# I. Introduction

The effect of fiscal policy on the stock market has been extensively discussed during the last three decades. Sucharita and Sethi (2011) demonstrated the nexus between fiscal policy and private investment (including the stock market). A looser budget balance (increasing government spending, decreasing tax rate, or both) leads to a higher government borrowing. A large outstanding government debt can reduce private investments (and stock prices) by raising the interest rate and vice versa for a tighter budget balance. Duy-Tung et al. (2018) supported this idea by showing that a fiscal consolidation attempt positively affects stock market performance in emerging Asian economies. In contrast, Qureshi et al. (2019) found that a looser budget balance positively influences the stock market in developed and developing economies. There is no consensus in the literature concerning the effect of fiscal policy on the stock market.

Fiscal news (e.g., budget expansion or budget contraction) could create significant fluctuations in an emerging stock market like Malaysia. The effect of fiscal policy on stock prices is ambiguous (Mumtaz & Theodoridis, 2020) from the perspective of economic theory. Blanchard (1981) pointed out that the impact of fiscal expansions on the stock market returns depends on investors’ expectations of future real interest rates and profits. However, higher (lower) government demand and companies’ sales under budget expansions (contractions) are confirmed. Budget expansions (contractions) are good (bad) news that positively affect stock market returns. According to Bird and Yeung (2012), investors have an asymmetric reaction to good news and bad news. The pessimistic bias of investors leads them to ignore good news in uncertain times. Therefore, the impact of budget expansion (good news) and budget contraction (bad news) on the stock market might have different magnitudes. A large and growing body of literature (Mumtaz & Theodoridis, 2020; Qureshi et al., 2019; Stoian & Iorgulescu, 2020) has investigated the symmetric impact of fiscal policy on financial markets but much less is known about its asymmetric impact on the stock market.

Figure 1.Price-budget correlation (Malaysia, 1997Q1-2018Q4)

This figure plots simple correlation between stock prices (KLCI) and lagged overall budget balance-to-GDP, lagged ascending budget balance-to-GDP, lagged descending budget balance-to-GDP in the left, central and right panel, respectively.

This study argues that the relationship between fiscal news (e.g., budget expansion or budget contraction) and stock market returns are asymmetric. Figure 1 provides a snapshot of the asymmetric reaction of the Kuala Lumpur Composite Index $(KLCI)$ to fiscal policy news. The left panel indicates no clear relationship between $KLCI$ and budget balance. The budget balance is also divided into budget expansion and budget contraction in the central and the right panels. In the budget expansion, a positive relationship exists between the budget balance and $KLCI$ performance, while the budget contraction phase is characterized by a negative relationship between the two variables. Overall, this stylized fact suggests that the impact of fiscal news on the stock market is asymmetric. However, most previous studies only focused on testing the symmetric relationship between fiscal news and stock market returns.

This study provides new insights into how fiscal policy affects stock market performance by decomposing the budget into expansions (good news) and contractions (bad news). Understanding how fiscal news affects the stock market asymmetrically will help fund managers and fiscal authorities to design a proactive investment strategy and appropriate fiscal policy respectively. Notably, fund managers can adjust their portfolios according to different types of fiscal news to maximize returns. On the other hand, stabilizing the stock market is as important as stabilizing the economy; the fiscal authorities are able to design a proper fiscal policy by understanding the various effects of budget expansion and budget contraction on the stock market.

# II. Methodology

This study focused on the impact of asymmetric fiscal policy on the $KLCI.$ Since the $KLCI$ is one of the largest and most active exchanges in Southeast Asia and is fully automated, it is well represented for the emerging stock markets. Furthermore, this study measured the fiscal policy information by the overall budget balance-to-GDP ratio $(BB)$ and the movement of Malaysia’s stock market prices by the logarithm of the $KLCI.$ Following Al-hajj et al. (2018) and Stoian and Iorgulescu (2020), this study used monetary policy which is proxied using the 3-month interest rate on the money market $(IR);$ inflation rate is measured by using the logarithm growth rate of the consumer price index $(CPI);$ the impact of general economic conditions are proxied using the logarithm of the real gross domestic product $(RGDP);$ the external conditions are proxied by the real effective exchange rate $(REER);$ and the influence of input costs are measured by logarithm of oil prices $(OIL).$ In addition, this study used government expenditure as an alternative fiscal variable for robustness test. Table 1 describes the data in detail where all data series are quarterly and cover the period between 1997: Q1 and 2018: Q4.

Table 1.A summary of data
 Variable name Variable description Source Stock market prices (KLCI) The logarithm of KLCI index Bloomberg Terminal Budget balance (BB) The overall budget balance-to-GDP ratio (%). Seasonally adjusted by Census X-13. The central bank of Malaysia Government Expenditure (GE) The total of government operating and development expenditure to GDP ratio. (%) The central bank of Malaysia Interest rate (IR) 3-month money market interest rate (%) The central bank of Malaysia Inflation (CPI) The logarithm of Consumer Price Index International Monetary Fund National production (RGDP) The logarithm if real Gross Domestic Product. Seasonally adjusted by Census X-13. International Monetary Fund Oil price (OIL) The logarithm of West Texas intermediate oil price Federal Reserve Bank St. Loius Exchange rate (REER) The logarithm of real effective exchange rate. International Monetary Fund

This table provides detail information of data used in this study.

The main model of this study is presented below:

where $\propto^{+}$and $\propto^{-}$ represent the long run parameter and $z_{t}$ is the vector regressor which is explained as:

where $z_{t}^{+}$ and $z_{t}^{-}$ represent the positive and negative partial sums which are computed as follows, respectively:

Additionally, the asymmetric error correction model (AECM) takes the following form:

where $\theta^{+} = \frac{\propto^{+}}{\rho y_{t - 1}}$ and $\theta^{-} = \frac{\propto^{-}}{\rho y_{t - 1}}.$

The non-linear autoregressive distributed lag (NARDL) approach includes the following steps: first, this study conducted conventional unit root tests to ascertain that no variable is I (2). The result of the F-test would be spurious in the presence of I (2) variables. Nevertheless, the presence of one or more breaks in the time-series data affects the reliability of conventional unit root tests (Sun et al., 2017). Therefore, this study adopted the break-point unit root test to verify stationarity and determine the structural break’s presence. Second, this study estimated Equation (6) using the Akaike Information Criterion (AIC) with a maximum lag of four to achieve the final specification.

Third, based on the estimated NARDL model, this study performed a test of co-integration among the variables using the bounds testing approach of Pesaran, Shin, and Smith (2001) and Shin and Greenwood-Nimmon (2014). The bound test in the NARDL framework had very similar procedures to the linear Autoregressive Distributed Lag. Case in point, the estimation of Eq. (6) the null hypothesis is $\rho = \theta^{+} = \theta^{-} = 0.$ Finally, in NARDL, the Wald test was employed to find the long-run coefficients by $\theta^{+} = \theta^{-}$ and the short-run coefficients are $\sum_{i = 0}^{p}\pi^{+} = \sum_{i = 0}^{p}\pi^{-}.$

# III. Result

The result of the unit root test[1] proved all variables are I (0) or I (1), thus confirming that none of the series is I (2). Table 2 reports the necessary tests and estimated results. The bound test result proves that F-statistic 3.15 and 3.5 is significant in 5%, suggesting a co-integration between these variables. Furthermore, the estimated model passed the standard diagnostic tests (normality, stability, serial correlation, and heteroscedasticity), indicating that the estimated model was free of traditional regression problems. Contrary to Al-hajj et al. (2018), who focused on long-run results, this study ran non-linear tests, both long run and short run, discovering the following impressive results. First, $OIL$ and $REER$ has an asymmetric impact in both the short run and long run. In the long run, only $RGDP$ and $CPI$ had a nonlinear relationship with $KLCI.$ Meanwhile, fiscal policy has an asymmetric effect on Malaysia’s market return only in the short run, while the impact of fiscal policy on $KLCI$ is symmetric.

Table 2.Bound, asymmetric, and diagnostic tests
 BB (2,1,2,2,0,0,0,0,1,2,2,2,1) GE (1,0,2,2,0,0,2,0,0,0,1,1,2) F-statistic Upper Bound F-statistic Upper Bound Bound test 3.15 3.04 3.5 3.04 Non-linear test F-statistic P-value F-statistic P-value Long run ${BB}^{+} = {BB}^{-}$ 0.6271 0.5378 3.7153** 0.0300 ${IR}^{+} = {IR}^{-}$ 1.0837 0.3452 1.4797 0.2358 ${RGDP}^{+} = {RGDP}^{-}$ 6.6168*** 0.0026 4.7128** 0.0125 ${CPI}^{+} = {CPI}^{-}$ 3.1672** 0.0496 0.7504 4.7650 ${WTI}^{+} = {WTI}^{-}$ 2.5092* 0.0903 2.7680* 0.0707 ${REER}^{+} = {REER}^{-}$ 9.6132*** 0.0003 7.7372*** 0.0010 Short run $\sum_{i = 0}^{p}{BB}^{+} = \sum_{i = 0}^{p}{BB}^{-}$ 9.8090*** 0.0002 - - $\sum_{i = 0}^{p}{IR3}^{+} = \sum_{i = 0}^{p}{IR3}^{-}$ - - - - $\sum_{i = 0}^{p}{RGDP}^{+} = \sum_{i = 0}^{p}{RGDP}^{-}$ - - - - $\sum_{i = 0}^{p}{CPI}^{+} = \sum_{i = 0}^{p}{CPI}^{-}$ - - - - $\sum_{i = 0}^{p}{OIL}^{+} = \sum_{i = 0}^{p}{OIL}^{-}$ 2.0449 0.1388 - - $\sum_{i = 0}^{p}{REER}^{+} = \sum_{i = 0}^{p}{REER}^{-}$ 1.2605 0.2126 1.2583 0.2914 Diagnostic test Jarque-Bera 0.2793 0.8697 0.057 0.9717 LM (1) 0.0223 0.8820 1.7374 0.1925 LM (2) 0.7513 0.4765 1.3498 0.2672 ARCH (1) 0.0001 0.9908 0.6132 0.4359 ARCH (2) 0.1802 0.8354 0.5014 0.6076 CUSUM Stable Stable CUSUMSQ Stable Stable

This table reports bound, asymmetric, and diagnostic test results. Lag length is selected using AIC and is reported in brackets. The model confirmed co-integration if F-statistic > upper bound in Bound test. *, **, *** indicates statistical significance at 10%, 5%, and 1% levels, respectively.

Table 3 shown the estimated result by NARDL. In the long run, only $RGDP$ and $REER$ had a significant impact on $KLCI.$ Moreover, a booming leading to a bull market and recession would create a bear market. Besides, the appreciation of Ringgit Malaysia has worsened the performance of $KLCI$ while the depreciation of local currency increased Malaysia’s stock market return. It could be observed that fiscal policy has no impact on Malaysia’s stock market return. In the short run, fiscal expansion has positively affected $KLCI,$ and fiscal contraction has reduced Malaysia’s market return. This finding supports the asymmetric impact of fiscal policy on $KLCI,$ as presented in Figure 1. Additionally, monetary expansion has harmed $KLCI$ with a one-quarter lag.

Notably, ascending oil prices has no impact on $KLCI,$ while descending oil prices raised Malaysia’s stock market return. The Malaysian government implemented a petrol subsidy to control oil price; therefore, the cost of ascending oil price did not transmit to the stock market while the benefit of descending oil price reflected in stock price hike. In contrast to the long run, the $RGDP$ did not affect $KLCI$ and $REER$ had a reverse impact on Malaysia’s stock market performance. In addition, the robustness test that uses GE as an alternative fiscal variable displays a similar result – budget balance as a fiscal variable.

Table 3.NARDL estimated results
 Variable BB GE Coefficient t-statistic Coefficient t-statistic Long Run ${Fiscal\ variable}^{+}$ -0.0108 -0.9342 0.0022 0.4977 ${Fiscal\ variable}^{-}$ -0.0070 -0.6843 -0.0059 -1.0925 ${IR}^{+}$ 0.0407 0.4709 0.1307 1.6560 ${IR}^{-}$ -0.0407 -0.9884 0.0378 1.0676 ${RGDP}^{+}$ 5.3243* 1.9914 1.2598 0.8059 ${RGDP}^{-}$ -3.3776* -1.8115 -5.8369** -2.2300 ${CPI}^{+}$ 1.7767 0.9726 1.4701 0.9816 ${CPI}^{-}$ 7.7914 1.4644 1.2743 0.4272 ${WTI}^{+}$ -0.3084 -0.5863 -0.6444 -1.5866 ${WTI}^{-}$ -1.0322 -1.4913 -0.6764 -1.4223 ${REER}^{+}$ -13.1374** -2.3907 -9.3904*** -2.6665 ${REER}^{-}$ 5.6655** 2.1351 6.2241*** 2.7031 Short Run Constant 0.0212 1.6375 3.2712*** 8.7493 ${\mathrm{\Delta}KLCI}_{- 1}$ -0.2116** -2.6260 - - ${\mathrm{\Delta}Fiscal\ variable}^{+}$ 0.0065*** 3.3113 - - ${\mathrm{\Delta}Fiscal\ variable}^{-}$ -0.0092*** -4.6991 -0.0001 -0.0476 ${\mathrm{\Delta}Fiscal\ variable}_{- 1}^{-}$ -0.0039** -2.2918 0.0033*** 2.6827 ${\mathrm{\Delta}IR}^{+}$ -0.0196 -1.0472 -0.0085 -0.3804 ${\mathrm{\Delta}IR}_{- 1}^{+}$ -0.0462** -2.5295 -0.0788*** -4.1002 ${\mathrm{\Delta}IR}^{-}$ - - -1.1373** -2.4959 ${\mathrm{\Delta}IR}_{- 1}^{-}$ - - 1.2932** 2.6219 ${\mathrm{\Delta}CPI}^{-}$ -0.8282 -0.7072 - - ${\mathrm{\Delta}WTI}^{+}$ -0.0019 -0.0167 - - ${\mathrm{\Delta}WTI}_{- 1}^{+}$ -0.2014* -1.7917 - -