Measuring International Uncertainty: the Case of Korea

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Measuring International Uncertainty: the Case of Korea Minchul Shin, Boyuan Zhang, Molin Zhong, and Dong Jin Lee 2017-066 Please cite this paper as: Shin, Minchul, Boyuan Zhang, Molin Zhong, and Dong Jin Lee (2017). “Measuring International Uncertainty: the Case of Korea,” Finance and Economics Discussion Series 2017-066. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2017.066. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Measuring International Uncertainty: the Case of Korea Minchul Shin Boyuan Zhang Molin Zhong Univ. of Illinois Univ. of Illinois Federal Reserve Board Dong Jin Lee ∗ Bank of Korea This version: April 20, 2017 Abstract We leverage a data rich environment to construct and study a measure of macroeconomic uncertainty for the Korean economy. We provide several stylized facts about uncertainty in Korea from 1991M10–2016M5. We compare and contrast this measure of uncertainty with two other popular uncertainty proxies, financial and policy uncertainty proxies, as well as the U.S. measure constructed by Jurado et al. (2015). Key words: Uncertainty, Stochastic volatility, Business cycle, Korean economy, Data rich environment. JEL codes: C11, C32, E32 Correspondence: Minchul Shin: 214 David Kinley Hall, 1407 W. Gregory Dr., Urbana, Illinois 61801. ∗ E-mail: mincshin@illinois.edu. Boyuan Zhang: Email: bzhang51@illinois.edu. Molin Zhong: 20th Street and Constitution Avenue N.W., Washington, D.C. 20551. E-mail: molin.zhong@frb.gov. Dong Jin Lee: E-mail: rheedj@bok.or.kr. We are grateful for comments from Sangyup Choi, Myungkyu Shim, and Rae Lee. Shin gratefully acknowledges financial support from the Bank of Korea. The views expressed in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System, any other person associated with the Federal Reserve System, or the Bank of Korea.

1 Measuring International Uncertainty: the Case of Korea 1 Introduction Macroeconomic uncertainty and its relationship with the business cycle has received much attention in the U.S. and internationally. Measuring uncertainty, however, has proven challenging because it is not directly observable. International studies largely proxy for uncertainty using the VIX or economic policy uncertainty index (e.g., Carriere-Swallow and Cespedes, 2013; Choi and Shim, 2016; Baker et al., 2017). Recently, Jurado et al. (2015) have provided a leading measure of “objective” uncertainty for the U.S. economy. This index has two attractive features: it is measured from macroeconomic data volatility in a reduced-form way and it covers a broad range of indicators spanning the entire macroeconomy. Internationally, this strategy for measuring uncertainty has remained unused because of data availability costs. This paper works to resolve this gap by using 112 data series to provide a broad-based measure of uncertainty for Korea. We hope that this paper encourages more work on uncertainty fluctuations in small open economies such as the Korean economy.1 2 Construction of the uncertainty measure 2.1 Methodology Following Jurado et al. (2015), we define the uncertainty of an individual series as the conditional volatility of the unforecastable component of the future values of that series. The h-period ahead uncertainty in the variable y Y = (y ,...,y ) is defined as jt ∈ t 1t Nyt 0 Uy(h) E y E y I 2 I (1) jt ≡ jt+h − jt+h t t r h (cid:12) i (cid:0) (cid:2) (cid:12) (cid:3)(cid:1) (cid:12) (cid:12) where the expectation is taken with respect to the information(cid:12)set I available to agents at t time t. If the expectation of the squared error in forecasting y rises, uncertainty in the jt+h 1We plan to make our estimated uncertainty indices available from our personal and journal websites.

2 variableincreases. Wecomputetheindividualuncertaintiesbymodelingtheindividualseries as factor augmented AR(p) models where both common factor and idiosyncratic shocks have stochastic volatility. The description of the model is in the appendix. A measure of macroeconomic uncertainty aggregates the individual uncertainties of each series at every date: Ny Uy(h) w Uy(h) (2) t ≡ j jt j=1 X where w isaweightassignedtotheuncertaintyinthe jthvariable. Ourbaselineuncertainty j measure is based on w = 1/N and h = 1. This index measures an average difficulty of j y predicting a time series in the economy.2 2.2 Data We use 112 monthly time series that represent the Korean economy from 1991M10-2016M5. We categorize these 112 individual series into 8 groups: (1) Output; (2) Labor market; (3) Housing market; (4) Consumption, orders, and inventories; (5) Money and loans; (6) Bonds and stocks; (7) Prices; (8) Imports and exports. We also compare our uncertainty index with two other popular uncertainty proxies. The firstistheVKOSPIindex,whichistheKoreanversionoftheVIX(optionimpliedvolatility).3 The second is the economic policy uncertainty (EPU) index (Baker et al., 2017), which is a news-based uncertainty measure meant to capture movements in policy-related economic uncertainty. 3 Korean Uncertainty Index 3.1 Aggregate uncertainty measure In Figure 1, we present our uncertainty measure for Korea overlaid with other uncertainty proxies. The shaded areas are Korea’s recessionary periods defined by the Korea National Statistical Office. We provide three stylized facts about uncertainty fluctuations. First, our uncertainty measure typically starts going up at the beginning of recessionary periods and has local peaks in the middle of recessionary periods. The exception is the 2We also compute the aggregate uncertainty index with w obtained from the first principal component. j The main features are not different from the simple average. 3We use the term VIX and VKOSPI interchangeably.

3 Figure 1 Korean Economic Uncertainty Index (UI) (a) Uncertainty Index (black) and VKOSPI/VIX (grey) (b) Uncertainty Index (black) and EPU (grey) Note: Uncertainty index overlaid with VIX index (upper panel) and EPU index (lower panel) for 1991M10- 2016M5. Pairwise correlations of these series are corr(UI, VIX) = 0.74, corr(UI, EPU) = -0.03, corr(VIX, EPU) = 0.17. 2000M8–2001M7 recession when the Korean economy suffered several damaging domestic (Daewoo Motors’ Bankruptcy) and foreign (9/11 attacks) economic events. Such shocks worsened Korea’s economic performance, but did not affect overall uncertainty. Second, the uncertainty index is persistent, positively-skewed, and fat-tailed. Table 1 presents descriptive statistics of our uncertainty measure. We also compute the same statistics for the other uncertainty proxies. They share similar properties although our measure exhibits higher persistence, skewness, and kurtosis. Third, our uncertainty index is countercyclical. Table 1 shows that the index’s contemporaneous correlation with industrial production growth is -0.28. The rest shows crosscorrelations between our uncertainty measure and IP growth (correlation between the uncertainty index at t and IP growth at t+k). They are negatively correlated within 6 month leads and lags. However, the correlation between uncertainty and IP growth becomes positive as k becomes large and is maximized at k = 18, which matches the average duration of

4 Table 1 Korean Economic Uncertainty Index (UI) UI VIX EPU Corr. UI VIX EPU Corr. UI VIX EPU AR(1) 0.97 0.82 0.72 with IP(k) with IP(k) Half Life 21.09 3.51 2.07 k = 1 -0.24 -0.06 -0.21 k =+1 -0.29 -0.14 -0.26 − Skewness 2.08 1.61 1.03 k = 3 -0.16 0.02 -0.15 k =+3 -0.25 -0.13 -0.28 − Kurtosis 8.83 6.26 4.34 k = 6 -0.07 0.08 -0.12 k =+6 -0.08 -0.01 -0.20 − IP-corr(0) -0.28 -0.11 -0.24 k = 12 0.05 0.13 0.00 k =+12 0.43 0.30 -0.01 − Table 2 Individual Uncertainties and Other Uncertainty Proxies with VIX Full sample Recession Expansion Top 1 KOSPI 0.73 Ex rate: UK 0.73 KOSPI 0.81 Top 2 Ex rate: UK 0.64 Ex rate: avg (nominal) 0.72 M2 0.62 Top 3 Ex rate: US 0.62 Ex rate: US 0.72 TB3y-MSB1y spread 0.57 with EPU Full sample Recession Expansion Top 1 BoP: CA 0.50 BoP: CA 0.54 BoP: CA 0.47 Top 2 Baltic Dry Index 0.47 Baltic Dry Index 0.48 Exports of goods 0.42 Top 3 PIP: Total 0.45 PIP: Total 0.47 Baltic Dry Index 0.42 Note: Numbers in the table are correlation between uncertainty proxies and corresponding individual uncertainties. A detailed description of the individual series is in the online appendix. recessions (18.4 months). 3.2 Comparisons with other uncertainty measures VIX. Panel (a) in Figure 1 shows the VIX index. It generally moves together with our uncertainty measure with a correlation of 0.74. Although both measures move closely, there are differences.4 One example is the 2002M12–2005M4 recession. At the beginning of this recession, both uncertainty measures increased. Our uncertainty measure had its highest peak on 2003M7, mostly due to heightened uncertainties of the variables in Group 1 (Output) and Group 4 (Consumption, orders, inventories) driven by the Korean credit card lending boom (1999– 2002) and bust (2003).5 Unlike our uncertainty measure, the VIX had its highest peak 4Moreover, as we show in a VAR exercise in the online appendix that includes both our uncertainty index and the VIX, shocks to our uncertainty index have different effects on the macroeconomy relative to shocks to the VIX. 5This recession is also known as the credit card crisis. See for example, Kang and Ma (2007).

5 around 2004M6 from news about policy rate increases by the Chinese government and the FederalReserve. Duringthisperiod, variablesinGroup4(Consumption, orders, inventories) and Group 9 (Imports and exports) contributed the most to our uncertainty measure, which distinguished the origin of this uncertainty from that of the 2003M7 peak. However, this heightened uncertainty in the international market did not translate into uncertainty about the overall Korean economy as our overall measure did not move much during this period. Table 2 reveals this disconnect between the two uncertainty measures. It reports the three individual uncertainties that are most associated with the volatility proxies. The VIX index is most related to uncertainty about financial variables such as the KOSPI index, exchange rates, and interest rate spreads. EPU. ThesecondpanelinFigure 1 showsthetimeseriesplotoftheEPUoverlaidwithour uncertainty measure. Our uncertainty measure and the EPU show quite different dynamics. For example, based on our measure, economic uncertainty was highest during the Asian financial crisis, while the EPU put low weight to the crisis. Other examples are periods with international affairs such as the 9/11 attacks, Gulf War II, the Eurozone debt crisis that may have increased uncertainty outside the Korean economy but not inside. The correlation between our uncertainty measure and the EPU is essentially zero (-0.03). One explanation for the low correlation is that some of the economic policy related uncertainty that the Korean news articles mentioned did not manifest as increases in the overall uncertainty of the Korean economy. Table 2 shows that the EPU index is correlated with uncertainties related to trade activities such as the current account, Baltic Dry index, import price index, and the exports of goods. This finding suggests that the index weighs heavily international affairs relative to domestic affairs. 3.3 Comparison to U.S. index The Korean and U.S. uncertainty measures share some common statistical properties. As we can see from Table 3a, economic uncertainty measures for both countries are persistent, positively skewed, fat-tailed, and countercyclical. In Table 3b, we report relationships among the various uncertainty measures from Korea and the U.S. We find that these uncertainty measures are correlated with each other with the exception of the Korean uncertainty index and the EPU index pair. Moreover, uncertainty in the Korean and U.S. financial markets are more related than are the broad-based uncertainty

6 Table 3 Korean and U.S. Uncertainty Measures (a) Descriptive statistics (b) Correlation among uncertainty measures Korea U.S. UI-K VIX-K EPU-K UI–US VIX-US UI VIX EPU UI VIX EPU UI-K 1 . . . . VIX-K 0.70 1 . . . AR(1) 0.96 0.78 0.73 0.99 0.89 0.73 EPU-K 0.03 0.25 1 . . Half Life 18.73 2.85 2.19 50.81 6.01 2.25 UI-US 0.33 0.38 0.45 1 . Skewness 2.23 1.59 1.19 2.33 1.79 1.48 VIX-US 0.50 0.69 0.50 0.63 1 Kurtosis 8.92 6.06 4.62 9.69 8.19 5.72 EPU-US 0.12 0.30 0.68 0.39 0.61 IP-corr(0) -0.49 -0.29 -0.22 -0.79 -0.49 -0.34 Note: Our sample for this table covers 1991M10–2011M12. K: Korea, US: United States. indices. This is sensible given that financial trading has less barriers than trading in other markets. Furthermore, the Korean EPU index is more associated with U.S. uncertainty measures than Korean ones. This suggests that Korean newspapers overweight news about foreign policy-related uncertainty which may not pass through to the domestic market. This is consistent with the previous subsection: the Korean EPU index is related to international trade activities variables, which may not be relevant for uncertainty about the Korean economy. The Korean EPU index is also less correlated with the Korean VIX and uncertainty index relative to the U.S. EPU index’s correlation with its U.S. counterparts. 4 Conclusion We construct an uncertainty measure based on 112 economic time series for Korea. We provide a set of stylized facts about Korean economic uncertainty. In addition, we find that other uncertainty proxies are associated with specific sectors and do not represent uncertainty of the whole economy. One needs to be cautious about the use of news-based measures because journalists’ view about uncertainty can be quite different across countries. For example, the EPU index for the Korean economy is more associated foreign uncertainty. References Baker, S., N. Bloom, and S. Davis (2017): “Measuring Economic Policy Uncertainty,” Quarterly Journal of Economics, Forthcoming. Carriere-Swallow, Y. and L. Cespedes (2013): “The Impact of Uncertainty Shocks in Emerging Economies,” Journal of International Economics, 90, 316–325.

7 Choi, S. and M. Shim (2016): “Financial vs. Policy Uncertainty in Emerging Economies: Evidence from Korea and the BRICs,” Working paper. Jurado, K., S. C. Ludvigson, and S. Ng (2015): “Measuring Uncertainty,” American Economic Review, 105, 1177–1216. Kang, T. and G. Ma (2007): “Recent Episodes of Credit Card Distress in Asia,” BIS Quarterly Review.

Online appendix (not for publication) Measuring International Uncertainty: the Case of Korea This version: April 20, 2017 This appendix has three sections: 1. Section 1 describes data used in this paper. 2. Section 2 provides the detailed description of the computation of the uncertainty index. 3. Section 3providesfurthercomparisonsoftheuncertaintyindexandotherproxiesbased on the VAR estimation. 4. Section 4 presents additional VAR results with different model specifications (different lag length, ordering of variables, and other combinations of variables). 1 Variables used in analysis In this section we present a list of variables used in the construction of our uncertainty measure. In each table, there are 5 columns: ID: Numeric number that identifies each series. • Group: Numeric number that identifies each group. • Trans: Numeric number that indicates the type of data transformation that is applied • to each individual series (Transformation code 3 and 5 are not used in this application). – Trans = 1: Transformation is not applied. – Trans = 2: X = log(Xraw) log(Xraw ). i,t i,t − i,t 1 − – Trans = 4: X = log(Xraw). i,t i,t – Trans = 6: X = Xraw 100. i,t i,t − Name: Name of each variable. • 1

2 Description: Description of each variable. • We apply X-12 to remove the seasonal component from each individual series and apply appropriate transformations to make all individual series stationary. In addition to these variables, we also use VKOSPI/VIX, EPU, and U.S. uncertainty measures. 1. The VKOSPI index (VIX), which is the Korean version of the VIX (the option implied volatility measure based on S&P500 index options). This measure is an implied volatility based on KOSPI 200 index options and it serves as a proxy for stock market uncertainty in Korea. More general discussion about VKOSPI can be found in Han etal.(2015). TheVKOSPIindexisonlyavailablefrom2003. WefollowChoiandShim (2016) and extend the VKOSPI index series back to 1991 using the realized volatility of the daily KOSPI200 index.. 2. The economic policy uncertainty (EPU) index constructed by Baker et al. (2017). This measure and its description are available from their webpage: http://www. policyuncertainty.com/ 3. U.S. uncertainty measure of Jurado et al. (2015) is taken from Serena Ng’s webpage: http://www.columbia.edu/~sn2294/pub.html

3 Table 1 Variables in Group 1: Output ID Group Trans Name Description 1 1 2 IP:Total Industry,IndustrialProductionIndexbyIndustry,AllGroups,ProductionIndex(2010),SA 2 1 2 IP:Manufacturing IndustrialProductionIndexbyIndustry,Manufacturing(2010),SA 3 1 2 IP:Chemical Industry,IndustrialProductionIndexbyIndustry,ManufactureofChemicalsandChemicalProductsexceptPharmaceuticals,ProductionIndex(2010),SA 4 1 2 IP:Equipment IndustrialProductionIndexbyIndustry,ElectricalEquipment(2010),SA 5 1 2 IP:Vehicles IndustrialProductionIndexbyIndustry,MotorVehicles,TrailersandSemitrailers(2010),SA 6 1 2 IP:Capitalgoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,CapitalGoods,ProductionIndex(2010),SA 7 1 2 IP:Intermediategoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,IntermediateGoods,ProductionIndex(2010),SA 8 1 2 IP:Consumersgoods IndustrialProductionIndexbyMarketGroup,WholeCountry,Consumers’Goods,ProductionIndex(2010),SA 9 1 2 SI:Total Industry,IndustrialProductionIndexbyIndustry,AllGroups,ShipmentIndex(2010),SA-Korea,Republicof 10 1 2 OI:Manufacturing MANUFACTURINGOPERATIONRATIOINDEX(2010),SA-SOUTHKOREA 11 1 2 OI:Chemicals ChemicalsOPERATIONRATIOINDEX(2010),SA-SOUTHKOREA 12 1 2 OI:Equipment ElectricalEquipmentOPERATIONRATIOINDEX(2010),SA-SOUTHKOREA 13 1 2 OI:Vehicles MotorVehicles,TrailersandSemitrailersOPERATIONRATIOINDEX(2010),SA-SOUTHKOREA 14 1 6 Businessconditions(total,actual) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Actual,CompositeBusinessConditionsIndex,Actual,NSA 15 1 6 Businessconditions(manufacturing,actual) SurveysandCyclicalIndicators,BusinessSurveyIndex,Manufacturing,Actual,CompositeBusinessConditionsIndex,Actual,NSA 16 1 6 Businessconditions(total,forecast) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Forecast,CompositeBusinessConditionsIndex,NSA-Kor 17 1 6 Businessconditions(manufacturing,forecast) SurveysandCyclicalIndicators,BusinessSurveyIndex,Manufacturing,Forecast,CompositeBusinessConditionsIndex,NSA- Kore

4 Table 2 Variables in Group 2: Labor Markets ID Group Trans Name Description 18 2 2 EMP:Employed LaborMarket,Population,Total,EmployedPersons,SA 19 2 2 EMP:Unemployed LaborMarket,Population,Total,UnemployedPersons,SA 20 2 1 EMP:Unemploymentrate LaborMarket,Population,Total,UnemploymentRate,SA 21 2 2 EMP:Emplyees,Manufacturing Employees: Manufacturing 22 2 2 EMP:regularworkers LaborMarket,Employed,Total,Wage&salaryworkers,Regularemployees,NSA 23 2 2 EMP:temporaryworkers LaborMarket,Employed,Total,Wage&salaryworkers,Temporaryemployees,NSA 24 2 1 EMP:participationrate LabourMarket,ParticipationRate,Total,NSA-Korea,Republicof 25 2 1 Hoursworked LabourMarket,EmployedPersonsbyHoursWorked,AverageWeeklyWorkingHours,NSA-Korea,Republicof 26 2 6 Businessconditions(emloyment,actual) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Actual,EmploymentIndex,Actual,NSA-Korea,Republic 27 2 6 Businessconditions(employment,forecast) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Forecast,EmploymentIndex,NSA-Korea,Republic of Table 3 Variables in Group 3: Housing Markets ID Group Trans Name Description 28 3 1 Permits: Dwellings(level) KORPermitsissuedfordwellingssa/MonthlyLevelSA-KOREA 29 3 2 Housingstarts Korea,Republicof(SouthKorea)-Housingstarts-Unit-Unit-NSA-Monthly 30 3 1 Permits: Dwellings(YoYgrowth) KORPermitsissuedfordwellingssa/GrowthratesameperiodpreviousyearSA-KOREA 31 3 2 Housingpriceindex Housingpurchasepricecompositeindices(seasonallyadjusted),(2015.12=100) 32 3 2 Housingjeonsepriceindex Housingjeonsepricecompositeindices(seasonallyadjusted)(*Jeonse: Keymoneydepositlease)(2015.12=100) 33 3 2 Constructioncontracts Constructioncontracts 34 3 2 Constructioncontracts: Public Constructioncontracts: Public 35 3 2 Constructioncontracts: Private Constructioncontracts: Private 36 3 2 Permits: Total Permits: Total

5 Table 4 Variables in Group 4: Consumption, Order, Inventory ID Group Trans Name Description 37 4 2 II:Total Industry,IndustrialProductionIndexbyIndustry,AllGroups,InventoryIndex(2010),SA 38 4 2 II:Manufacturing Industry,IndustrialProductionIndexbyIndustry,Manufacturing,ProductionIndex(2010),SA 39 4 2 II:Capitalgoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,CapitalGoods,InventoryIndex(2010),SA 40 4 2 II:Intermediategoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,IntermediateGoods,InventoryIndex(2010),SA 41 4 2 II:Consumersgoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,Consumers’Goods,InventoryIndex(2010),SA 42 4 2 Automobileregistration WholesaleTradeandRetailSales,AutomobileRegistration,Total,NSA-Korea,Republicof 43 4 6 Businessconditions(demand,actual) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Actual,DomesticDemandIndex,Actual,NSA-Korea,Repu 44 4 6 Businessconditions(demand,forecast) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Forecast,DomesticDemandIndex,NSA-Korea,Republico 45 4 2 Salesofgoods KORSalesoftotalmanufacturedgoods(Volume)sa/IndexpublicationbaseSA-KOREA 46 4 2 Salesofconsumergoods KORSalesoftotalmanufacturedconsumergoods(Volume)sa/IndexpublicationbaseSA-KOREA 47 4 2 Salesofintermediategoods KORSalesofmanufacturedintermediategoods(Volume)sa/IndexpublicationbaseSA-KOREA 48 4 2 Salesofinvestmentgoods KORSalesofmanufacturedinvestmentgoods(Volume)sa/IndexpublicationbaseSA-KOREA 49 4 2 Retailtradevolume KORTotalretailtrade(Volume)/Indexpublicationbase-KOREA 50 4 2 SI:Manufacturing Industry,IndustrialProductionIndexbyIndustry,Manufacturing,ShipmentIndex(2010),SA-Korea,Republicof 51 4 2 SI:Chemicals Industry,IndustrialProductionIndexbyIndustry,ManufactureofChemicalsandChemicalProductsexceptPharmaceuticals 52 4 2 SI:Equipment Industry,IndustrialProductionIndexbyIndustry,ManufactureofElectricalEquipment,ShipmentIndex(2010),SA 53 4 2 SI:Vehicles Industry,IndustrialProductionIndexbyIndustry,MotorVehicles,TrailersandSemitrailers,ShipmentIndex(2010),SA 54 4 2 SI:Capitalgoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,CapitalGoods,ShipmentIndex(2010),SA 55 4 2 SI:Intermediategoods Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,IntermediateGoods,ShipmentIndex(2010),SA 56 4 2 SI:Consumersgoods IndustrialProductionIndexbyMarketGroup,WholeCountry,Consumers’Goods,ShipmentIndex(2010),SA 57 4 2 SI:Capitalgoods(domestic) Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,CapitalGoods(domestic),ShipmentIndex(2010),SA 58 4 2 SI:Intermediategoods(domestic) Industry,IndustrialProductionIndexbyMarketGroup,WholeCountry,IntermediateGoods(domestic),ShipmentIndex(2010),SA 59 4 2 SI:Consumersgoods(domestic) IndustrialProductionIndexbyMarketGroup,WholeCountry,Consumers’Goods(domestic),ShipmentIndex(2010), SA

6 Table 5 Variables in Group 5: Money and Loan ID Group Trans Name Description 60 5 2 M1 MoneyandBanking,MoneySupply,SeasonallyAjustedM1(Endof),SA-Korea,Republicof 61 5 2 M2 Supply,SeasonallyAjustedM2(Endof),SA-Korea,Republicof 62 5 2 LF Supply,SeasonallyAdjustedLf(EndOf),SA-Korea,Republicof 63 5 2 TotalDeposits MoneyandBanking,MoneySupply,TotalDepositsofCBs&SBs. (EndOf),NSA-Korea,Republicof 64 5 2 TotalLoans MoneyandBanking,MoneySupply,LoansofCBs&SBs(EndOf),NSA-Korea,Republicof 65 5 1 Turnoverratio MoneyandBanking,MoneySupply,TurnoverRatioofDemandDeposits,CBs&SBs,NSA-Korea,Republicof 66 5 2 Internationalreserves InternationalLiquidity,InternationalReserves,Korea,Republic Of Table 6 Variables in Group 6: Bond and Stock ID Group Trans Name Description 67 6 1 Callrate MoneyandBanking,MarketInterestRates,UncollateralizedCallRates(AllTransactions),NSA-Korea,Republicof 68 6 1 Housingbonds(5y) KORYield5-yearhousingbonds/Quantum(non-additiveorstockfigures)-KOREA 69 6 1 MSB(1y) MoneyandBanking,MarketInterestRates,YieldsofMonetaryStab. Bonds(364-day),NSA-Korea,Republicof 70 6 1 Fin. debenturesbonds(1y) MoneyandBanking,MarketInterestRates,YieldsofFinancialDebentures(1-year),NSA-Korea,Republicof 71 6 1 Fin. debenturesbonds(3y) MoneyandBanking,MarketInterestRates,YieldsofFinancialDebentures(3-Year),NSA-Korea,Republicof 72 6 1 Corp. bonds(AA-,3y) MoneyandBanking,MarketInterestRates,YieldsofCorporateBonds: O.T.C(3-year,AA-),NSA-Korea,Republicof 73 6 1 CDs(91days) KORYield91-dayCDs/Quantum(non-additiveorstockfigures)-KOREA 74 6 1 Treasurybonds(3y) MoneyandBanking,MarketInterestRates,YieldsofTreasuryBonds(3-year),NSA-Korea,Republicof 75 6 1 Treasurybonds(5y) MoneyandBanking,MarketInterestRates,YieldsofTreasuryBonds(5-year),NSA-Korea,Republicof 76 6 1 FD3y-MSB1yspread FD3y-MSB1yspread 77 6 1 CB3y-MSB1yspread CB3y-MSB1yspread 78 6 1 TB3y-MSB1yspread TB3y-MSB1yspread 79 6 1 TB5y-MSB1yspread TB5y-MSB1yspread 80 6 1 CB3y-TB3y CB3y-TB3y 81 6 2 KOSPI KoreaCompositeStockPriceIndex

7 Table 7 Variables in Group 7: Price ID Group Trans Name Description 82 7 2 CPI:All Prices,ConsumerPriceIndex,CPI:AllItems(2010),NSA-Korea,Republicof 83 7 2 CPI:exceptAgriandOils Prices,ConsumerPriceIndex,CPI:Allitems,excludingAgriculturalProductsandOils(2010),NSA-Korea,Republicof 84 7 2 CPI:exceptFoodandEnergy Prices,ConsumerPriceIndex,CPI:AllItems,excludingFoodandEnergy(2010),NSA-Korea,Republicof 85 7 2 PPI:Total Prices,ProducerPriceIndexPPI:Total(2010=100)NSA-Korea,Republicof 86 7 2 PIP:Total Prices,ImportPriceIndexbySpecialGroups,AllCommodities(DollarBasis)(2010),NSA-Korea,Republicof 87 7 2 CrudeOilPrice Oil;Dubai,medium,Fateh32API,fobDubaiCrudeOil(petroleum),DubaiFatehFateh32API,US$perbarrel 88 7 2 MetalsPrice MetalsPriceIndex,2005=100,includesCopper,Aluminum,IronOre,Tin,Nickel,Zinc,Lead,andUraniumPriceIndices 89 7 2 AgriculturalPrice AgriculturalRawMaterialsIndex,2005=100,includesTimber,Cotton,Wool,Rubber,andHidesPrice Indices Table 8 Variables in Group 8: Export, Import, and Trade Condition ID Group Trans Name Description 90 8 2 Exportvol EXPORTVOLUMEINDEX(2010)-TOTAL,NSA-SOUTHKOREA 91 8 2 Importvol IMPORTVOLUMEINDEX(2010)-TOTAL,NSA-SOUTHKOREA 92 8 2 Exportsofgoods EXPORTSOFGOODS-MILLIONSOFUSDOLLARS,NSA-SOUTHKOREA,thenchangetoin1000dollars 93 8 2 Importsofgoods IMPORTSOFGOODS-MILLIONSOFUSDOLLARS,NSA-SOUTHKOREA,thenchangetoin1000dollars 94 8 1 BoP:CA BalanceofPayments,CurrentAccount,SA-Korea,Republicof 95 8 6 CLI:OECD OECDCompositeLeadingIndicators,amplitudeadjusted 96 8 2 WTMindex CPBWorldTradeMonitor(WTM),index2005=100 97 8 2 CLI:France OECDCompositeLeadingIndicators,France 98 8 2 CLI:Germany OECDCompositeLeadingIndicators,Germany 99 8 2 CLI:Japan OECDCompositeLeadingIndicators,Japan 100 8 2 CLI:Japan OECDCompositeLeadingIndicators,UnitedKingdom 101 8 2 CLI:US OECDCompositeLeadingIndicators,US 102 8 2 CLI:G7 OECDCompositeLeadingIndicators,G7 103 8 2 CLI:Europe OECDCompositeLeadingIndicators,Europe 104 8 2 CLI:Total OECDCompositeLeadingIndicators,OECDall 105 8 2 BalticDryIndex BalticDryIndex 106 8 2 Exrate: US ForeignExcahngeRate: US 107 8 2 Exrate: Yen ForeignExchangeRate: Japan 108 8 2 Exrate: UK ForeignExchangeRate: UnitedKingdom 109 8 2 Exrate: avg(real) RealEffectiveExchangeRate 110 8 2 Exrate: avg(nominal) NominalEffectiveExchangeRate 111 8 6 Businessconditions(export,actual) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Actual,ExportsIndex,Actual,NSA-Korea,Republicof 112 8 6 Businessconditions(export,forecast) SurveysandCyclicalIndicators,BusinessSurveyIndex,AllIndustries,Forecast,ExportsIndex,NSA-Korea,Republic of

8 2 Uncertainty index computation The main step in computing the individual uncertainties is to approximate the purely unforecastable component of the future values of individual series, y E y I , and its jt+h jt+h t − 2 variance, E y E y I I . To do so, we compute the conditional mean of y jt+h jt+h t t (cid:2) (cid:12) (cid:3) jt+h − (cid:12) as the h-steph ahead point prediction(cid:12) miade at time t based on the following model: (cid:0) (cid:2) (cid:12) (cid:3)(cid:1) (cid:12) (cid:12) (cid:12) p1 p2 y = φ y + Γ Z +v j,t j,l j,t l 0j,l t l j,t − − l=1 l=1 X X (1) p3 Z = Φ Z +vZ t l t 1 t − l=1 X where Z = [(F ),(F2 ),(G )]. F is a vector of the first r principal components of Y , t t 0 1,t 0 t 0 0 t f t (F2 ) is the squared first principal component of Y , and G is a vector of the first r principal 1,t t t g components of Y2. Based on this model, we can approximate each individual uncertainty t by assuming distributional characteristics of the shocks in the system above. To take into account the time-varying forecast error variance, we assume that both innovations v and jt vZ follow the stochastic volatility model: t v N(0, exp(h )) where h = ch +ϕhh +σhη , η N(0,1) j,t ∼ j,t j,t j j j,t − 1 j j,t j,t ∼ i.i.d. (2) vZ N(0, exp(w )) where w = cw +ϕww +σwζ , ζ N(0,1) k,t ∼ k,t k,t k k k,t − 1 k k,t k,t ∼ i.i.d. for j = 1,...,N and k = 1,...,(r + r + 1). In our empirical application, we select the y f g number of factor predictors based on the Bayesian information criteria and follow Jurado et al. (2015) to select other tuning parameters (r = 9, r = 1, p = 4, p = 2, p = 4). As f g 1 2 3 in Jurado et al. (2015), individual uncertainties are computed in two steps. In the first step we obtain forecast errors by estimating the model in Equation 1 via OLS estimation. Then, we run an MCMC algorithm (Kastner and Fruhwirth-Schnatter, 2014) to generate posterior draws for (h ,ch,ϕh,σh,w ,cw,ϕw,σw) in Equation 2. j,1:T j j j k,1:T k k k To see how the forecast error variances fluctuate over time, consider a case with p = p = 1 2 p = 1. The one-step-ahead forecast error is v and its variance is 3 j,t Uy (1) = E y E y I 2 = exp(h ). j,t j,t+1 − j,t+1 t j,t+1 h i (cid:0) (cid:2) (cid:12) (cid:3)(cid:1) (cid:12) When h > 1, predictor uncertainties also play a role in measuring uncertainty. For example,

9 if h = 2, then, Uy (2) = E y E y I 2 = φ2exp(h )+Γ ΣZ Γ +exp(h ) j,t j,t+2 − j,t+2 t j j,t+1 0j t+1t j j,t+2 | h i (cid:0) (cid:2) (cid:12) (cid:3)(cid:1) whereΣZ istheforecasterrorvaria (cid:12) nce-covariancematrixfortheone-step-aheadprediction t+1t | made for Z at time t. As the above equation reveals, the two-step-ahead prediction error t+1 variance depends also on any uncertainty variation from predicting Z . The same logic t+1 applies to the case with h > 2.

10 3 Real effects of uncertainty shocks To investigate the dynamic relationships between our uncertainty measure and aggregate economic activity, we fit VAR models to monthly Korean data from 1991M10 to 2014M12. Our main focus is to study the effects of uncertainty shocks on economic activity. To identify the uncertainty shocks, we use a Cholesky decomposition with our uncertainty measure ordered first (Baker et al., 2017; Choi and Shim, 2016). The main VAR specification includes threelagsofthelogoftheKOSPIindex,thepolicyrate(overnightcallrate),logemployment, andlogindustrialproduction. Asacomparison, wealsoidentifytheuncertaintyshocksusing both the VKOSPI index or the EPU index. Figure 1 presents the impulse response functions of the identified uncertainty shocks using our economic uncertainty index (UI, blue) and using the EPU index (EPU, red). These two shocks are identified based on separate estimated VAR models. Bands around the thick lines are 90% confidence sets. There are significantly negative and prolonged effects of uncertainty shocks to the KOSPI index, employment, and industrial production based on our economic uncertainty index. On impact, the policy rate increases for 6 months and remains positive for one year. These positive responses are due to the so-called the “flight to safety” motive where the central bank increases the policy rate to prevent capital outflows (Gourio et al., 2014; Choi, 2016; Rey, 2016). On the other hand, uncertainty shocks based on the EPU index have very little (and insignificant) impact on all other variables. The signs of these impacts on employment and industrial production are negative, but their magnitudes are small and insignificant. The impact on the policy rate is almost zero for all horizons. There is a negative effect on the stock price index on impact, but the response becomes positive after 8 months. Figure 2 shows impulse response functions of uncertainty shocks based on the VKOSPI index (Blue) and the EPU index (Red). The effects of uncertainty shocks based on the VKOSPI are qualitatively and quantitatively similar to those from our uncertainty measure. As we argued in the main text, the VKOSPI and the EPU indices capture uncertainty aboutspecificaspectsoftheKoreaneconomy. Therefore, theymaynotserveasacomprehensive uncertainty measure. In addition, uncertainty originated from the financial market may have different real effects than those originated from other sources (Ludvigson et al., 2015; Shin and Zhong, 2016; Carriero et al., 2016). To disentangle these effects, we re-estimate our VAR model by including both the VKOSPI and our economic uncertainty measure. To make sure that we separate the uncertainty shocks originating from the financial market from uncertainty variations due to other sources, we order the VKOSPI index first and our

11 Figure 1 Impulse responses of uncertainty shocks (separate estimation) uncertainty measure second. In this way, the second shock contains exogenous variation that does not move financial uncertainty contemporaneously.1 Figure 3 presents impulse responses of the two different uncertainty shocks. The red lines are impulse response functions of the uncertainty shocks originating from the financial market (financial uncertainty shocks) and the blue lines are impulse response functions of the uncertainty shocks that move the overall uncertainty index without affecting the VKOSPI contemporaneously (real activity uncertainty shocks). As we can see from the upper left panel, the impact of financial uncertainty shocks to the KOSPI index is significantly negativeforatleast7monthsandarearoundfourtimeslargerthantheimpactoftherealactivity uncertainty shocks contemporaneously. Unlike the results based on the separate identification of different uncertainty shocks, it turns out that the “flight to safety” story only holds for the real activity uncertainty shocks. The financial uncertainty shocks actually decrease the policy rate on impact. The real activity uncertainty shocks have a much larger and more 1Thisorderingassumptionisquiteimportantbecauseouruncertaintymeasurecontainsuncertaintyfrom theKOSPIindex. However,changingtheorderofvariablesdoesnotalterourmainresults. Seetheappendix for robustness checks.

12 Figure 2 Impulse responses of uncertainty shocks (separate estimation) persistent effect on real variables when compared to the financial uncertainty shocks. Caveat. Even though our results are quite robust as shown in the next section, we want to comment that these results have some limitations. First, unlike the U.S., Korea is a small open economy. Therefore, its international economic activities play an important and significant role. Second, when we include two uncertainties at the same time in our VAR model, we need to be careful about the interpretation of the two uncertainty shocks. Our ordering of variables (or exclusion restriction) decomposes unexpected movements in uncertainty measures into two pieces. One moves VIX and UI contemporaneously and another moves only UI within a month. Our categorization of the two uncertainty shocks (financial versus real activity uncertainty shocks) comes from the additional assumption that any uncertainty shock that moves both the VIX and UI contemporaneously (within a month) originated from the financial market. However, there may be an uncertainty shock that originated from another source that has an impact on the financial market in a month. For these reasons, we view our VAR exercise as a convenient way to present and compare the dynamic relationships between the uncertainty measures and other key macroeconomic variables.

13 Figure 3 Impulse responses of uncertainty shocks (both measures are included in VAR) 4 Robustness checks for VAR analysis In this section, we present other VAR results where 1. we include two lags as opposed to three lags and include a time trend (Figure 4 and Figure 5). 2. we show that the ordering does not matter for conclusion made from results in Figure 3. Our economic uncertainty measure is ordered first and the VKOSPI index is ordered second (Figure 7). 3. weincludeboththeEPUandtheoureconomicuncertaintymeasureintheVARmodel. Impacts of our uncertainty shocks do not change much from our baseline estimation results (Figure 6).

14 Figure 4 Impulse responses of uncertainty shocks (separate esimation with 2 lags and time trend) Figure 5 Impulse responses of uncertainty shocks (separate esimation with 2 lags and time trend)

15 Figure 6 Impulse responses of uncertainty shocks (both measures are included in VAR) Figure 7 Impulse responses of uncertainty shocks (both measures are included in VAR)

16 References Baker, S., N. Bloom, and S. Davis (2017): “Measuring Economic Policy Uncertainty,” Quarterly Journal of Economics, Forthcoming. Carriero, A., T. Clark, and M. Marcellino (2016): “Measuring Uncertainty and Its Impact on the Economy,” Working paper. Choi, S. (2016): “The Impact of US Financial Uncertainty Shocks on Emerging Market Economies: An International Credit Channel,” Working paper. Choi, S. and M. Shim (2016): “Financial vs. Policy Uncertainty in Emerging Economies: Evidence from Korea and the BRICs,” Working paper. Gourio, F., M. Siemer, and A. Verdelhan (2014): “Uncertainty Betas and International Capital Flows,” Working paper. Han, H., A. Kutan, and D. Ryu (2015): “Effects of the US Stock Market Return and Volatility on the VKOSPI,” Economics: The Open-Access, Open-Assessment E-Journal, 9, 1–34. Jurado, K., S. C. Ludvigson, and S. Ng (2015): “Measuring Uncertainty,” American Economic Review, 105, 1177–1216. Kastner, G. and S. Fruhwirth-Schnatter (2014): “Ancillarity-sufficiency interweaving strategy (ASIS) for boosting MCMC estimation of stochastic volatility models,” Computational Statistics Data Analysis, 76, 408–423. Ludvigson, S. C., S. Ma, and S. Ng (2015): “Uncertainty and Business Cycles: Exogenous Impulse or Endogenous Response?” NBER Working Papers 21803, National Bureau of Economic Research, Inc. Rey, H. (2016): “International Channels of Transmission of Monetary Policy and the Mundellian Trilemma,” IMF Economic Review, 64, 6–35. Shin, M. and M. Zhong (2016): “A New Approach to Identifying the Real Effects of Uncertainty Shocks,” Working paper.