reaction to vary over time and take into account endogeneities by casting the analysis in a multivariate framework. Our findings are: (1) the response of the government's primary surplus to a 1-percent exogenous debt shock displays first a deterioration of the primary balance in 1981Q1-1991Q4 and 2001Q1-2008Q1. It followed by a sustained improvement, consistent with the definition of a fiscally responsible policy. (2) However, there are the periods when there is a sustained improvement from beginning without any deteriorations of the primary balance, that is in 1992Q1-2000Q4 and 2008Q2- 2013Q1. (3) These differences in the fiscal policy reactions might be due to recession and wars. If some emergency expenditures are needed during recessions or wars, it would be harder to make fiscal policy more sustainable. (4) The effect on output gap becomes more negative during the resession along with no immediate reaction in fiscal policy to reduce the level of debt in response to an exogenous debt shock. (5) The robustness test shows that our time-varying coefficient model is more superior to the standard vector autoregressive (VAR) model in terms of marginal density of data.