14
The effects of government spending shocks on the trade account balance in Korea Hyeongwoo Kim a, * , Daeyup Lee b a Department of Economics, Auburn University, 217 Miller Hall, Auburn, AL 36849, United States b Economic Research Institute, Bank of Korea, Seoul, South Korea ARTICLE INFO JEL classication: E32 E62 F32 F41 Keywords: Fiscal policy Trade account Twin decit Twin divergence Real exchange rate VAR ABSTRACT This paper investigates scal policy effects on the trade account balance in Korea using open economy vector autoregressive (VAR) models. In response to scal spending shocks, real appre- ciations occur, followed by deteriorations in the trade account balance with the pre-2000 sample period, while real depreciations occur leading to an improvement in the trade account balance with the post-2000 sample period. We explain this structural break in relation to Korea's transition from a tightly managed exchange rate regime to a more market-oriented oating system after Korea's foreign exchange crisis in 199798. Furthermore, such scal policy effects from our post- 2000 sub-sample imply that expansionary scal policy can be effective in stimulating private sector economy in Korea. 1. Introduction This paper empirically investigates the effects of scal expansions on the trade account balance in Korea using an open economy vector autoregressive (VAR) model. Specically, we study how the trade account balance responds to government spending shocks through the propagation mechanisms of the shock to other key macroeconomic variables such as consumption and the real exchange rate in Korea. We often discuss scal policy effects on the current account balance and the trade account balance interchangeably, since dynamic responses of these two series to the scal shock are qualitatively similar. This is because their business cycle components, which are relevant to VAR analysis, tend to exhibit very high correlation, even when the raw data show sizable differences over time. 1 Standard economic theories (e.g., Baxter, 1995; Kollmann, 1998; Erceg, Guerrieri, and Gust, 2005) provide straightforward ex- planations for the so-called twin decit hypothesis. When government spending increases with no matching increases in tax revenues (scal decit), Ricardian consumers, rationally expecting a tax hike in the near future, reduce consumption (increase saving) and increase labor hours. As capital becomes more productive due to an increase in labor hours, investment rises which then offsets the increase in private saving. Therefore, the current account deteriorates (current account decit), resulting in a twin decit. The views expressed herein are those of the authors and do not necessarily reect the ofcial views of the Bank of Korea. When reporting or citing this paper, the authors' names should be explicitly stated. * Corresponding author. E-mail addresses: [email protected] (H. Kim), [email protected] (D. Lee). 1 In the case of the US, the difference is very small, so the current account and the trade account are often used interchangeably. Contents lists available at ScienceDirect International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref https://doi.org/10.1016/j.iref.2017.10.001 Received 18 April 2017; Received in revised form 27 September 2017; Accepted 2 October 2017 Available online 9 October 2017 1059-0560/© 2017 Elsevier Inc. All rights reserved. International Review of Economics and Finance 53 (2018) 5770

The effects of government spending shocks on the … · The effects of government spending shocks on the trade ... Standard economic theories (e.g., Baxter ... When government spending

Embed Size (px)

Citation preview

International Review of Economics and Finance 53 (2018) 57–70

Contents lists available at ScienceDirect

International Review of Economics and Finance

journal homepage: www.elsevier .com/locate/ i ref

The effects of government spending shocks on the trade accountbalance in Korea☆

Hyeongwoo Kim a,*, Daeyup Lee b

aDepartment of Economics, Auburn University, 217 Miller Hall, Auburn, AL 36849, United StatesbEconomic Research Institute, Bank of Korea, Seoul, South Korea

A R T I C L E I N F O

JEL classification:E32E62F32F41

Keywords:Fiscal policyTrade accountTwin deficitTwin divergenceReal exchange rateVAR

☆ The views expressed herein are those of the authorsauthors' names should be explicitly stated.* Corresponding author.E-mail addresses: [email protected] (H. Kim), dy

1 In the case of the US, the difference is very small, s

https://doi.org/10.1016/j.iref.2017.10.001Received 18 April 2017; Received in revised form 27 SAvailable online 9 October 20171059-0560/© 2017 Elsevier Inc. All rights reserved.

A B S T R A C T

This paper investigates fiscal policy effects on the trade account balance in Korea using openeconomy vector autoregressive (VAR) models. In response to fiscal spending shocks, real appre-ciations occur, followed by deteriorations in the trade account balance with the pre-2000 sampleperiod, while real depreciations occur leading to an improvement in the trade account balancewith the post-2000 sample period. We explain this structural break in relation to Korea's transitionfrom a tightly managed exchange rate regime to a more market-oriented floating system afterKorea's foreign exchange crisis in 1997–98. Furthermore, such fiscal policy effects from our post-2000 sub-sample imply that expansionary fiscal policy can be effective in stimulating privatesector economy in Korea.

1. Introduction

This paper empirically investigates the effects of fiscal expansions on the trade account balance in Korea using an open economyvector autoregressive (VAR) model. Specifically, we study how the trade account balance responds to government spending shocksthrough the propagation mechanisms of the shock to other key macroeconomic variables such as consumption and the real exchangerate in Korea. We often discuss fiscal policy effects on the current account balance and the trade account balance interchangeably, sincedynamic responses of these two series to the fiscal shock are qualitatively similar. This is because their business cycle components, whichare relevant to VAR analysis, tend to exhibit very high correlation, even when the raw data show sizable differences over time.1

Standard economic theories (e.g., Baxter, 1995; Kollmann, 1998; Erceg, Guerrieri, and Gust, 2005) provide straightforward ex-planations for the so-called twin deficit hypothesis. When government spending increases with no matching increases in tax revenues(fiscal deficit), Ricardian consumers, rationally expecting a tax hike in the near future, reduce consumption (increase saving) andincrease labor hours. As capital becomes more productive due to an increase in labor hours, investment rises which then offsets theincrease in private saving. Therefore, the current account deteriorates (current account deficit), resulting in a twin deficit.

and do not necessarily reflect the official views of the Bank of Korea. When reporting or citing this paper, the

[email protected] (D. Lee).o the current account and the trade account are often used interchangeably.

eptember 2017; Accepted 2 October 2017

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

On the other hand, most undergraduate level macroeconomics textbooks such as Mankiw (2014) provides an alternative but simplerexplanation on the twin deficit hypothesis. As the government spending increases, national saving decreases, which results in anincrease in the real interest rate. Higher real interest rate then attracts more capital toward the domestic capital market, resulting ina decrease in the net capital outflow. At the same time, the decrease in the net capital outflows causes the real exchange rate to go up(real appreciation), which deteriorates the trade account balance.

On the contrary, Kim and Roubini (2008) report a short-run improvement in the current account balance in response to the fiscaldeficit shock using the current floating exchange rate regime data in the US. The real interest rate increases as the public saving de-creases, which then lowers private investment. Private saving rises due to a decrease in consumption. Overall, the increase in privatesaving and the decrease in domestic investment jointly outweigh the decrease in public saving, leading to an improvement in the currentaccount balance. They named this empirical finding the twin divergence hypothesis. Their model, however, does not explain why the realexchange rate depreciates, which is crucial to explain the improvement in the trade account balance, even though the real interest rateincreases in their model in response to a positive fiscal shock.

Jia and Kim (2016b) revisit this issue introducing the role of consumer sentiment. Their VAR model estimates imply that privatespending declines as the fiscal shock generates consumer pessimism. That is, unexpected increases in the fiscal spending confirmincoming decreases in productivity in the near future (news effect). Anticipating recessions, private sector agents reduce both privateconsumption and investment. Unlike Kim and Roubini (2008), they report a decrease in the real interest rate, which then results in a realdepreciation. The trade account balance improves accordingly. Therefore, their work confirms Kim and Roubini’s (2008) twin diver-gence hypothesis, though their model also explains why the real exchange rate depreciates leading to an improvement in the tradeaccount balance as well.

In addition to these studies, Müller (2008), Corsetti and Müller (2008), Ravn, Schmitt-Groh�e, and Uribe (2007), Beetsma, Giuliodori,and Klaassen (2008), Kollmann (2010), Enders, Müller, and Scholl (2011), and Kim (2015) also investigated fiscal policy effects on thereal exchange rate (terms of trade) and/or the trade account balance.

In the present paper, we study the fiscal policy propagation mechanism from Korean experiences that adds new insights to thecurrent literature. Among earlier related research work, Kim and Kim (2006) implemented the Granger Causality test for the govern-ment budget balance, the current account balance, and the exchange rate using Korean data. Unfortunately, their empirical findings arenot very informative.2

We employ open economy VAR models for quarterly observations from 1980:I to 2015:I to investigate how fiscal shocks influencethe trade account in Korea. We obtained strong evidence of a structural break in the dynamic relationship between the governmentspending and the trade account balance around the year 2000 when Korea completed economic overhauls since Korea's foreign ex-change crisis in the late 1990's. The trade account balance deteriorates in response to the fiscal spending shock only whenwe employ thepre-2000 data. We obtained similar responses of the current account balance, which is consistent with the twin deficit hypothesis.

With the post-2000 sample period, however, the trade account balance improves in response to the fiscal shock, which is similar tothe US evidence reported by Kim and Roubini (2008). That is, in contrast to previous results, we observed real depreciations thatimproved the trade account balance.

We noticed that Jeong, Kang, and Kim (2017) reported similar findings for Korea and Japan in their three-country study for China,Korea, and Japan. That is, they also reported a structural break in the response of the trade account balance to the government spendingshock in Korea. However, they are primarily interested in comparing the fiscal policy effects/multipliers in the three countries.3

On the other hand, we focus on the fiscal policy effects on the trade (current) account balance in Korea using various models andspecifications. Furthermore, we provide detailed explanations on what caused this structural break in sections 3.3 and 4.1. Korea hasundergone a major overhaul since its financial crisis in 1997–98. In what follows, we clarify the role of more flexible prices in financialmarkets in understanding our findings about fiscal policy effect on the trade account in Korea, relating it to Korea's transition to a marketoriented exchange rate regime after the crisis. In response to fiscal policy shocks, substantial nominal depreciations occur dominatingrising domestic inflation rate, resulting in real depreciations under the flexible exchange rate regime in the 2000's but not in the pre-2000 sample period.

We also implement an array of robustness check analysis which confirm our major findings from the baseline model. Also, using thepost-2000 data, we provide some evidence of effective fiscal policy in stimulating private activity in Korea via responses of consumption,consumer sentiment, and government tax receipts.

The rest of the paper is organized as follows. Section 2 describes the econometric model. Section 3 provides data descriptions andpreliminary analysis, then we report our major empirical findings. Section 4 provides further discussions and robustness check analysis.Section 5 concludes.

2. The empirical model

We employ the following pth order vector autoregressive (VAR) model for open macroeconomy data in Korea. Abstracting from

2 They report that the fiscal balance fails to Granger cause the current account, whereas the current account Granger cause the fiscal budget. Since Granger Causalitydoes not imply causal relationships, their findings provide not much useful information as to the validity of the twin deficit hypothesis in Korea.

3 In a similar study, Ilzetzki, Mendoza, and Vegh (2013) reported that the fiscal multiplier tends to be large in economies under the fixed exchange rate regime. Forcountries with flexible exchange rates, however, they reported a negative fiscal multipliers both in the short- and in the long-run, which is in sharp contrast with resultsof Jeong, Kang, and Kim (2017) who found substantial and significantly positive output effects of fiscal policy for Korea and Japan during flexible exchange rateregimes.

58

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

deterministic terms,

xt ¼ Φ1xt�1 þΦ2xt�2 þ⋯þΦpxt�p þ Cut; (1)

where

xt ¼ ½ gt z0t �0

and C is a lower-triangular matrix and ut is a vector of mutually orthonormal structural shocks, that is, Eutu0t ¼ I.4 Our key variable gt

denotes a measure of the government spending (govt) such as the government consumption. For our baseline model, zt is the following.

zt ¼ ½ yt nxtt nirt πt rert �0; (2)

where yt is a private spending variable such as private consumption (cont) or private investment (ivtt). nxtt is the trade account balance asa share of the total GDP.5 nirt is the nominal interest rate and πt denotes the inflation rate. Lastly, rert denotes the real effective exchangerate. We demean and detrend xt prior to estimations.

We are particularly interested in the j-period ahead orthogonalized impulse-response functions (OIRF) of zt to the governmentspending shock, defined as follows.

IRFðjÞ ¼ E�ztþj

��ug;t ¼ 1;Ωt�1

�� E�ztþj

��Ωt�1

�; (3)

where ug,t is the structural shock to gt in (1) and Ωt�1 is the adaptive information set at time t � 1.6

Note that gt is ordered first in xt, meaning that the fiscal policy variable is not influenced by other variables in zt within one quarter.This assumption is often employed in the current fiscal policy literature (e.g., Ramey, 2011), which seems to be an innocuousassumption, because implementation of discretionary fiscal policy normally requires congressional approval.

It is well documented that econometric inferences based on recursively identified VARmodels might not be robust to alternative VARorderings. Our major findings, however, are not subject to this problem. As Christiano, Eichenbaum, and Evans (1999) show, allresponse functions to the fiscal spending shock are invariant because gt is ordered first. Put it differently, all response functions to thefiscal shock are numerically identical even when one randomly rearranges variables in zt.

3. Empirical results

3.1. Data description

We obtained all data from the Bank of Korea unless stated otherwise. Observations are quarterly and span from 1980:I to 2015:I. nxttis the trade account balance as a share of the total GDP (gdpt), which is the exports (extt) of goods and services minus the imports (imtt) ofgoods and services. The current account balance (curt) is also defined as a share of gdpt.

All public and private spending variables, the government consumption (govt), the private consumption (cont), and the total in-vestment (ivtt), are divided by the GDP deflator, then log-transformed. The BIS real effective exchange rate (rert) is obtained from theFRED and log-transformed. The nominal interest rate (nirt) is the money market interest rate, obtained from the IFS CD-ROM. Weobtained the core Consumer Price Index (CPI) inflation rate from the same source.

Our baseline VAR models utilize govt to identify the fiscal spending shock. In the current literature, however, researchers haveemployed alternative measures of the fiscal spending variables such as the government budget balance (or deficit) or the government nettax to control for the tax effect. Unfortunately, the real net tax (taxt) data is available only for the sample period that begins in 2000:I.

In what follows, however, we show that empirical evidence from an extended VAR model with taxt provides qualitatively similarresults. And Kim (2015) discussed the benefits of using the government consumption shock in the current literature of the twin deficit/divergence. Therefore, our identification strategy with the government consumption seems to the best option available for a Korean casestudy, even though it is not an ideal choice.

3.2. Preliminary analysis

Fig. 1 reports the Hodrick-Prescott (HP) cyclical (business cycle) components of our key variables: government consumption (govt),total GDP (gdpt), the current account balance (share of GDP; curt), the trade account balance (share of GDP; nxtt), and the real effectiveexchange rate (rert).7 We also report correlations of these variables in Table 1. Some of notable findings are as follows.

First, we find empirical evidence of counter-cyclical fiscal policy in Korea only for the post-2000 sample period (2000:I-2015:I).

4 We employed an intercept and linear time trend for our VAR estimations. Our estimates are qualitatively very similar when we add quadratic trend term to thosedeterministic terms.

5 We also employ the current account balance (curt) instead of the nxtt, which yielded qualitatively similar results.6 That is, the information set has the following property, Ωt�1⊇Ωt�2⊇Ωt�3⊇⋅⋅⋅.7 We set the smoothing parameter at 1600 following the suggestion for quarterly data by Hodrick and Prescott (1997).

59

Fig. 1. Hodrick-Prescott Cyclical Components of the Data.Note: We extracted the cyclical components of the data via the Hodrick-Prescott filter with a smoothing parameter of 1600 for quarterly data.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

When the economy enters a period of recession, government may implement expansionary fiscal policy that may come in the form ofincreased government spending. As the economy improves, the government may switch to contractionary policy to reduce its fiscaldeficit and possibly to dampen inflationary pressure. In the pre-2000 era (1980:1–1999:IV), however, the correlation between govt andgdpt was negligibly small (�0.045), whereas it has increased to � 0.131 in absolute value during the post-2000 era. That is, we find ansubstantial role of output stabilization in the conduct of fiscal policy in Korea only from observations since 2000.

Second, as we can clearly see in Fig. 1, curt and nxtt exhibit very strong co-movement. The correlations were consistently high, 0.777,0.811, and 0.696 in the full, the pre-2000, and the post-2000 sample periods, respectively.8 On the other hand, nxtt (or curt) and rertmove in opposite directions, which is not surprising. Correlations were �0.602, �0.629, and �0.551 in the full, the pre-2000, and thepost-2000 sample periods, respectively. In response to a sharp real depreciation triggered by Korea's foreign exchange crisis in 1997, nxttand curt both exhibit an abrupt and rapid hike. Similar dramatic changes are observed in around 2008 when the US Financial Crisis

8 The difference between the current account and the trade account balance is the net amount received for domestically-owned factors of production used abroad.

60

Table 1Correlation estimations.

1980:I-2015:I

govt gdpt curt nxtt rert

govt 1.000gdpt �0.066 1.000curt �0.188 �0.496 1.000nxtt �0.138 �0.694 0.777 1.000rert 0.157 0.412 �0.555 �0.602 1.000

1980:I-1999:IVgovt gdpt curt nxtt rert

govt 1.000gdpt �0.045 1.000curt �0.273 �0.481 1.000nxtt �0.189 �0.710 0.811 1.000rert 0.342 0.392 �0.613 �0.629 1.000

2000:I-2015:Igovt gdpt curt nxtt rert

govt 1.000gdpt �0.131 1.000curt 0.011 �0.590 1.000nxtt 0.011 �0.627 0.696 1.000rert �0.272 0.523 �0.434 �0.551 1.000

Note: We used the Hodrick-Prescott cyclical components of the data with a smoothing parameter of 1600 for quarterly data.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

began, triggered by the collapse of Lehman Brothers in September 2008. We also note that nxtt exhibit counter-cyclical movements. Thatis, it seems that the trade account balance tends to improve in Korea during recessions. This is not necessarily an improvement, however,because Korean imports tend to fall more than exports during global recessions.

Most importantly, we obtained negative correlation between govt and nxtt (or curt) only from the pre-2000 era, while very smallpositive correlation was observed when we use the post-2000 observations. As we discussed before, Baxter (1995), Kollmann (1998),and Erceg, Guerrieri, and Gust (2005) among others predicted a deterioration of the current/trade account balance when there is a fiscalexpansion shock (twin deficit), whereas Kim and Roubini (2008) proposed an alternative hypothesis of the twin divergence based on theirempirical findings for the post-war US data.9 That is, we observe evidence in favor of the twin deficit hypothesis from the pre-2000sample period, while the data supports the twin divergence hypothesis for the latter sample period.

To put in a nutshell, we obtained strong empirical evidence of a negative (static) relationship between the government spending andthe trade account balance only for the earlier sample period. We also noticed a similar structural break between the governmentspending and the real exchange rate. In what follows, we investigate the dynamic effects of the fiscal expansion shock to the trade (andcurrent) account balance in Korea using an array of open economy VAR models that are robust to alternative orderings.

3.3. Government spending shocks and the trade account

We report dynamic responses of the trade account balance to the government spending (gt ¼ govt) shock via the orthogonalizedimpulse-response function (OIRF) estimate for the VAR model (1) and (2). We employed an intercept and linear time trend for esti-mations.10 Our benchmark model employs the private consumption (cont) for the private spending yt. Using the total investment (invt)instead produces qualitatively similar results.11 We also estimated the VARmodel that replaces the trade account balance (nxtt) with thecurrent account balance (curt), which again provided qualitatively similar results.12 All OIRF estimates are accompanied by one standarderror confidence bands that are obtained by 500 nonparametric bootstrap simulations.

Fig. 2 provides strong empirical evidence of a negative dynamic relationship between the govt and nxtt from the full sample period.Since the impact, the trade account balance (nxtt) starts to deteriorate persistently and significantly in response to the fiscal expansionshock. Initial adjustments are somewhat sluggish and the maximum response occurs in about 1.5 years. We also obtained quite similardynamics for curt, which is consistent with the twin deficit hypothesis.

The real exchange rate exhibits persistent increases (real appreciations), which explains the negative responses of the trade accountbalance. The responses are overall statistically insignificant. It turns out, however, that this is due to pooling the two sub-sample periodstogether that generate qualitatively different responses of the real exchange rate to the fiscal shock.

9 Kim (2015) reported real depreciations in response to the government consumption shock using a panel VAR framework for developed countries. He reportedsomewhat mixed evidence as to its effect on the trade account balance.10 We obtained qualitatively very similar estimates when quadratic trend term is added to those deterministic terms. All results are available upon request.11 The “total” investment is the sum of the private sector gross investment and the government gross investment. It is ideal to use the private investment, but the dataavailability is limited for the post-2000 period.12 All results are available from authors upon requests.

61

Fig. 2. IRF Estimates to the Fiscal Spending Shock: 1980:I-2015:I.Note: We obtained one standard deviation confidence bands (dashed lines) by 500 nonparametric bootstrap simulations.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

In Figs. 3 and 4, we report OIRF estimates to the fiscal spending shock from the two sub-sample periods in order to investigatedynamic evidence of a structural break between govt and nxtt. When we employ the pre-2000 sample period, see Fig. 3, the trade accountbalance decreases persistently and significantly in response to the fiscal spending shock, reaching the maximum response in around 1year. Also, the fiscal spending shock yields real appreciations of Korean won, which explains the deterioration of the trade accountbalance. Both the nominal interest rate (nirt) and inflation (πt) rise when the fiscal shock occurs. It should be noted that private con-sumption (cont) shows persistent increases that are highly significant, which is at odds with Ricardian consumers’ behavior. That is,consumers in Korea decrease private saving when the government implements expansionary fiscal policy. This might happen if con-sumer confidence rises following expansionary fiscal policy as shown by Bachmann and Sims (2012).

This strong evidence in favor of the twin deficit hypothesis disappears when we use the post-2000 observations as can be seen inFig. 4. Responses of nxtt to the government consumption shock are overall significantly positive, although weaker than those from theearlier sample period. When fiscal spending shock occurs, the trade account balance increases for a short-period of time followed bya short correction, then go up again for over 5 years. We obtained very similar OIRF estimates from a VAR model with the currentaccount balance (curt) that replaces nxtt.

Interestingly, Kim and Roubini (2008) also find similar short-run improvement of the current account balance to the fiscal shock forthe US. 13 That is, we observe this so-called twin divergence only for the sample period when Korea transitioned into a highly marketoriented economy after Korea's foreign exchange crisis in late 1990's. Responses of other variables, cont, nirt, and πt are qualitativelysimilar as those from the earlier observations.

We also investigate this evidence of a structural break in the effect of the fiscal spending shock on the trade account balance via theforecast error variance decomposition (FEVDEC) analysis for nxtt. Results are reported in Fig. 5.

The most notable finding is that govt plays an important role in explaining the k-period ahead forecast error variance for nxtt only inthe pre-2000 era. That is, govt explains roughly 25% in the long-run, which is the second important factor next to cont. However, govt'srole becomes negligible in the post-2000 period. That is, the share of govt falls to around 10% in the long-run. Putting it differently, theeffect of the government spending shock becomes a minor factor in predicting the trade account balance in a more market orientedKorean economy since 2000. That is, the FEVDEC analysis is also consistent with our findings on a structural break reported earlier.

Also, we notice that cont has been always an important factor in explaining variations in future nxtt. This is because that contmoves intandemwith the total GDP. Replacing nxtt with the GDP yields qualitatively similar results.14 We also note that the nominal interest rate(nirt) plays an important role in the latter sample period as its share increases from around 3%–19%. This may reflect a drastic overhaulof financial regulatory system in Korea since the outburst of the Korean Financial Crisis in late 1990's, which eliminated direct

13 They use the government deficit, government spending minus the tax revenue, to identify the fiscal spending shock. We also implemented similar investigationsreplacing the government consumption with the government deficit. Results were very similar and are available upon requests.14 Results are available upon requests.

62

Fig. 3. IRF Estimates to the Fiscal Spending Shock: 1980:I-1999:IV.Note: We employed the first 80 observations to estimate the fiscal spending shock effects. Dashed lines are 95% confidence bands obtained from 500 nonparametricbootstrap simulations.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

regulations of market interest rates. When interest rates are mainly determined by the market, nirt may play an important role inexplaining the trade account via its effect on the nominal exchange rate, thus the real exchange rate. 15

We further investigate this issue by estimating the time-varying relationship between govt and nxtt. For this, we employ a 20-yearfixed-size rolling window scheme for our VAR model in (1). That is, we repeatedly estimate the OIRFs with a 20-year memory windowstarting from 1980:I, which helps identify possible structural changes in the data generating process if any.

The first column in Fig. 6 provides three-dimension surface graphs for the OIRF of the trade account balance and the real exchangerate to the government spending shock. Similar estimates with the current account balance are reported in the second column. Lightcolored areas of the response functions indicate negative responses, while darker areas are positive responses. The first set of theresponse function estimates on the Date 2000 are obtained from observations between 1980:I and 1999:IV, while the last responsefunction estimates on Date 2015 come from the sample period from 1996:II to 2015:I.

The surface graph of the trade account as well as that of the current account balance clearly show that empirical evidence in favor ofthe twin deficit hypothesis disappears when estimations are implemented with newer observations. In fact, when the last set of 20-yearlong observations are used, the response function of nxtt remains positive since the impact, implying that expansionary fiscal policyimproves the trade account balance. The current account balance also overall increases in response to the fiscal shock, which is con-sistent with the twin divergence hypothesis.

The lower panel graphs exhibit dramatic changes in the response of the real exchange rate to the government spending shock in bothVAR models. Response surface functions are qualitatively very similar whether we use the trade account or the current account in theVARmodel. These findings are consistent with those of Kim (2015) who reported strong panel evidence of real depreciations in responseto fiscal shocks in developed countries under the flexible exchange rate regime. Overall, we obtain solid evidence that the fiscal spendingshock tends to generate real appreciations of Korean won in the earlier period, whereas it yields real depreciations in the latter sampleperiod.

4. Further discussion

4.1. Changes in exchange rate regimes

Previous section presented a possibility of a structural break in the propagation of fiscal policy to the trade account around the year2000. This subsection discusses possible explanations on this issue in relation to a change in foreign exchange rate regimes in Koreaaround that period of time.

15 Higher nominal interest rates imply a greater return on the domestic currency, which tends to result in appreciations of the currency.

63

Fig. 4. IRF Estimates to the Fiscal Spending Shock: 2000:I-2015:I.Note: We employed the last 80 observations to estimate the fiscal spending shock effects. Dashed lines are 95% confidence bands obtained from 500 nonparametricbootstrap simulations.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

It is well-known that Korea had maintained a tightly managed foreign exchange rate system until Korea's foreign exchange crisis in1997. Korea's exchange rate policy has focused mainly on the US dollar exchange rate because most Korea's international trade wasconducted in the US dollar. As can be seen in Fig. 7, the US dollar exchange rate, Korean won price of 1 US dollar, exhibits steady longswings around 800 until Korea got hit by massive speculative attacks in 1997. Similar pattern can be also observed in the HP cyclical(business cycle) component.

Since this foreign exchange crisis, Korean government pursued a major economic overhaul based on recommendations of the In-ternational Monetary Fund (IMF) that provided a bailout program. Especially, Korea's foreign exchange rate regime has transitionedfrom a tightly managed float system to a more market-oriented floating exchange rate system. Further, Korean government grantedforeign investors broad access to domestic financial markets. See Dooley, Dornbusch, and Park (2002) for detailed explanations.

In order to appraise the impact of this change in exchange-rate regimes on our analysis, we implemented VAR estimations for the twosub-sample periods, 1980:I to 1996:IV (Pre-Crisis) and 2000:I to 2015:I (Post-Crisis). We report responses of the dollar exchange rate,which is transformed to the (log) dollar price of one unit of Korean won to be consistent with our previous analysis. Results are reportedin Fig. 8.

We first note that the response of the dollar exchange rate is a lot weaker in the pre-crisis period than in the post-crisis period, whichis consistent with the regime change in Korea's exchange rate system. Inflation rises in response to the fiscal spending shock immediatelyin the pre-crisis period and with a lag in the later period, which generates an upward pressure to the real exchange rate of Korean won.Nominal interest rates move in accordance with inflation, confirming the Fisher effect.

In the pre-crisis period, Korean won appreciates slightly, though overall insignificantly, reflecting an increase in the nominal interestrate. Note that inflation rises in response to the fiscal spending shock, leading towards a real appreciation of Korean won. Real ap-preciations then deteriorate the trade account balance.

In the post-crisis period, we noticed that Korean won depreciates substantially in response to unexpected increases in the govern-ment spending. After a brief recovery, which is statistically insignificant, Korean won overall stays depreciated for a while. As can beseen in Fig. 8, the responses of the nominal dollar exchange rate dominate the responses of inflation, resulting in real depreciations ofKorean won in response to the fiscal spending shock, which explains an improvement of the trade account balance in the later period.

Taken all together, this change in the exchange rate regimes in Korea implies a stronger role of the nominal exchange rate indetermining the real exchange rate in the post-crisis period, and provide an explanation on our empirical evidence in favor of the twindeficit hypothesis only for the pre-2000 era, but not for the post-2000 era.

4.2. Robustness analysis with augmented VAR models

Lastly, we provide robustness check analysis via augmented VAR models with three additional macroeconomic variables:

64

Fig. 5. Forecast error variance decomposition of the trade Account balance.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

government tax receipts (taxt), consumer sentiment (sentt), and private investment (pinvt). That is, we estimate OIRFs from the VARspecification in (1) and (3) with the following.

xt ¼ ½ govt z0t at �0; (4)

where at is one of the three additional variables. These three variables are available from 2000:I, so we estimate and report OIRFs onlyfor the post-2000 sample period. It should be noted that the ordering of at does not matter as long as it is placed next to govt, because weare interested only in the OIRFs to the government spending shock as explained in Section 2.

Fig. 9 reports responses of all 6 variables plus taxt to the fiscal spending shock.We note OIRFs in Fig. 10 confirm our previous findingsin Fig. 5 even after controlling for the tax revenues. The trade account balance overall improves in response to the fiscal shock, and rertpersistently declines resulting in significant real depreciations.

taxt increases after about a half-year delay. To understand why, we look at the response of cont that is highly positively correlatedwith the private GDP. It seems that the fiscal spending shock stimulates private activity as can be seen in persistent and significantincreases in cont for about two years. Increases in tax revenues imply a sustainable fiscal stimulus policy in Korea because suchendogenous increases in taxt would help balance the government budget.16

Fig. 10 presents OIRF's from a VAR model that employs at ¼ sentt. Results again confirm our findings in Fig. 4.

16 We also implemented similar analysis using the fiscal deficit variable, which is the government consumption minus tax. That is, instead of using it as a separatecontrol variable, tax is incorporated to the policy variable. We obtained similar results that are available upon requests. We also used the government investment insteadof the government consumption for a policy variable. Again, we obtained similar findings and results are available upon requests.

65

Fig. 6. Fiscal Spending Shock Effects: Fixed-Size Rolling Window Scheme.Note: We repeat estimations of the impulse-response functions using a 20-year fixed-size rolling windowsscheme.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

We obtain strong evidence that shows an improvement of the trade balance (and the twin divergence) as well as real depreciations ofKorean won when the fiscal shock occurs. The response of sentt implies that consumers become optimistic in the short-run when govtincreases unexpectedly, which might explain the increases in cont. Bachmann and Sims (2012) also reported similar consumer optimismin response to the expansionary fiscal shock during recessions in the US, while Jia and Kim (2016a) argue that fiscal spending shockstend to generate consumer pessimism using an array of alternative identification schemes.

Lastly, we employed at ¼ pinvt and report OIRFs in Fig. 11. Interestingly, the response of pinvt exhibits a mirror image of the responseof the nxtt. This is consistent with the response of curt (available upon requests) that implies the twin divergence. Since cont increaseswhen the fiscal spending shock occurs, domestic saving, both the private and the public saving, must decrease. Then, investment mustfall by more than savings to be consistent with an improvement of the current account balance.

5. Conclusion

This paper investigates the effects of the fiscal expansion shock on the trade account balance in Korea using an open economy VARmodel framework. Finding a strong empirical support for the twin divergence hypothesis may imply that fiscal expansions can be aneffective tool to stimulate the economy during recessions for a small open economy like Korea. The twin deficit hypothesis, however,suggests that policy makers should be cautious in implementing expansionary fiscal policy because it may deteriorate the trade accountbalance, which may not be negligible cost for a small open economy.

Our impulse-response function analysis reveals a time-varying relationship between the government spending and the trade accountbalance over time. We obtained persistent and statistically significant decreases of the trade account balance as well as the currentaccount balance with the pre-2000 sample period, whereas the trade account balance overall improves when we use the post-2000 data,

66

Fig. 7. US dollar exchange rates.

Fig. 8. IRF Estimates to the Fiscal Spending Shock for the Pre- and the Post-Crisis Periods.Note: The sample period is between 1980:I and 1996:IV for the pre-crisis period, whereas the post-crisis period is between 2000:I and 2015:I. We obtained one standarddeviation confidence bands (dashed lines) by 500 nonparametric bootstrap simulations.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

67

Fig. 9. IRF Estimates to the Fiscal Spending Shock with Tax Revenues.Note: The sample period is between 2000:I and 2015:I. We obtained one standard deviation confidence bands (dashed lines) by 500 nonparametric bootstrapsimulations.

Fig. 10. IRF Estimates to the Fiscal Spending Shock with Consumer Sentiment.Note: The sample period is between 2000:I and 2015:I. We obtained one standard deviation confidence bands (dashed lines) by 500 nonparametric bootstrap simulations.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

68

Fig. 11. IRF Estimates to the Fiscal Spending Shock with Private Investment.Note: The sample period is between 2000:I and 2015:I. We obtained one standard deviation confidence bands (dashed lines) by 500 nonparametric bootstrapsimulations.

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

which confirms the finding by Jeong, Kang, and Kim (2017).We explain this structural break as follows. Korea has implemented a drastic overhaul in financial system since its foreign exchange

crisis in 1997–98, which resulted in more market-oriented financial markets since then. For instance, Korea discarded their tightlymanaged float exchange rate system following the IMF's recommendation, and transitioned to a free floating system after the crisis.

Inflation responds positively to the fiscal expansion shock, resulting in an upward pressure on the real exchange rate. Sluggishadjustments of the nominal exchange rate under the managed float system result in real appreciations in response to the governmentspending shock, followed by deteriorations of the trade account balance. On the contrary, the nominal exchange rate depreciatedsharply when the fiscal spending shock occurs in the post-2000 era, which overall dominated inflation effects on the real exchange rate,resulting in an improvement in the trade account balance.

In addition, we employed a smooth transition VAR model via a 20-year fixed-size rolling window scheme. That is, we investigatedtime-varying fiscal policy effects on the real exchange rate and on the trade/current account balance under the rolling window scheme.Results again strongly confirm the existence of the aforementioned structural break.

We also implemented an array of extended VARmodels that confirm our major findings for the post-2000 sample period. We furtherreport evidence of effective fiscal policy in stimulating private spending in Korea via the OIRFs of consumption, consumer sentiment,and the government net tax.

References

Bachmann, R., & Sims, E. (2012). Confidence and the transmission of government spending shocks. Journal of Monetary Economics, 59(3), 235–249.Baxter, M. (1995). International trade and business cycles. In G. M. Grossmann, & K. Rogoff (Eds.), Handbook of international economics (Vol. 3, pp. 1801–1864).

Amsterdam: North-Holland.Beetsma, R., Giuliodori, M., & Klaassen, F. (2008). The effects of public spending shocks on trade balances and budget deficits in the European union. Journal of the

European Economic Association, 6, 414–423.Christiano, L. J., Eichenbaum, M., & Evans, C. L. (1999). Monetary policy Shocks: What have we learned and to what end?. In J. B. Taylor, & M. Woodford (Eds.),

Handbook of macroeconomics (Vol. 1, pp. 65–148) Elsevier.Corsetti, G., & Müller, G. J. (2008). Twin deficits, openness, and the business cycle. Journal of the European Economic Association, 6, 404–413.Dooley, M., Dornbusch, R., & Park, Y. C. (2002). A framework for exchange rate policy in Korea. Finance Working Papers 21757. East Asian Bureau of Economic Research.Enders, Z., Müller, G. J., & Scholl, A. (2011). How do fiscal and technology shocks affect real exchange rates? new evidence for the United States. Journal of International

Economics, 83(1), 53–69.

69

H. Kim, D. Lee International Review of Economics and Finance 53 (2018) 57–70

Erceg, C. J., Guerrieri, L., & Gust, C. (2005). Expansionary fiscal shocks and the U.S. Trade deficit. International Finance, 8(3), 363–397.Hodrick, R., & Prescott, E. (1997). Postwar U.S. Business Cycles: An empirical investigation. Journal of Money, Credit, and Banking, 29(1), 1–16.Ilzetzki, E., Mendoza, E. G., & Vegh, G. A. (2013). How big (small) are fiscal multipliers? Journal of Monetary Economics, 60, 239–254.Jeong, M., Kang, J., & Kim, S. (2017). Effects of government consumption shocks in China, Japan, and Korea. China Economic Journal, 10, 194–225.Jia, B., & Kim, H. (2016a). Government spending shocks and private Activity: The role of sentiments. MPRA Working Paper No. 71554.Jia, B., & Kim, H. (2016b). Fiscal expansions and the current account in the US:: The role of sentiments. manuscript.Kim, S. (2015). Country characteristics and the effects of government consumption shocks on the current account and real exchange rate. Journal of International

Economics, 97, 436–447.Kim, C., & Kim, D. (2006). Does Korea have twin deficits? Applied Economics Letters, 13, 675–680.Kim, S., & Roubini, N. (2008). Twin deficit or twin Divergence? Fiscal policy, current account, and the real exchange rate in the U.S. Journal of International Economics,

74, 362–384.Kollmann, R. (1998). U.S. Trade balance Dynamics: The role of fiscal policy and productivity shocks and of financial market linkages. Journal of International Money and

Finance, 17, 637–669.Kollmann, R. (2010). Government purchases and the real exchange rate. Open Economies Review, 21, 49–64.Mankiw, G. (2014). Principles of macroeconomics. South-Western College Publication.Müller, G. J. (2008). Understanding the dynamic effects of government spending on foreign trade. Journal of International Money and Finance, 27, 345–371.Ramey, V. A. (2011). Identifying government spending Shocks: It's all in the timing. Quarterly Journal of Economics, 126(1), 1–50.Ravn, M. O., Schmitt-Groh�e, S., & Uribe, M. (2007). Explaining the effects of government spending shocks on consumption and the real exchange rate. NBER Working Papers

13328. National Bureau of Economic Research.

70