Testing for Nonlinear Threshold Cointegration in the Monetary Model of Exchange Rates with a Century of Data

The monetary model suggests that nominal exchange rates between two countries will be determined by important macroeconomic variables. The existence of a cointegrating relationship among these fundamental variables is the backbone of the monetary model. In a recent paper, Rapach and Wohar (2002, Journal of International Economics) advance the literature by testing for linear cointegration in the monetary model using a century of data to increase power. They find evidence of cointegration in five or six of ten countries. We extend their work to the nonlinear framework by performing threshold cointegration tests that allow for asymmetric adjustments in two regimes. Asymmetric adjustments in exchange rates can occur, for example, if transactions costs are present or if policy makers react asymmetrically to changing fundamentals. Moreover, whereas Rapach and Wohar (2002) found it necessary to exclude the relative output variable in some cases to maintain the validity of their cointegration tests, we can include this variable as a stationary covariate to increase power. Overall, using their same long-span data, we find more support for cointegration in a nonlinear framework. 환율 결정 모형의 근간이 되는 이론으로 널 리 알려져 온 화폐모형은 두 국가 간의 환율 이 각국의 통화량과 소득 수준에 의해 결정된 다고 설명하고 있다. 그러나 이 이론이 성립 하려면 이 모형에 내포된 변수 간에 공적분이 성립해야 하는데, Rapach and Wohar(2002) 의 논문은 10개 국가의 자료 중 대 여섯개 의 자료에만 (선형) 공적분이 존재한다는 결과를 제시하였다. 본 논문은 그들이 사용 한 100년간에 걸친 자료를 사용하되, 환율 결정과정에서 발생할 수 있는 비대칭적 조정과정을 감안하여 비선형 공적분이 성립하는가를 검증하였다. 또한 독립변수 가 불안정적이 아닐 경우에는 공적분 관 계를 설정하기 곤란하다는 이유로 누락시 키는 경우가 많은데 본 논문에서 사용되 는 방법론에서는 그러한 문제가 제기되지 않는다. 본 논문에서는 선형 공적분 검정 결과에 비해 더 많은 경우에 있어서 비선 형 공적분 관계가 있다는 검정 결과가 산 출되었다. Testing for Nonlinear Threshold Cointegration in the Monetary Model of Exchange Rates with a century of Data 3


ABSTRACT
The monetary model suggests that nominal exchange rates between two countries will be determined by important macroeconomic variables. The existence of a cointegrating relationship among these fundamental variables is the backbone of the monetary model. In a recent paper, Rapach and Wohar (2002, Journal of International Economics) advance the literature by testing for linear cointegration in the monetary model using a century of data to increase power. They find evidence of cointegration in five or six of ten countries. We extend their work to the nonlinear framework by performing threshold cointegration tests that allow for asymmetric adjustments in two regimes. Asymmetric adjustments in exchange rates can occur, for example, if transactions costs are present or if policy makers react asymmetrically to changing fundamentals. Moreover, whereas Rapach and Wohar (2002) found it necessary to exclude the relative output variable in some cases to maintain the validity of their cointegration tests, we can include this variable as a stationary covariate to increase power. Overall, using their same long-span data, we find more support for cointegration in a nonlinear framework.

I . Introduction
The monetary model suggests that nominal exchange rates between two countries will be determined by important macroeconomic fundamentals. Two early references to the model are Mussa (1976) and Bilson (1978). While the monetary model is intuitively appealing, empirical support for the model is often difficult to find. Perhaps most critical in this regard are the findings in Meese and Rogoff (1983), where the authors obtain better forecasts of nominal exchange rates in a simple random walk as compared to the monetary model. If the monetary model is valid and the fundamental variables are nonstationary, then a cointegrating relationship must exist. Many empirical studies, however, fail to find support for (linear) cointegration in the monetary model (e.g., Meese, 1986, Baillie and Selover, 1987, and Sarantis, 1994. More recently, Rapach and Wohar (2002, RW) advance the literature by performing (linear) cointegration tests of the monetary model using a century of data. By using long-span data to increase power, RW find greater support for the monetary model than in many previous tests and find evidence of cointegration in 5 or 6 of 10 countries. 1 In this paper, we re-examine the long-span data in RW and perform nonlinear threshold cointegration tests. If the underlying model is nonlinear and linear cointegration tests are adopted, then lower power can result. As such, it is possible that greater support for cointegration will be found when adopting nonlinear tests. In this regard, a growing number of recent studies document evidence of nonlinear dynamics in exchange rates (e.g., Taylor and Peel, 2000, Guerra, 2001, Kilian and Taylor, 2003. Nonlinear dynamics in exchange rate might arise, for example, if reaction to fundamentals and adjustment depends on the magnitude or sign of the deviation from the equilibrium. For instance, Taylor and Peel (2000) find evidence that deviations in exchange rates from the monetary model follow a nonlinear adjustment process. Although Taylor and Peel (2000) note that a tractable way to model nonlinear adjustment is to adopt a threshold model, they adopt an exponential smooth transition autoregressive (ESTAR) model perhaps for convenience of estimation. While these and other recent papers find greater support for the monetary model in nonlinear models, these papers do not provide formal tests for nonlinear cointegration. Analogous to the linear case, if the variables in a nonlinear monetary model are nonstationary and not cointegrated, then spurious estimates can result. It remains to be seen whether nonlinear cointegration holds or not, but this important question was not examined in the 1 Rapach and Wohar (2002) initially consider fourteen countries, but some of the countries contain a mix of I(0) and I (1) variables that cannot be cointegrated, and in one country, The Netherlands, all of the variables in the model are I(0) so cointegration tests are not performed for four of these countries. However, in this paper, we utilize I(0) regressors in our testing scheme rather than discarding them as we shall see more details shortly. Thus, our procedure permits us to overcome a limitation of Rapach and Wohar (2002) in this regard. above papers.
To perform our empirical tests, we first consider the ordinary least squares based autoregressive distributed lag (ADL-OLS) threshold cointegration test developed by Li and Lee (2008). We utilize two different threshold effects hypothesized to arise from asymmetric policy responses and/or transactions costs. In particular, we consider threshold models where adjustment to the long-run equilibrium can depend on the level or change in the deviations from the long-run equilibrium. Moreover, in some countries, the nominal exchange rate and the relative money supply series are each I(1) while the deviation in output series is I(0). While RW (2002) omit the output deviation variable in these cases, we want to include this variable as a stationary covariate in our cointegration tests to increase power. In these cases, we utilize the instrumental variables based autoregressive distributed lag (ADL-IV) threshold cointegration test as suggested in Enders, Im, Lee and Strazicich (2009). The ADL-IV threshold cointegration test is well suited to this task, since the test statistics are unaffected by including stationary covariates. 2 Our data set is the same as in RW and consists of over 100 years of annual data on nominal exchange rates (foreign currency per U.S. dollar), national money supplies relative to the U.S. money supply, and real GDPs relative to the U.S. real GDP for fourteen industrialized countries. 3 The nominal exchange rate series come from Taylor (2001). The money supply and real GDP data come from Bordo and Jonung (1998) and Bordo, Bergman, and Jonung (1998), respectively. The specific sample periods for each country are reported in our Tables below. Using a long-span data set has the distinct advantage of potentially more observations in each regime and greater power in inference tests. Overall, we find greater support for cointegration in a nonlinear framework as compared to the linear tests. Combining results, we reject the null of no cointegration (in at least one regime) in 8 of the 10 countries examined. These findings provide new support to the growing number of papers by Taylor and Peel (2000) and others who find more support for the monetary model in a nonlinear framework.
The remainder of the paper proceeds as follows. In Section 2, we briefly describe the monetary model and our test methodology. In Section 3, we discuss our empirical findings. We summarize and conclude and Section 4.

Ⅱ. Monetary Model and Testing for Threshold Cointegration
The monetary model can be described by: where e denotes the nominal exchange rate (foreign currency per unit of domestic currency), m* denotes the foreign money supply, m denotes the domestic money supply, y* denotes the foreign country output, y denotes the domestic country output, and t is a time subscript. The United States is the domestic country in each case and all variables are in natural logarithms. 4 If e t , (m t * − m t ), and (y t * − y t ) are each I(1), then the long-run equilibrium condition implies that these variables are cointegrated and v t = e t -β 0 -β 1 (m t * − m t ) -β 2 (y t * − y t ) will be a stationary process. 5 While the ADL-OLS threshold cointegration test can have greater power than the ADL-IV test, the ADL-OLS based test has nonstandard distributions that depend on the nuisance parameter when stationary covariates are included. In contrast, the ADL-IV threshold cointegration test is invariant to nuisance parameters in such cases. Therefore, in countries where y* -y, e, and m t * − m t are nonstationary, we will utilize the ADL-OLS threshold cointegration test. Then, in countries where y* − y is stationary, while e and m t * − m t are nonstationary, we will utilize the ADL-IV test.
The ADL-IV based test is well suited in this case, since the same standard normal critical values can be adopted with stationary covariates in the testing equation.
The nonlinear specification of the monetary model in the ADL threshold cointegration test can be described as follows: 6 Δe t = I t [ρ 1 e t -1 +a 1 (m t *−m t ) +a 2 (y t * − y t ) +b 1 Δ(m t * − m t ) +b 2 Δ(y t * − y t )] + (1−I t ) [ρ 2 e t -1 +c 1 (m t *−m t ) +c 2  (y t *−y t ) +d 1 Δ(m t *−m t ) +d 2 Δ(y t *−y t )] +u t . ( Lags of Δe t , Δ(m t * − m t ), and Δ(y t * − y t ) can be included as necessary to correct for serial correlations. There are clear advantages to using ADL models; see Li and Lee (2008), and Enders, Im, Lee, and Strazicich (2009) for more details. Following these methods, we consider two threshold indicators. The first is the so-called threshold autoregressive (TAR) model: where is the threshold value. The second threshold indicator is the so -called momentum threshold autoregressive (M-TAR) model: We test the following null hypothesis in each case: H o : ρ 1 = 0 and ρ 2 = 0 vs. H 1 : ρ 1 < 0 and/or ρ 2 < 0.
Thus, under the alternative hypothesis the deviation from the equilibrium will be stationary in at least one regime. We transform the threshold parameter into its percentile and determine this value by minimizing the sum of squared residuals. Specifically, since the threshold parameter cannot be greater or less than the maximum or minimum value of the threshold variable, we first sort the threshold variable e t-1 into e t-1 *, which takes the ordered values of e t-1 from the minimum to maximum value of e t-1 . Then, we consider the following transformation scheme: where c* = As a result, the threshold parameter τ is transformed into a percentile parameter c defined over the interval 0 and 1, and the asymptotic distribution of the corresponding threshold tests will depend only on the percentile parameter c. We can therefore provide critical values based on the percentile parameter defined on the interval between 0 and 1, rather than on a real value that can potentially vary over -∞ to +∞. We estimate the threshold percentile parameter by a grid search to find the value of e t-1 (or Δe t-1 ) that minimizes the sum of squared residuals from the regression. For the grid search procedure, we use each value of the sorted data of e t-1 (or Δe t-1 ) from the minimum to maximum value, while trimming values at the <Table 1> ADL-OLS Threshold Cointegration Test Results, I t = 1 if Δe t-1 ≥ τ and I t = 0 if Δe t-1 < τ Note: The Wald statistic tests the null hypothesis of no cointegration in two regimes (ρ1 = ρ2 =0). All models include a constant term without trend. Critical values come from Table 1 in Li and Lee (2008) for the Boswijk version of the ADL-OLS threshold cointegration test with n = 2 conditioning variables. The percentile threshold value was determined by minimizing the sum of squared residuals. *, **, and *** denote rejection of the null of no cointegration at the 10%, 5%, and 1% levels of significance, respectively. lower and upper 10% of the data. Chan (1993) showed that this type of procedure can estimate the threshold consistently under the null and alternative hypotheses. The threshold parameter estimator is super-consistent under the alternative, implying that the estimated value is expected to converge to its true parameter value more quickly under the alternative hypothesis than under the null.
In the ADL-OLS test, we utilize the Boswijk (1994) version of the Wald test to test the null hypothesis as recommended by Li and Lee (2008). The critical values come from Table 1 in Li and Lee (2008). We use critical values corresponding to each of the indicator functions defined in (3) and (4), respectively. In the ADL-IV test we utilize the usual t-statistics to test the significance of ρ 1 and ρ 2 , since these test statistics have standard distributions and are unaffected by including y*-y as a stationary covariate.

Ⅲ. Empirical Results
We now examine the results of testing for threshold cointegration. To be consistent in our comparisons to the linear tests in RW, we utilize their same unit root test results and the same long-span data. In the ADL-OLS tests, we determine the optimal number of lags in the testing regression by employing the Schwarz information criteria (SIC). In the ADL-IV tests, we jointly determine the optimal value of m to construct a proper IV (w t ) and the optimal number of lags to correct for serial correlations. We first search for the optimal lag for a given value of m, for m = 1 and maxm, where maxm is given as T 0.5 . We then determine the optimal value of m as the value that minimizes the residual sum of squares (RSS) from the regression using the optimal number of lags.

Asymmetric Momentum Threshold Effects
We first examine the ADL-OLS threshold cointegration test results with asymmetric effects modeled by the change in deviations from the equilibrium in the monetary model. This is the momentum threshold model, where the speed of adjustment to the equilibrium will depend on whether the change in the deviation is above or below the threshold level (I t = 1 if Δe t-1 ≥ τ and I t = 0 if Δe t-1 < τ). To obtain valid ADL-OLS test results, we will consider only threshold cointegration tests for the eight countries where e, m*-m, and y*-y were each identified as I(1) variables in RW. The test results are displayed in Table 1. Looking at the results, we observe that 6 of the 8 countries reject the null of no cointegration in at least one regime (Belgium, Canada, France, Italy, Spain, and the UK) at the 1% level of significance. Moreover, in each country, except Spain, the speed of adjustment to the monetary model equilibrium is fastest when the rate of depreciation is above the threshold level. Given that the threshold level is close to zero in each case, these findings suggest that nominal exchange rates adjust more quickly to the equilibrium predicted by the monetary model when they are depreciating rather than appreciating. For example, in Canada the estimated persistent parameter when the change in the deviation from the equilibrium is above the threshold level (in regime 1) is -0.315, which is clearly stationary. In contrast, the estimated persistent parameter when the change in the deviation from the equilibrium is below the threshold level (in regime 2) is -0.079, implying that nominal exchange rate behave as a random walk. While less extreme, the differences in the estimated persistent parameters are similar in four of the other five countries that reject the null of no cointegration (Belgium, France, Italy, and the UK). One possible explanation for these findings could be that policy makers are more likely to intervene in currency markets when their currency is depreciating than when their currency is variables should not lead to any serious bias.
appreciating. This is an example of policy response to different economic conditions; see also Lee (2006) for the case of Korea regarding fiscal policy response to economic cycles.

Asymmetric Deviation Threshold Effects
We next examine the ADL-OLS threshold cointegration test results with asymmetric threshold effects modeled by the level of the deviations from the equilibrium. This is the autoregressive threshold model, where the speed of adjustment to the equilibrium depends on whether the level of the deviation is above or below the threshold level (I t = 1 if e t-1 ≥ τ and I t = 0 if e t-1 < τ). Again, to obtain valid ADL-OLS test results we will consider only threshold cointegration tests for the eight countries where e, m*-m, and y*-y were each identified as I(1) variables in RW. The test results are displayed in Table 2. Looking at the results, we observe that 4 of the 8 countries reject the null of no cointegration in at least one regime (Canada, Italy, Switzerland, and the UK) at the 1% or 5% level of significance. In two of the four countries (Canada and Italy) that reject the null of no cointegration, the difference in the adjustment speeds is similar to that in the momentum models of Table 1. In Canada, the estimated persistent parameter when the deviation from the equilibrium is above the threshold level (in regime 1) is -0.710 while the estimated persistent parameter when the deviation is below the threshold level is -0.310. In Italy, the <Table 2> ADL-OLS Threshold Cointegration Test Results, I t = 1 if e t-1 ≥ τ and I t = 0 if e t-1 < τ Note: The Wald statistic tests the null hypothesis of no cointegration in two regimes (ρ1 = ρ2 =0). All models include a constant term without trend. Critical values come from Table 1 in Li and Lee (2008) for the Boswijk version of the ADL-OLS threshold cointegration test with n = 2 conditioning variables. The percentile threshold value was determined by minimizing the sum of squared residuals. *, **, and *** denote rejection of the null of no cointegration at the 10%, 5%, and 1% levels of significance, respectively. estimated persistent parameter when the deviation from the equilibrium is above the threshold level (in regime 1) is -0.441, while the estimated persistent parameter when the deviation is below the threshold level is -0.136. In the other two countries (Switzerland and the UK) that reject the null of no cointegration, the results are less clear. In Switzerland, the adjustment speeds are the same in each regime, while in the UK the speed of adjustment to the equilibrium is fastest when the deviation from the equilibrium is below the threshold level rather than above. Given the lack of a consistent pattern in the estimated threshold values and/or the persistent parameters in these four countries, it is more difficult to provide a general explanation using the levels of the deviations from the equilibrium for the threshold indicator as compared to the results in the momentum models. Overall, we conclude that momentum threshold models provide the clearest and most intuitive evidence of nonlinear adjustments in nominal exchange rates to the equilibrium predicted by the monetary model.

Allowing For Stationary Output Deviations
In the two countries where y*-y is stationary, while e and m*-m are nonstationary (Finland and Portugal; see RW), RW omit y*-y to maintain the validity of their (linear) cointegration tests. In contrast, rather than omit y*-y from our cointegration tests we want to include this fundamental variable as a stationary covariate to increase power. While omitting this stationary variable can be seen as a limitation of the OLS based cointegration tests, this limitation does not occur in the IV based tests. In contrast, the test statistic in the ADL-IV threshold cointegration test that we consider retains an asymptotic standard distribution even when a stationary covariate is included. Our test results are displayed in Table 3. 8 Looking at the results, we observe that the null of no cointegration is rejected in at least one regime for Finland at the 5% level of significance. Moreover, it is clear that adjustment to the equilibrium is faster when the change in the deviation is above the threshold level (in regime 1) than when the change is below the threshold level (in regime 2). In particular, the estimated persistent parameter is -0.410 when the change in the deviation from the equilibrium is above the threshold level. This indicates that the nominal exchange rate is clearly stationary and supports adjustment to the equilibrium predicted by the monetary model. However, when the change in the deviation is below the threshold level, the estimated persistent parameter is 0.05 and implies that the nominal exchange rate will behave as a random walk. Overall, including the results for Finland, we can reject the null of no cointegration in the momentum model in 7 out of 10 countries at the 1% or 5% level of significance. If we combine these results with those for Switzerland in Table 2, we can reject the null hypothesis of no cointegration in at least one regime in 8 of 10 countries.

Ⅳ. Conclusion
In this paper, we adopt nonlinear threshold cointegration tests to test for cointegration in the monetary model of exchange rates. While previous researchers have estimated nonlinear versions of the monetary model, they were unable to test for cointegration in a nonlinear framework due to nuisance parameter problems in the existing tests. In this paper, we strive to make a contribution towards filling this gap in the literature. To compare results, we utilize the same long-span data that was previously adopted by Rapach and Wohar (2002) to test for linear cointegration in the monetary model. Given that adopting linear tests can lead to lower power if the underlying model is nonlinear, we test for nonlinear cointegration to see if greater support for the monetary model will occur. We first adopt the ADL-OLS threshold cointegration test developed by Li and Lee (2008) and consider two different threshold models. Following this, we utilize the ADL-IV threshold cointegration test developed by Enders, Im, Lee, and Strazicich (2009). The ADL-IV threshold cointegration test has the distinct advantage that we can include relative output as a stationary covariate to increase power, while the test statistic maintains a standard distribution. Overall, we find greater support for cointegration in the nonlinear framework as compared to the linear cointegration tests in Rapach and Wohar (2002). Moreover, our findings suggest that adjustment to the long-run equilibrium predicted by the monetary model is faster when nominal exchange rates are depreciating as compared to when than appreciating. Finally, our findings complement the growing number of papers that find greater support for the monetary model in a nonlinear framework and perhaps help to explain why Meese (1986), Baillie and Selover (1987), and Sarantis (1994), among others, fail to find support for cointegration in a linear framework.

ABSTRACT
This paper estimates the term structure of interest rates with the setup of 3-factor no arbitrage model and investigates the trend of term premia and the effectiveness of changes in policy interest rates. The term premia are found to be high in a three-year medium term objective, which can be interpreted as reflecting the recognition of investors who expect a higher uncertainty in real activities for the coming three years than for a longer term. Then, in order to look into the effect of policy interest rates after the recent change of benchmark interest rate, this paper analyzes the effects of the changes in short-term interest rates of the financial market on the yield curve of the bond market at time of change. Empirical results show that the discrepancy between call rate, short-term rate in money market, and instantaneous short rate, short-term rate in the bond market, is found to be significantly widened, comparing to the periods before the change in benchmark interest rate. It is not easy to conclude clearly for now whether such a widening gap is caused by the lack of experiences with managing new benchmark interest rate or is just an exceptional case due to the recent turmoil in the global financial market. However, monetary policy needs to be operated in a manner that could reduce the gap to enhance its effectiveness. 본고에서는 3요인 무재정거래 (3- Fama and Bliss(1987), Mishkin(1990), Fama(1990), Campbell and Shiller(1991), Cochrane and Piazzesi(2005), Hamilton and Kim(2002)을 참조하기 바란다. 4) 이에 대한 연구로는 Ang and Piazzesi(2003), Diebold, Rudebusch, and Aruoba(2006), Rudebusch and Wu (2008), Bekaert, Cho, and Moreno(2005) 등을 참조하기 바란다.  Duffie and Kan(1996) 이러한 방식에 의해 추정된 결과는 <Table 2>에 제시되어 있다. 추정된 모수 9) 이와 관련해서 좀 더 자세한 내용은 James and Webber(2000), Kim and Orphanides(2006)

ABSTRACT
With a relatively simple quantitative method, this study comprehensively analyzes the characteristics related to business cycles represented by macroeconomic variables of Korea since 1970. This empirical analysis deals with roughly following three topics: How to identify cyclical component with respect to trend; with what characteristics and how the economic variables of each sector move with in the phases of business cycle, and; whether there are signs of a structural change in the phases of business cycle. Section 2 discusses how to identify trends and cycle components, the basis assumption for the analysis of business cycle. Like the Korean economy, where a relatively high growth rate has been maintained, it is appropriate to determine its economic recession based on the fall in the growth trend, not in the absolute level of real output. And, it is necessary to apply the concept of growth cycle against a traditional concept of business cycle. Accordingly the setting of growth trend is of preliminary importance in identifying cyclical fluctuations. The analysis of Korea's GDP data since 1970, the decomposition of trends and cycles through the Band-pass filter is found to appropriately identify the actual phases of busyness cycle. Section 3 analyzes what particular relationship various economic variables have with output fluctuations during the phases of economic cycle, using the corss-correlation coefficients and prediction contribution. Section 4 monitors the stability of the phases of Korea's business cycle and quantitatively verifies whether there is a structural break, and then reviews the characteristics of variations in each sector. And, stylized facts observed through these studies are summarized in the conclusion.
The macroeconomic stability of Korea, in particular, is found to continue to improve since 1970, except for the financial crisis period. Not only that, it is found that its volatility of economic growth rate as well as inflation have been reduced gradually. Meanwhile, until recently since 2000, the volatility in domestic demand has remained stable, while that in exports and imports has been increased slightly. But, in an over all perspective, Korea's business cycle variation is on the decline due to shorter response period to shocks and the formation of complementary relationship among economic sectors. [ Figure 2] Deviations from Linear Trend for GDP    1970 197 5 1980 1985 1990 1995 2000 2005 [ Figure 5] Cyclical Component from B-P Filtering for GDP Note: GDP, constant Won, seasonally adjusted, quarterly, natural logs.     Note: The Cyclical Component is extracted from B-P Filtering which drops data of 12 quarters from the initial data point. Therefore Cyclical component series start at the first quarter of 1973.
[ Figure 9] Volatility Trend of CPI and Cyclical Component of GDP

71
[ Figure 10] Time-Varying Parameter Estimation on   using GARCH Model      In the internet portal industry, the indirect network externality from portal visitors to advertisers and the direct network externality among portal visitors have important implications for anti-trust policies. This paper examines the existence and the magnitude of the direct/indirect network externality in the Korean internet portal industry and measures its effect on the market power of the internet portals. The results show that the direct/indirect network externality is substantive in the industry hence the market share of a portal in the visitors' side has the 'leverage' effect on its market power in the advertisers' side. 2) Farrell and Shapiro (1990), Willig(1991), Landes and Posner(1981

ABSTRACT
There is a growing concern about potential harmful effect of second-hand or environmental tobacco smoking. As a result, smoking bans in workplace become more prevalent worldwide. In Korea, workplace smoking ban policy become more restrictive in 2003 when National health enhancing law was amended. The new law requires all office buildings larger than 3,000 square meters (multi-purpose buildings larger than 2,000 square meters) should be smoke free. Therefore, a lot of indoor office became non smoking area. Previous studies in other counties often found contradicting answers for the effects of workplace smoking ban on smoking behavior. In addition, there was no study in Korea yet that examines the causal impacts of smoking ban on smoking behavior. The situation in Korea might be different from other countries. Using 2001 and 2005 Korea National Health and Nutrition surveys which are representative for population in Korea we try to examine the impacts of law change on current smoker and cigarettes smoked per day. The amended law impacted the whole country at the same time and there was a declining trend in smoking rate even before the legislation update. So, the challenge here is to tease out the true impact only. We compare indoor working occupations which are constrained by the law change with outdoor working occupations which are less impacted. Since the data has been collected before (2001) and after (2005) the law change for treated (indoor working occupations) and control (outdoor working occupations) groups we will use difference in difference method. We restrict our sample to working age (between 20 and 65) since these are the relevant population by the workplace smoking ban policy. We also restrict the sample to indoor occupations (executive or administrative and administrative support) and outdoor occupations (sales and low skilled worker) after dropping unemployed and someone working for military since it is not clear whether these occupations are treated group or control group. This classification was supported when we examined the answers for workplace smoking ban policy existing only in 2005 survey. Sixty eight percent of indoor occupations reported having an office smoking ban policy compared to forty percent of outdoor occupation answering workplace smoking ban policy. The estimated impacts on current smoker are 4.1 percentage point decline and cigarettes per day show statistically significant decline of 2.5 cigarettes per day. Taking into account consumption of average sixteen cigarettes per day among smokers it is sixteen percent decline in smoking rate which is substantial. We tested robustness using the same sample across two surveys and also using tobit model. Our results are robust against both concerns. It is possible that our measure of treated and control group have measurement error which will lead to attenuation bias. Note: See Notes on Table 2 for the first column and Table 3

ABSTRACT
The Korean economy successfully overcame the macroeconomic downturns driven from the Asian financial crisis in a very short period of time. The economic shock, however, generated a variety of social problems, one of which was the increase in felonies (homicides, robbery, rape, and arson), or degradation of public safety.
We argue that the Korean criminal policy has not been effective to ameliorate the rising trends in crime caused by the financial crisis. In order to substantiate this claim, we assess the effectiveness of criminal policy: policing, sentencing, and corrections.
First, there has been resource shortage in policing since the 1997 financial crisis. For the past ten years, the investment of human resource and budget in the police has been virtually stagnant, as well as in prosecutors' investigation activities. The insufficient resource allocation in policing caused a huge decline in arrest rates and prosecution rates.
Second, the Korean judicial system has not increased the severity of punishment. Comparing the pre-and the post-financial crisis period, the average length of prison sentence by the courts has declined. Given the degrading in the quality of crime and the decreasing amount of inputs into the policing and prosecution, the government should have increased the severity of punishment to deter crime.
Third, we found that the government hired more officers and allocated larger budget into prison and probation. However, it is difficult to suggest that the increased level of resources in correctional programs have been effective in preventing released prisoners from committing future crimes. This is because the number of repeat offenders convicted of more than a third offense increased dramatically since 1997, pushing felonies upward.
In sum, the government organizations failed to respond respectively or to make coordinated actions, eventually causing a dramatic increase in crimes. This research brings explicit policy implications. In order to prevent possible additional degradation of public safety, the government must put more efforts into increasing the effectiveness of policy and to investing more resources into said policies. We also emphasize the importance of the institutional mechanisms which foster policy coordination among the Police, the Prosecutor's Office, the Ministry of Justice, and other relevant government organizations.    n it e d S t a t e s T u r k e y S w it z e r l a n d F in l a n d S w e d e n S l o v e n ia C z e c h R e p u b li c K o r e a H n g r a r y U n i t e d K i n g d o m C a n a d a P o r t u g a l F r a n c e P o l a n d B e lg iu m A u s t r a l ia N e w Z e a l a n d I t a l y S p a i n G e r m a n y I c e l a n d Notes: Homicide rate is computed per 100,000. The homicide rates of Mexico marks 13 per 100,000 people, the highest among the OECD nations, but excluded in this figure.     1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Year Proportion of labor cost in total police expenditure (%)    1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Year Assistant employees per prosecutor  1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Year  1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 Year  Notes: The number of cases sentenced to "Less than 3 years" is the sum of "Less than 1 year" and "Less than 3 years".      1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Year 항등식 (   This paper investigates changes in the extent of exchange rate pass-through to export price in Korea. First, empirical results show that export prices have become less responsive to the exchange rate since the financial crisis in 1997. The decline of exchange rate pass-through to export prices suggests that Korean exporters are more likely to use profit margins to absorb part of the impact of exchange rate changes, consistent with pricing to market phenomenon. Second, this paper finds asymmetries in the response of export prices to exchange rate changes. In the post-crisis period. appreciations are more likely to be offset by markup adjustment than depreciations. Third, this paper documents that a significant portion of the decline of exchange rate pass-through is a result of both increased volatility of exchange rate and increased competition with China in the world market.   1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 400 600 800 1000 1200 1400 1600 1800