Mutual fund flows and investors’ expectations in BRICS economies: Implications for international diversification

Economic Systems 43 (2019) 130–150

Contents lists available at ScienceDirect

Economic Systems journal homepage: www.elsevier.com/locate/ecosys

Mutual fund flows and investors’ expectations in BRICS economies: Implications for international diversification

T

Fiza Qureshia, , Habib Hussain Khanb, Ijaz Ur Rehmanc, Abdul Ghafoord, Saba Qureshia ⁎

a

Institute of Business Administration, University of Sindh, Jamshoro, Pakistan Faculty of Management Science, International Islamic University, Islamabad, Pakistan c College of Business Administration, Al Falah University, Dubai, United Arab Emirate d Management Science Department, Bahria University Lahore Campus, Lahore, Pakistan b

ARTICLE INFO

ABSTRACT

Keywords: Mutual fund flows Macroeconomic variables Investors’ expectations BRICS economies

This paper empirically examines the relationship between different classes of mutual funds, measures of investors’ expectations and business cycle movements in the BRICS markets over the 1996Q1-2017Q3 period. Applying the Panel Vector Autoregressive (PVAR) model in a Generalized Method of Moments (GMM) setting, the results suggest a strong causal relationship between mutual fund flows and measures of investors’ future expectations. In particular, fund flows are forward-looking and assist in forecasting real economic conditions. Moreover, investors choose to invest in riskier funds when economic conditions are good, while they prefer safer options in poor economic situations. These findings have important implications for international diversification.

1. Introduction Mutual funds play a significant role for financial markets and hence economic growth. They represent a major portion of investors and influence the economy in three different ways. First, mutual fund investors incorporate macroeconomic information in their portfolios and reallocate their investments to safer investment avenues so as to safeguard themselves from expected (unexpected) losses (Flannery and Protopapadakis, 2002; Kaul and Phillips, 2008). Similarly, investment strategies such as holding more cash in a recession, lowering the portfolio beta and sector rotation (investing more in defensive industries during recession and in cyclical industries in boom periods) entail that mutual funds formulate investment modifications over the business cycle (Kacperczyk et al., 2013). Second, mutual fund flows predict economic conditions that help policymakers forecast macroeconomic conditions (Kaul and Phillips, 2008; Ferson and Kim, 2012; Jank, 2012; Kopsch et al., 2015). Third, mutual fund flows provide capital to the economy, thereby injecting liquidity into the capital market and the real economy while assisting diversification (Halim, 2007). The funds are invested in different securities (such as stocks, bonds, real estate, commodities) in both domestic and international markets and thus influence the overall economy. Finance theory suggests that both economic information and news affect asset prices. Numerous channels highlight the relationship between macroeconomic variables (such as the Gross Domestic Product (GDP), inflation, un-

Corresponding author at: Institute of Business Administration, University of Sindh, Jamshoro. E-mail addresses: [email protected] (F. Qureshi), [email protected] (H.H. Khan), [email protected] (I.U. Rehman), [email protected] (A. Ghafoor), [email protected] (S. Qureshi). ⁎

https://doi.org/10.1016/j.ecosys.2018.09.003 Received 24 August 2017; Received in revised form 29 March 2018; Accepted 25 September 2018 Available online 09 March 2019 0939-3625/ © 2019 Elsevier B.V. All rights reserved.

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

employment and the interest rate) and financial market securities (such as stocks, bonds, money market and other market securities).1 Early studies on fundamental macroeconomic variables and stock markets by Bodie (1976) and Fama (1981) explain the relationship between inflation and stock returns concerning real economic activity. Similarly, Kaul (1987) and Du (2006) consider the relationship between inflation and stock returns given the monetary policy effect (demand and supply shocks). Studies show that the macroeconomic risk is highly related to aggregate investments, employment and output dynamics (Bloom et al., 2007; Bloom, 2009). These potential links suggest that the price of financial securities is associated with changes in macroeconomic variables.2 Hence, the expected performance of market participants in financial markets is likely to be influenced by macroeconomic news. In this vein, Jank (2012) suggests that mutual funds react to new macroeconomic information, which is then reflected in both fund flows and market prices. He argues that the predictor variables can forecast variations in mutual fund flows better than market returns. Therefore, we expect that, as institutional investors, mutual funds are affected due to the funds’ sensitivity to macroeconomic dynamics. Despite the extensive literature on the relationship between financial markets and the macroeconomy, studies on the linkage between financial market investors (e.g. mutual funds) and the macroeconomy are scarce and limited in scope. For instance, limited studies address the questions whether mutual fund flows affect economic conditions and whether fund flows contain information about future economic conditions.3 Some of the existing literature explains that investment by funds is mostly driven by investor sentiment rather than information about real economic fundamentals (Kaul and Phillips, 2008).4 In this vein, Oh and Parwada (2007) argue that it is controversial whether fund flows contain information about real economic activity or not. Several studies have been conducted on the flow-performance relationship of mutual funds at the micro level (Ferson and Schadt, 1996; Ferreira et al., 2012; Kacperczyk et al., 2013). However, there has been limited work on the performance of mutual funds at the macro level and the associated macroeconomic risk (e.g., Jank, 2012; Ferson and Kim, 2012; Bali et al., 2014). Although mutual funds have expanded globally, evidence of their flow-macroeconomic nexus is scarce and geographically limited. Previous studies have mostly been conducted in developed economies in single country settings (Cao et al., 2008; Jank, 2012; Alexakis et al., 2013; Aydogan et al., 2014; Bali et al., 2014). Studies addressing the relationship between mutual funds and the business cycle in emerging markets are nearly non-existent. Our interest is to examine the relationship of equity, bond, balanced and money market mutual funds with the business cycle in BRICS (Brazil, Russia, India, China and South Africa) markets using measures of macroeconomic conditions. The BRICS economies are the chief beneficiaries of international investments and capital flows. International investors pay special attention to the comovement of the BRICS stock markets with international economic factors and global economic financial conditions, given the opportunity of investment and risk diversification (Mensi et al., 2014). Studies suggest that integrated regional markets are more proficient compared to fragmented national markets (Click and Plummer, 2005). While many developed countries face severe adverse economic problems and recessions, emerging economies such as the BRICS countries are less affected by the economic and financial crunches and have maintained vigorous growth (Samargandi and Kutan, 2016b). Moreover, the financial sectors of the BRICS have developed noticeably over the last two decades. The growing international ties and financial trade among the BRICS economies have termed them potential economic superpowers (O’Neill, 2011). Given the progressive role of the BRICS countries as emergent developing economies, examining the role of mutual funds as institutional investors is relevant and interesting. The contributions of this study are manifold. First, it extends the work of Jank (2012) and provides empirical evidence on four different types of mutual funds (i.e., equity, bond, balanced and money market mutual funds) and measures of investors’ future expectations in the BRICS countries. Unlike Jank (2012), who focuses only on equity funds and their association with market returns and macroeconomic variables, this study provides a comparative analysis of four mutual fund classes in relation to investors’ future expectations and business cycle variables. Moreover, the work of Jank (2012) is limited to the US financial market, whereas our study explores the relationship in multiple country settings where we believe the findings can differ from those of developed countries due to the well-developed financial markets, higher access to information, lower participation costs and highly secured regulatory systems of the latter (Ferreira et al., 2012). On the other hand, financial markets in developing economies are characterized by fragile market mechanisms, lesser access to information, higher participation costs, improper regulatory systems and high volatility (Khorana et al., 2005; Halim, 2007). Moreover, several studies5 have indicated that mutual funds are more sophisticated in their investment decisions and tend to be less behaviorally biased in developed countries. These features inspire us to examine the role of mutual funds in developing economies that lack proper information mechanisms and market structures to facilitate investment. 1 See, for example, Fama (1981); Geske and Roll (1983); Chen et al. (1986); Kaul (1987); Barro (1990); Fama (1990); Schwert (1990); Choi et al. (1999); Goetzmann and Massa (1999); Lettau and Ludvigson (2001); Du (2006); Du et al. (2012) and Narayan et al. (2014). 2 Another branch of the literature, popularly known as “literature on return predictability”, documents that “predictive variables are closely linked with economic variables”. These predictive variables forecast changes in the returns of securities in the market (Fama and French, 1989; Campbell and Thompson, 2008). They are, for example: dividend yield (Keim and Stambaugh, 1986; Fama and French, 1989; Hodrick, 1992; Lamont, 1998; Lettau and Ludvigson, 2001, 2005; Westerlund et al., 2015), term spread (Fama and French, 1989; Schwert, 1990) and the Treasury bill rate (Campbell and Viceira, 1996). These are some of the common predictive variables documented in previous studies, which not only capture investor expectations about future returns but also establish the link between market returns and fundamental economic variables (Chen et al., 1986). 3 Ferson and Kim (2012) find that lagged flows have predictive ability for future economic conditions, indicating that fund flows not only follow past market performance but also forecast the future conditions of variables symbolizing economic conditions. 4 According to the information response hypothesis, fund flows are determined by fundamentals, whereas under the price pressure hypothesis fund flows are distinct from fundamentals. For more details, please refer to the studies by Jank (2012) and Ben-Rephael et al. (2011). 5 These include Binswanger (2004), Alexakis et al. (2013); Chatziantoniou et al. (2013) and Thomas et al. (2014).

131

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Second, this study analyses the behavior of risky mutual funds in comparison to less risky ones in different economic setups.6 This is particularly important given the risky and volatile nature of financial markets in developing countries, which entails comprehensive insights into asset allocation and forecasting decisions for portfolio managers, market analysts and policymakers. Third, we extend the Jank (2012) model by including additional variables such as money supply, the budget deficit to GDP ratio, the investment ratio and unemployment to provide further insights into the relationship of funds flows and business cycle variables. Finally, we apply panel data and thus take advantage of the cross-country dimensions of the dataset. Overall, the contribution of this study provides additional evidence from BRICS economies on the relationship between mutual fund flows, investors’ expectations and business cycle variables. The findings suggest that mutual fund flows and measures of investors’ future expectations are significantly interrelated. Furthermore, as risky securities, equity and balanced fund flows are more reactive to information related to macroeconomic variables compared to less risky securities such as bonds and money market fund flows. In addition, investors switch from equity based funds to fixed income type funds in times of economic crises. The rest of the paper is organized as follows. The data and variables are discussed in Section 2. A description of the model and the estimation techniques is provided in Section 3. Section 4 proceeds with the empirical analysis and a discussion of the results. Section 5 concludes the paper. 2. Data and variables Constrained by the availability of high-frequency data for key measures, this paper considers quarterly data for the sample of BRICS (Brazil, Russia, India, China and South Africa) economies over the 1996Q1- 2017Q3 period (the latest data obtainable at the time of extraction). We feel this sample is large enough to capture business cycle behavior over a long horizon. Usually, the effects of changing economic variables or economic policies are perceptible quarterly or annually (Fama, 1981, 1990; Binswanger, 2000).7 It is also pertinent to note that major growth of the mutual fund industry can be observed in the BRICS countries after the Asian financial crisis of 1997 (Klapper et al., 2004). Moreover, Klapper et al. (2004) state that the mutual fund industry came into the limelight globally, particularly in middle-income countries, during the 1990s after the Asian financial crisis. In addition, while the growth of mutual funds has doubled globally during this period, it has increased particularly in emerging economies.8 2.1. Mutual fund flows data As illustrated in Table 1, the total sample consists of 2605 equity mutual funds, 3799 balanced mutual funds, 3522 bond mutual funds, and 1973 money market mutual funds. Table 1 provides a detailed breakdown of the types and total number of mutual funds in each country. We calculate the aggregate fund flows on a quarterly basis. Flows are defined as the net growth in mutual fund assets excluding reinvested dividends (Sirri and Tufano, 1998). Net flows are described as net trading (net buying less net selling), which is a proxy for mutual fund trading behavior in financial markets (Warther, 1995; Ferreira et al., 2012; Thomas et al., 2014). This shows that fund flows represent net trading or net investment by mutual funds in financial markets. According to Ferreira et al. (2012), the flows are defined as the new money growth rate. This is because the net growth in total net assets (TNA) is not dominated due to dividends and capital gains on the assets under management but due to new external money earned through investment (net trading by mutual funds) in the financial markets. Therefore, the trading behavior of mutual funds can only be captured by net fund flows in the financial markets. Previous studies use fund flows as a proxy for the trading activity of mutual funds as institutional investors in financial markets, see, for example, Aydogan et al. (2014); Jank (2012); Thomas et al. (2014); Cao et al. (2008); Sirri and Tufano (1998); Edelen (1999); Edelen and Warner (2001); Ferreira et al. (2012); Ferson and Kim (2012) and Warther (1995). We follow Sirri and Tufano (1998); Ferreira et al. (2012) and Ferson and Kim (2012) in calculating fund flows in Eq. (1).

Flowsi, t = [TNAi, t

TNAi, t

1

(1 + Ri, t )]/ TNAi, t

(1)

1

where TNAi, t is the total net asset in dollars of fund i at the end of month t, and Ri,t Ri, t is fund і’s raw return in dollar value in month t. The quarterly TNA and fund returns data of each fund in each category of funds (equity, bond, balanced and money market funds) have been extracted. Then the flows of each individual fund are calculated through Eq. (1) for each country. Finally, the flows of each period of all individual funds for each country are summed up to obtain the aggregated fund flows of each period (in our case, one 6 Equity mutual funds invest in medium- to long-term equities, which generally tend to be risky investments that provide returns in the form of dividend and capital gains, whereas bond mutual funds invest specifically in fixed income and less risky securities. Moreover, money market funds invest in liquid, short-term, low risk securities. Therefore, we categorize bond mutual funds and money market mutual funds as less risky securities. Balanced funds are investments in a combination of both equity and bond securities. We found that balanced funds follow a moderate investment approach due to the high correlation between equity and balanced fund flows (see Table 2). Therefore, we categorize equity and balanced funds as risky securities. Note that a moderate investment approach entails a higher equity component in the mix of securities by balanced funds/hybrid funds. An opposite investment strategy is a conservative investment approach, which implies a higher fixed-income component in hybrid securities. 7 For example, Fama (1981 and Binswanger (2000) show that stock returns on a monthly basis have lower predictive ability for succeeding real activity growth rates. This is because the effects of a certain production period are extended over a large number of prior periods. Moreover, Fama (1981) suggests that annual data may not be feasible because some of the subsample periods under investigation are short and using annual data may lead to overlapping issues in the observations in the regressions. 8 Please refer to Figs. 1 and 2. Fig. 1 shows the worldwide growth trend of mutual funds from 1996 to 2015. Fig. 2 shows the double digits growth of mutual funds in all major regions of the world in the years 2000 and 2015.

132

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 1 Total number of different types of mutual funds included in the sample. Countries

Equity Funds

Balanced Funds

Bond Funds

Money Market Funds

Brazil Russia India China South Africa Total

1030 192 375 806 202 2605

3023 84 116 378 198 3799

2579 58 335 473 77 3522

349 8 65 1518 33 1973

month). For instance, the flows in the sample of 1030 equity mutual funds in Brazil have been summed up for each period (i.e., from January 2001 to December 2015) to obtain the quarterly aggregated equity fund flows for Brazil. We follow this procedure of aggregating flows as suggested by Ferreira et al. (2012); Ferson and Kim (2012) and Cao et al. (2008). All data are expressed in US dollars. Mutual fund data is winsorized at 1% to counter the problem of outliers (Verardi and Dehon, 2010). 2.2. Measures of investors’ future expectations data We incorporate measures that can capture the investors’ future expectations of returns to test the flow-economy relationship. We therefore consider the dividend yield (DY), the term spread (TS) and the relative Treasury bill ratio (TB) as alternative measures of investors’ future expectations. The dividend yield is computed by the ratio of average annual dividends, and end of quarter market prices (Jank, 2012; Shiller et al., 1984; Campbell and Shiller, 1988; Keim and Stambaugh, 1986; and Fama and French, 1988, 1989, 1992) determine that dividend yields foretell stock returns. Lettau and Ludvigson (2001) find that dividend yields capture longerterm trends in financial markets and provide accurate long-term predictions of booms or collapses. Following Campbell (1991); Hodrick (1992) and Jank (2012), we computed the relative T-bill rate as the 30-day T-bill rate minus its 12-month moving average. The term spread is calculated using the 10-year Treasury bond yield less the 1-year Treasury bill rate following Fama and French (1989). Fama and French (1988, 1990, 1993) note that the term spread captures economic conditions. Campbell (1991); Hodrick (1992); Fama and Schwert (1977); Merton (1973) and Shanken (1990) identify the short-term Treasury bill rate as a predictive variable for a state that tracks variations in investment growth. The first difference in variables is taken into account while identifying the relationship between the business cycle variables and fund flows.9 2.3. Alternative business cycle variables data This study uses alternative business cycle variables data that are highly linked to the financial market variables. Following Thomas et al. (2014), the GDP growth rate and the inflation rate are used. The GDP growth rate is used instead of per capita income growth as it is a better measure of economic stability (Bali et al., 2014; Thomas et al., 2014; Samargandi and Kutan, 2016a). The CPI is used as a proxy for inflation (measure of the price level). Moreover, additional variables such as the money supply growth rate, the budget deficit to GDP ratio, the unemployment rate, the exchange rate and the investment rate are used to provide a deeper understanding of changes in the macroeconomic conditions. Macroeconomic data are extracted from seasonally adjusted data sources (Figs. 1 and 2).10 The study also includes the exchange rate among the explanatory variables as an indicator of a country’s macroeconomic stability. Exchange rate fluctuations (particularly regarding the dollar) have a strong effect on international trade given that the BRICS, as open economies, confront hostile jolts due to international exposure and risk sharing among economies (Haddad et al., 2013; Kopsch et al., 2015). The unemployment rate is also tested as another measure of macroeconomic conditions because unemployment is associated with business cycle fluctuations and stock market variables (Geske and Roll, 1983; Flannery and Protopapadakis, 2002; Bali et al., 2014). Another variable used to measure macroeconomic conditions is the real investment rate. Greater variation in stock prices and returns can be captured by real investment in the economy (Fama, 1981, 1990; Geske and Roll, 1983; Kaul, 1987; Barro, 1990; Galeotti and Schiantarelli, 1994). Likewise, Galeotti and Schiantarelli (1994) find that stock price movements are associated with changes in investment. Finally, monetary and fiscal policy indicators are incorporated to see their impact on fund flows. Monetary policy, government revenue, expenditure and taxes are influential for stock market movements and are closely inter-related (Geske and Roll, 1983; Kaul, 1987; Laopodis, 2009; Chatziantoniou et al., 2013). Following Kaul (1987); Geske and Roll (1983) and Laopodis (2009), we use the narrow money (M1) growth rate as a measure of money supply growth and the government budget deficit to GDP ratio as a proxy for the fiscal policy effect. Table 2 contains a description of the variables, data sources and the descriptive statistics of the variables.

9 Jank (2012) follows the approach of Chen et al. (1986) and uses the first difference of all predictive variables to identify the fund flows’ reaction to news (changes in predictive variables) about the real economy. 10 The macroeconomic data has been taken from the Thomson DataStream and International Monetary Fund (IMF) websites.

133

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Fig. 1. Total number of mutual funds worldwide and worldwide growth in total net asset values (NAVs) of mutual funds (millions of US dollars, year-end). Source: Authors’ calculations based on data collected from the Investment Company Institute (ICI), Mutual funds worldwide market statistics, 2015

Fig. 2. Total number of mutual funds in different regions of the world in 2000 and 2015. Source: Authors’ calculation based on data collected from the Investment Company Institute (ICI), Mutual funds worldwide market statistics, 2015

3. Methodology and estimation techniques 3.1. Panel VAR models Following Abrigo and Love (2016), we take into account a k-variate identical panel VAR of order p with panel-specific fixed effects by the system of linear equations shown below:

Yit = Yit

1

A1 + Yit

2

A2 + ...... + Yit

p+1

Ap

1

+ Yit

p

(2)

Ap + Xit B + µi + eit

where Yit is a (1 x k ) vector of dependent variables, Xit is a (1 x l) vector of independent variables, and µi and eit are (1 x k ) vectors of dependent variable-specific panel fixed effects and idiosyncratic errors, respectively. The (k x k ) matrices A1 , A2 , ... , Ap 1 , Ap and the (l x k ) matrix B are coefficients to be estimated. 3.2. Panel VAR model by GMM estimation Holtz-Eakin et al. (1988) find that the equation-by-equation GMM estimation produces reliable estimates of panel VAR and may kp + l instruments is given by the row vector Zit , where provide efficient estimates. Suppose the common set of L kp+ l L Xit Zit , and the equations are indexed by a number in superscript. Consider the following transformed panel VAR model based on Eq. (4) but represented in a more compact form:

Yit* = Y¯it*A + eit* Yit* = [Yit1 *....... + Yitk *] Yit* = [Yit* 1.....Yit* p. + Xit*] eit* = [eit1*....... + eitk *] A = [A1 .....Ap B]

(3)

where the asterisk denotes some transformation of the original variable. If we denote the original variable as mit then the first difference transformation implies that m *it = mit mit 1 mit* = mit mit 1, while for the forward orthogonal deviation

mit ,

m *it = (mit

m¯it ) Tit /(Tit + 1) mit* = (mit

m ¯ it )

T it (T it + 1)

, where Tit T it is the number of available future observations for panel i at

time t , and m ¯ it is its average. Suppose we stack observations over panels and then over time. The GMM estimator is given by 134

135

Thomson DataStream IMF Thomson DataStream

The first difference of the difference between the 10-year and 1-year maturity treasury rates at the end of each quarter The first difference of the 3-month T-bill rate minus its 12- month moving average The first difference of ratio of average annual dividends and end of quarter prices

Notes: The table offers an explanation of the types of variables used in the study. It provides a detailed discussion of their definitions, sources and statistics.

IMF IMF IMF IMF IMF Thomson DataStream Thomson DataStream

Bloomberg

Sources

Growth rate of real gross domestic product Rate of inflation based on the consumer price index Percentage change in the exchange rate USD/local currency Unemployment rate defined as the number of unemployed as percentage of the labor force Growth rate of money supply (M1) Budget deficit as percentage of GDP Growth rate of net investment, i.e. gross investment minus depreciation

TNAi, t 1 (1 + Ri, t )]/TNAi, t 1], Percent change in flows calculated with the formula [Flowsi, t = [TNAi,t where TNAi,t is the total net asset of fund i at the end of quarter t, and Ri, t is fund і’s raw return in quarter t

Net fund flows

Business cycle variables GDP growth ( GDP) Inflation rate ( Inf) Exchange rate ( Ex) Unemployment rate ( UE) Money supply growth ( MS) Deficit to GDP ratio ( DG) Investment rate ( Inv) Predictable variables Term spread ( TS) Treasury bill ratio ( TB) Dividend yield ( DY)

Definition

Variables

Table 2 Variable description, sources and statistics.

.649 −.719 .901

.0473 .0507 .021 .113 .0378 −.129 .0318

.0448

Mean

0.89435 0.31596 0.93787

.04398 .042304 .03929 .15719 .040248 .29711 .0918225

.094214

Std. Dev.

F. Qureshi, et al.

Economic Systems 43 (2019) 130–150

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

A = (Y¯ * Z W Z Y¯ *)

1

(4)

(Y¯ * Z W Z Y¯ *)

where W is a (L x L) weighting matrix assumed to be non-singular, symmetric and positive semi-definite. Assuming that E [Z e] = 0 ˆ W may be selected to maximize efficiency and rank E [Y¯ * Z ] = kp + l , the GMM estimator is consistent. The weighting matrix W (Hansen, 1982).11 Since we expect the presence of the endogeneity problem among fund flows and business cycle variables as reported by earlier studies,12 we estimate the panel VAR model. The panel vector autoregressive model (PVAR) is applied to assess the interaction between endogenous variables and permits an unobserved heterogeneity (Love and Zicchino, 2006). Earlier studies13 also estimate the VAR model. However, they apply VAR in a time series setting. This study uses a reduced-form panel VAR in a generalized method of moments (GMM)14 environment following Love and Zicchino (2006).15, 16 Before the VAR analysis, it is important to check the stationarity, model selection and the optimal lag order selection of variables. To test the stationarity of the variables, we perform both the Fisher type augmented Dickey Fuller unit root test and the Philips Perron unit root test (with and without drift). The result suggests that all variables are stationary at level.17 3.2.1. Flows-measures of investors’ future expectations model To examine the mutual fund flows and measures of investors’ future expectations, we estimate the panel VAR model. n

Flowsi, t =

1

+

n 1 Flowsi, t 1+

i=1 n

Zi, t =

2

+

2 Zi, t 1

+

1i, t

(5)

i=1 n

3 Flowsi, t 1+ i=1

4 Zi, t 1

+

2i, t

(6)

i=1

where Flowsi,t Flowsi, t stands for net fund flows or net sales and Z i,t Zi, t stands for the vector of alternative measures of investors’ future expectations (dividend yield, Treasury bill rate, term spread). i,t 1i, t stands for the error term. The expected relationship between dividend yield and fund flows is negative, since a higher dividend yield ratio suggests lower market returns and vice versa. Kaul and Phillips (2008) note that a high term spread indicates bad economic conditions and vice versa. Thus, the expected result is based on the criteria that if term spread is expected to be high, more outflows of mutual funds are expected in the market and vice versa. Kaul and Phillips (2008) and Jank (2012) report that high T-bill values indicate good economic conditions and vice versa. This implies the expected positive relationship between the relative Treasury bill rate and flows into mutual funds. 3.2.2. Robustness check: flows-business cycle model We also conducted a robustness check by empirically examining the model of the flows-business cycle variable relationship. We estimate the following panel VAR equation: n

Flowsi, t =

1

+

n 1 Flowsi, t 1+

i=1

n

Xi, t =

2

+

+

1i, t

(7)

n 3 Flowsi, t 1+

i=1

2 Xi, t 1 i=1

4 Xi, t 1

+

2i, t

(8)

i=1

where X refers to the vector of business cycle variables (GDP, Inf, UE, Ex, Inv, MS and DG).18 This model tests the flow-business cycle relationship to ascertain concurrent relationships among the variables. The expected outcome is a positive relationship between the fund flows and the macroeconomic variables because fund flows increase with good economic news and vice versa. 11

Roodman (2006) provides a detailed explanation of GMM estimation using Stata in a dynamic panel setting. See the studies by Bali et al. (2014) and Kopsch et al. (2015). 13 These include Edwards and Zhang (1998), Ben-Rephael et al. (2011); Jank (2012) and Kopsch et al. (2015). 14 Wooldridge (2001) and Assenmacher and Gerlach (2008) state that GMM is feasible for estimating interesting extensions of the basic unobserved effects model, for example models where unobserved heterogeneity interacts with observed covariates. 15 To avoid the problem of the mean-differencing procedure in eliminating fixed effects, the Helmert procedure transformation is used to estimate coefficients by GMM. For a detailed discussion, refer to Arellano and Bover (1995); Love and Zicchino (2006) and Assenmacher and Gerlach (2008). 16 We use the PVAR command with fod, instlags, gmmstyle and porder options in STATA-14. The fod option uses the Helmert transformation to remove panel-specific fixed effects, instlags specifies the lag orders of variables to be used as instruments; gmmstyle uses “GMM-style” instruments, and the porder option specifies the Cholesky ordering of endogenous variables. For robustness purposes, we also use an alternative ordering of endogenous variables to generate PVAR estimates. However, the main findings are qualitatively similar. The results are not reported for brevity purposes. 17 The results are not reported for brevity purposes but can be provided on request. 18 Refer to Section 2.2 for details of the macroeconomic variables. 12

136

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 3 Correlation matrix.

(1) Equity flows (2) Bond flows (3) Balanced flows (4) Money market flows (5) GDP (6) Inf (7) MS (8) UE (9) EX (10) DG (11) Inv (12) TBΔTB (13) TS (14) DY

1

2

3

4

5

6

7

8

9

10

11

12

13

14

1 −0.43* 0.79** −0.252 0.256 0.201 0.520* 0.212** 0.051 0.361* 0.22 0.40* −0.310 −0.410

1 −0.291 0.352 0.355 0.302 0.314 0.163** 0.041 0.321* 0.221 0.102 −0.005 −0.101

1 0.39 0.38 0.252 0.139 −0.071 0.486** 0.081 0.061 0.252 0.171 −0.43

1 0.34 0.31 0.34 0.12 0.03 0.35* 0.20 0.21 −0.001 −0.112

1 0.67* 0.778 −0.088 −0.105 −0.045 0.382** 0.040 −0.032 −0.044

1 0.630* −0.119 −0.097 0.11* 0.376** −0.052 −0.041 −0.050

1 0.014 −0.050 −0.027 −0.052 0.037 −0.047 −0.048

1 −0.024 0.119* 0.011 −0.003 −0.013 −0.081

1 −0.040 −0.152** 0.036 −0.009 −0.048

1 0.108* 0.041 0.018 0.059

1 −0.007 −0.015 −0.052

1 .003 .019

1 −.010

1

Notes: The table reports the correlation among variables, where ‘flows’ stands for net flows or net sales (in percent) and Δ represents the first difference of variables. ΔGDP is the gross domestic product growth rate, ΔInf is the inflation rate, ΔMS is the money supply growth rate, ΔUE is the unemployment rate, EX is the exchange rate, ΔDG is the deficit to GDP ratio, Inv is the investment growth rate, ΔTB is the relative Treasury bill rate, ΔTS is the Treasury spread, and ΔDV is dividend yield. ** and * indicate the statistical significance of correlations at 1% and 5% levels, respectively.

4. Empirical results 4.1. The correlation matrix Table 3 presents the correlation matrix for the preliminary analysis of all variables. We note that the correlations among the variables are not too high to create problems of multicollinearity. The only noticeable correlation is between equity flows and balanced flows with (0.85).19 Other variables, such as money supply, dividend yield and the deficit to GDP ratio, also exhibit a moderate correlation with fund flows. 4.2. Selection order criteria The selection of the model is supported by the maximum likelihood-based Bayesian Information Criteria (MBIC), the maximum likelihood-based Akaike Information Criteria (MAIC) and the maximum likelihood-based Hannan-Quinn Information Criteria (MQIC) by Andrews and Lu (2001) reported in Table 4. Based on the results and the overall coefficient of determination, second-order panel VAR is the preferred model since this has the smallest MBIC, MAIC and MQIC. Hansen’s J-statistic, which shows that validity of the instruments, is also reported.20 4.3. Results and discussion 4.3.1. Flows-measures of investors’ future expectations model Tables 5 and 6 report the VAR model and the Granger causality Wald test respectively for fund flows and measures of investors’ future expectations. In order to test all variables in one model, we apply the PVAR model to determine the relationship among fund flows and investors’ future expectations using two lags. In Table 5, we find that changes in the investors’ expectations variables at the lagged period explain mutual fund flows in the current period, whereas in Table 6 we find that changes in mutual funds at the lagged period affect changes in the measures of investors’ expectations. For almost all variables of investors’ future expectations, the study finds consistent patterns; all fund flows are significantly associated with variables. The results suggest that the measures of investors’ future expectation and fund flows are significantly associated. The Treasury bill ratio (TB), term spread (TS) dividend yield (DY) are significantly associated with fund flows. Both term spread and dividend yield are inversely related to equity flows and balanced flows but positively related to bond and money market flows. Moreover, we find that the Treasury bill ratio is positively associated with equity and balanced flows and negatively associated with bond and money market flows. We find that an increase in the TB ratio indicates better expected economic conditions and thus has a positive effect on equity and balanced flows and a negative effect on bond and money market flows, whereas an increase in dividend yield and term spread signals a poor state of the economy, thus reducing equity and balanced flows, while increasing bond and money market flows. These findings are in line with Kaul (1987); Barro (1990) and Laopodis (2009). 19

A correlation between equity and balanced flows is not problematic as the analysis of each fund flow is carried out separately. See Abrigo and Love (2016) for lag selection criteria under the Panel VAR methodology. The lag selection table for the flow-economy model is not reported for brevity purposes. 20

137

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 4 Selection order criteria for flow-measures of investors’ future expectations model. lag

CD

J

J p-value

MBIC

MAIC

MQIC

1 2 3 4 5 6 7 8 9 10

0.091785 0.070974 0.131451 0.201299 0.192067 0.193847 0.186351 0.211824 0.221724 0.284535

59.85378 25.32114 41.56261 23.42153 22.13772 15.40425 9.600256 5.40267 3.103567 1.29E-30

0.10478 0.5828 0.047648 0.495042 0.333085 0.495253 0.650984 0.713798 0.547229 .

−149.262 −175.125 −117.194 −112.656 −91.2599 −75.3138 −58.4383 −39.9564 −19.476 1.29E-30

−15.2462 −28.2886 −14.4374 −24.5785 −17.8623 −16.5958 −14.3997 −10.5973 −4.98643 1.29E-30

−66.0783 −74.1396 −55.6068 −59.8665 −47.269 −40.1211 −32.0438 −22.36 −11.9778 1.29E-30

Notes: The table reports the lag selection criteria of the flow-return PVAR model. CD stands for the coefficient of determination; J stands for Hansen’s J statistic and Hansen’s J p-value determines the validity of the instruments. MBIC stands for maximum likelihood-based Bayesian Information Criteria (MBIC), MAIC for maximum likelihood-based Akaike Information Criteria (MAIC) and MQIC for maximum likelihood-based Hannan-Quinn Information Criteria. Table 5 Panel VAR model of total fund flows and measures of investors’ future expectations.

Flows t-1 Flows t-2 Wald test p-value TB t-1 TB t-2 Wald test p-value TS t-1 TS t-2 Wald test p-value DY t-1 DY t-2 Wald test p-value

Equity Flows

Bond Flows

Balanced Flows

Money Market Flows

0.710 (3.82)** 0.450 (2.92)** 0.00 0.54 (3.94)** 0.34 (2.74)** 0.00 −0.32 (5.18)** −0.37 (3.18)** 0.00 −0.10 (2.81)** −0.20 (2.99)** 0.00

0.012 (0.22) 0.034 (0.32) 0.13 −0.15 (2.89)** −0.25 (2.09)* 0.02 0.91 (2.96)** 0.23 (2.06)* 0.00 0.18 (0.78) 0.17 (4.48)** 0.02

0.156 (3.96)** 0.460 (2.96)** 0.00 0.13 (2.14)* 0.24 (2.84)** 0.00 −0.51 (2.03)* −0.14 (2.33)* 0.04 −0.17 (2.67)** −0.119 (3.67)** 0.00

−0.098 (3.07)** −0.100 (2.17)* 0.00 −0.12 (2.05)* −0.32 (2.50)* 0.04 0.10 (2.82)** 0.72 (4.82)** 0.00 0.12 (1.02) 0.220 (2.92)** 0.05

Notes: The table reports the result of the panel VAR model estimated by GMM of net fund flows, and measures of investors’ future expectations. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

The lags of all fund flows are associated with measures of investors’ future expectations, suggesting that mutual funds predict future economic conditions. This is consistent with the findings of Jank (2012). who shows that mutual funds are forward-looking and predict expected real economic activity. Overall, fund flows are highly correlated with the variables, suggesting that fund flows convey information about future macroeconomic variables as they respond to changes in the state variables that are usually related to a future stage of the cycle. The lags of all measures of investors’ future expectations are related to fund flows, indicating that mutual funds react to new macroeconomic information and incorporate economic information when making investment and asset allocation decisions. Moreover, poor economic conditions imply an increase in investments in bond funds and money market funds, thus providing support for the notion that investors switch to safer investment avenues (like bonds) in times of poor economic conditions. Our findings are consistent with Ferson and Kim (2012), who find that investors reduce equity based investments and increase fixed based investments in times of a poor economic state. For further analysis of this relationship, we split the flows into expected and unexpected components following Warther (1995) and Jank (2012). The unexpected flows are obtained by calculating the residuals from a panel fixed effect model when flows are regressed on other explanatory variables.21 The predicted values of flows from this model represent the expected flows. The estimation results for the expected and unexpected flows are reported in Tables 7 and 8. The coefficients of unexpected fund flows and

21

Jank (2012) states that changes in the economic variables (unexpected) determine unexpected changes in mutual fund flows. 138

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 6 Panel VAR model of total fund flows and measures of investors’ future expectations.

Equity flows(t-1) Equity flows(t-2) Wald test p-value Bond flow(t-1) Bond flow(t-2) Wald test p-value Balanced flows(t-1) Balanced flows(t-2) Wald test p-value Money market flow(t-1) Money market flow(t-2) Wald test p-value

Flows

TB

TS

DY

1.129 (2.78)* 0.211 (3.59)** 0.00 1.129 (3.78)** 0.105 (2.20)* 0.00 −0.021 (2.02)* −0.012 (3.14)** 0.00 0.041 (1.66) 0.058 (4.16)** 0.00

0.475 (2.56)* 0.867 (4.91)** 0.00 −0.275 (2.31)* −0.207 (2.30)* 0.05 0.314 (2.13)* 0.216 (2.80)** 0.00 −0.028 (2.39)* −0.086 (2.99)** 0.00

−0.938 (2.06)* −3.193 (4.35)** 0.00 0.938 (2.06)* 0.706 (2.49)* 0.03 −0.015 (2.28)* −0.071 (2.89)** 0.00 −0.051 (0.68) 0.031 (2.91)** 0.05

−0.412 (1.73) −1.791 (3.51)** 0.00 0.412 (2.83)** 0.301 (2.54)* 0.00 −0.010 (3.29)** −0.013 (0.11) 0.04 0.082 (1.72) 0.150 (3.54)** 0.02

Notes: The table displays the results of the panel VAR model estimated by GMM of the net fund flows and measures of investors’ future expectations. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are in parentheses. ** and * indicate significance at the 1% and 5% level, respectively. Table 7 Panel VAR model of total expected fund flows and measures of investors’ future expectations. Response to

Flows t-1 Flows t-2 Wald test p-value TB t-1 TB t-2 Wald test p-value TS t-1 TS t-2 Wald test p-value DY t-1 DY t-2 Wald test p-value

Response of Equity Flows

Bond Flows

Balanced Flows

Money Market Flows

0.008 (0.16) 0.123 (2.12)* 0.05 1.554 (2.07)* 2.066 (2.18)* 0.00 −2.713 (2.58)* −2.764 (2.20)* 0.00 −1.192 (2.32)* −0.390 (0.85) 0.04

−0.143 (2.13)* 0.011 (0.21) 0.02 −1.926 (2.48)* −3.525 (2.53)* 0.04 2.98 (2.05)* 1.006 (2.02)* 0.04 0.856 (1.68) 0.703 (1.71) 0.12

−0.020 (0.47) −0.061 (1.34) 0.15 1.370 (0.33) 2.121 (1.81) 0.15 −2.207 (2.90)** −2.077 (2.17)* 0.00 −0.527 (1.26) −0.220 (0.62) 0.14

−0.176 (2.37)* 0.014 (0.26) 0.03 −1.932 (1.34) −0.594 (0.43) 0.20 3.954 (2.46)* 2.140 (2.14)* 0.05 0.737 (1.29) 0.930 (0.16) 0.11

Notes: The table reports the results of the panel VAR model estimated by GMM of expected fund flows and measures of investors’ future expectations. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

measures of investors’ expectations are greater than the expected fund flows and measures of investors’ expectations. This indicates that unexpected fund flows are more related to all measures of investors’ future expectations. Table 8 shows that all fund flows are significantly associated with almost all lagged measures of investors’ expectations. This implies that unexpected fund flows contain information on the real economy and assist in predicting economic conditions. We perform the Granger causality Wald test and report the results to validate the PVAR estimates. The Wald p-value confirms that fund flows Granger-cause all measures. Our findings are different from Jank (2012), who concludes that innovations in equity mutual fund flows are related to the measures of investors’ future expectations except term spread, whereas our findings suggest that all four types of mutual funds are related to all measures of investors’ future expectations.

139

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 8 Panel VAR model of total unexpected net flows and measures of investors’ future expectations. Response to

Flows t-1 Flows t-2 Wald test p-value TB t-1 TB t-2 Wald test p-value TS t-1 TS t-2 Wald test p-value DY t-1 DY t-2 Wald test p-value

Response of Equity Flows

Bond Flows

Balanced Flows

Money Market Flows

0.105 (1.71) −0.061 (0.91) 0.10 1.754 (2.17)* 2.96 (2.88)* 0.00 −2.90 (2.98)** −2.79 (2.90)** 0.00 −1.92 (2.32)* −0.401 (2.85)* 0.00

0.331 (6.10)** 0.208 (3.18)** 0.00 −1.988 (2.68)* −3.599 (2.98)** 0.00 3.08 (2.95)** 1.976 (2.92)** 0.00 1.56 (2.68)* 0.803 (2.71)* 0.01

−0.427 (4.95)** −0.167 (2.02)* 0.02 1.270 (2.33)* 2.16 (2.71)* 0.01 −2.271 (2.99)** −2.79 (2.23)* 0.00 −0.770 (2.96)** −0.223 (0.62) 0.04

−0.059 (2.83)* 0.106 (1.49) 0.04 −2.98 (2.64)* −0.99 (2.49)* 0.02 3.99 (3.96)** 2.91 (2.99)** 0.00 0.937 (2.29)* 0.099 (2.16)* 0.03

Notes: The table reports the results of the panel VAR model estimated by GMM of unexpected fund flows and measures of investors’ future expectations. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

4.3.2. Impulse response function (IRF) The stability condition of the estimated PVAR is checked before the estimation of the IRF and FEVD of the model. Fig. A1 in the Appendix A shows that all the eigenvalues lie inside the unit circle. PVAR satisfies the stability condition.22 A graph of OIRFs is presented in Fig. 3. The 5% error bands are estimated using the Gaussian approximation generated by the Monte Carlo simulation with 1000 reps.23 Fig. 3 depicts graphs of impulses and responses of MF flows and measures of investors’ future expectations. The response of equity flows to TB ratio shocks shows a slightly increasing and decreasing trend in the first five periods and an increase to 0.01 standard deviation to the 10th period. This is expected, as the equity flows and the TB ratio are positively correlated. Similar patterns can be observed in response to balanced flows to TB ratio shocks. Balanced flows have a significant pronounced increasing response of about 0.03 standard deviation, since the TB ratio signals good economic conditions. However, the response of bond flows and money market flows is negative to TB ratio shocks. The response of equity flows to the TS ratio and the DY ratio is significantly negative in the subsequent periods, ahead of about 0.03 and 0.02 standard deviation. The response of balanced flows is similar. However, bond flows and money market flows are positively reactive to shocks of the TS ratio and the DY ratio. This implies that an increase in TS and DY signal an expected negative economic state and may thus reduce equity and balanced flows and increase bond and money market flows. The response of the TB ratio to shocks in equity flows increases to 0.03 standard deviation to the 10th period. This entails that shocks of equity flows influence the TB ratio. Moreover, the response of the TS ratio and DY to shocks in equity flows is negative. This is observed in the PVAR results reported in Section 4.3.1. The response of the TB ratio to shocks in bond flows is significantly negative. A similar case is observed with money market flow shocks. The responses of all measures of investors’ future expectations to MF flow shocks are significant. The results substantiate with the PVAR results that the Treasury bill ratio is causally associated with fund flows in such a way that increases in the TB ratio leads to increases in equity and balanced flows and decreases in bond and money market flows. On the contrary, both term spread and dividend yield are inversely related to equity and balanced flows but positively related to bond and money market flows. 4.3.3. Factor error variance decomposition (FEVD) The FEVD for the PVAR model is presented in Table 9. The predictive variables explain more of the fund flows variation ten periods ahead. The proportion of the forecast error in fund flows is largely explained by its own shock in the 10th quarter with equity 55%, bond 65%, balanced 60% and money market 50%, respectively. It is observed that equity flows explain 55% of total variations in flows themselves, bond flows 65%, balanced 60% and money market 50%. The TB ratio explains 25% of money market flows followed by balanced, equity and bond flows with 22%, 20% and 15%, respectively. The TB ratio explains 25% of money market 22 The result of stability is the same for all PVAR models reported in Tables 5–8. All PVAR models are stable. The figures are not shown for brevity purposes. 23 The study follows the procedure of generating impulse response functions by Love and Zicchino (2006) and Abrigo and Love (2016).

140

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Fig. 3. Mutual fund flows and measures of investors’ expectations. Notes: The X-axis shows the number of time periods and the Y-axis shows the unit shock standard deviations. The blue line shows the orthogonalized shock and response of one variable on another variable. The red lines represent the plus and minus two standard deviation bands.

Table 9 FEVD of fund flows and measures of investors’ expectations. Flows

Flows

TB

TS

DY

Equity Bond Balanced Money Market

0.55 0.65 0.60 0.50

0.20 0.15 0.22 0.25

0.10 0.12 0.10 0.15

0.15 0.08 0.18 0.10

Notes: The percent of variation in the row variable (10 periods ahead) is explained by the column variable. FEVD standard errors and confidence intervals are based on Monte Carlo simulations.

flows followed by balanced, equity and bond flows with 22%, 20% and 15%, respectively. TS explains 15% and 12% variations in money market and bond flows, respectively. DY explains 18% and 15% variation in balanced and bond flows, respectively. The proportion of forecast error in the predictive variables shows that they display about 40%–50% variation in fund flows and the remaining variations are explained by the flows’ own shocks. The results from the variance decomposition imply that the predictive variables are important in determining the relationship with fund flows. The association of fund flows with predictive variables entails that fund flows can predict the macroeconomic conditions. Overall, the findings corroborate the PVAR results reported in Section 4.3.1. 4.3.4. Flows-market returns-investors’ expectations model So far, we have followed Jank (2012) in analyzing the relationship among fund flows, market returns and investors’ expectations. However, it is possible that the relationship between fund flows and investors’ expectations is driven by market returns such that the

141

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 10 Panel VAR model of total fund flows, market returns and measures of investors’ future expectations. Equity Flows

Flows t-1 Flows t-2 Wald test p-value MR t-1 MR t-2 Wald test p-value TB t-1 TB t-2 Wald test p-value TS t-1 TS t-2 Wald test p-value DY t-1 DY t-2 Wald test p-value

Bond Flows

Balanced Flows

Money Market Flows

Model (a)

Model (b)

Model (a)

Model (b)

Model (a)

Model (b)

Model (a)

Model (b)

0.27 (2.92)** 0.480 (2.91)** 0.00 0.21 (0.14) 0.09 (2.90)** 0.00

0.21 (2.42)* 0.420 (2.01)** 0.04 0.31 (0.14) 0.06 (2.29)* 0.05 0.006 (3.92)** 0.032 (2.93)** 0.00 −0.08 (3.00)** −0.032 (3.07)** 0.00 −0.011 (2.91)** −0.021 (2.98)** 0.00

0.006 (0.29) 0.030 (0.24) 0.11 −0.358 (2.92)** −0.398 (2.89)** 0.00

0.002 (0.21) 0.033 (0.30) 0.13 −0.253 (2.42)* −0.377 (2.12)* 0.05 −0.007 (2.81)** −0.065 (2.08)* 0.02 0.09 (2.96)** 0.023 (2.06)* 0.00 0.008 (0.78) 0.070 (4.48)** 0.02

0.106 (2.91)** 0.461 (2.97)** 0.00 0.253 (3.73)** 0.049 (0.75) 0.00

0.101 (2.11)* 0.432 (2.89)** 0.00 0.233 (2.93)** 0.038 (0.71) 0.00 0.003 (2.14)* 0.054 (2.84)** 0.00 −0.051 (2.03)* −0.001 (2.33)* 0.04 −0.017 (2.67)* −0.119 (3.67)** 0.00

−0.097 (3.05)** −0.123 (2.69)* 0.00 −0.266 (5.09)** −0.103 (2.12)* 0.00

−0.082 (2.35)* −0.121 (2.19)* 0.00 −0.243 (3.01)** −0.007 (0.12) 0.03 −0.02 (2.05)* −0.03 (2.50)* 0.04 0.010 (2.82)** 0.072 (4.82)** 0.00 0.002 (1.02) 0.220 (2.92)** 0.05

Notes: The table reports the results of the panel VAR model estimated by GMM. Model (a) presents the estimations of net fund flows and market returns. Model (b) presents the estimations of net fund flows, market returns and measures of investors’ future expectations. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

movements in fund flows and investors’ expectations are both associated with market returns.24 In this section, we further analyze this relationship to see if the presence of market returns affects the explanatory power of fund flows with investors’ expectations. We use the popular stock market index for each country as a proxy for market returns.25 The estimation results of total fund flows, market returns and investors’ expectations are reported in Table 10. Model (a) presents the estimations of net fund flows and market returns and model (b) presents the estimations of net fund flows, market returns and measures of investors’ future expectations. A visible observation is that measures of investors’ expectations in the model reduce the explanatory power of market returns. The coefficient of market returns is lower in magnitude in model (b) than in model (a). This indicates that mutual funds are more strongly associated with investors’ expectations than market returns. Thus, we infer that fund flows convey information about the future macroeconomic variables as they respond more to changes in the state variables that are usually related to a future stage of the cycle. 4.3.5. Robustness check: flow-business cycle relationship As a robustness check, we use alternative measures of macroeconomic variables. Tables 11 and 12 depict the results of four mutual fund flows and macroeconomic variables. The findings are similar to what we noted in the flow-measures of the investors’ future expectations model. Lags of most of the business cycle variables significantly affect current fund flows. GDP, exchange rate and real investment are positively related to all fund flows. However, inflation has a negative but insignificant relationship with fund flows. One possible explanation is that GDP and money supply are correlated to inflation and pick up some of its explanatory power. It is interesting to note that money supply has a positive relationship with equity and balanced flows, and a negative relationship with bond and money market flows. This shows that money supply, which is a proxy for monetary policy, has a different impact on risky and less risky mutual funds. This indicates that a lower money supply curtails the purchasing power of investors and switches investors’ preferences to invest in secure avenues like fixed income securities. A similar effect is also noted in the case of money market flows. We find a negative relationship between money supply and money market flows. On the contrary, the budget deficit ratio, which is a proxy for fiscal policy, has a negative relationship with equity flows and balanced flows. A similar relationship is noted with the unemployment rate, which also has a negative relationship with equity and balanced flows. Further, both budget 24 The relationship between fund flows and market return is well documented in Warther (1995) and Edelen and Warner (2001), while the relationship of fund flows and predicted variables is documented in Jank (2012). 25 Stock market index data includes the Shanghai composite index, the Indian SENSEX Index, the Ibovespa Brazil Sao Paulo Stock Exchange Index, the Russia MICEX Stock Market Index and the South Africa FTSE/JSE Index. The data have been taken from Thomson Reuters DataStream.

142

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 11 Panel VAR model of total fund flows and business cycle variables. Response to

Flows t-1 Flows t-2 Wald test p-value GDP t-1 GDP t-2 Wald test p-value Inf t-1 Inf t-2 Wald test p-value Ex t-1 Ex t-2 Wald test p-value UE t-1 UE t-2 Wald test p-value MS t-1 MS t-2 Wald test p-value DG t-1 DG t-2 Wald test p-value Inv t-1 Inv t-2 Wald test p-value

Response of Equity Flows

Bond Flows

Balanced Flows

Money Market Flows

0.710 (8.82)** 0.062 (1.03) 0.00 4.231 (2.92)** 2.291 (2.05)* 0.01 −1.146 (0.52) −6.803 (1.34) 0.13 1.748 (3.86)** −3.723 (1.87) 0.00 −1.530 (0.50) −1.444 (4.94)** 0.04 1.483 (3.48)** 7.888 (4.32)** 0.00 −0.965 (1.45) −3.251 (3.66)** 0.00 2.118 (3.73)** 0.781 (0.64) 0.01

0.001 (0.22) −0.015 (0.25) 0.15 3.562 (3.77)** 2.010 (2.36)* 0.00 −1.124 (1.31) −0.467 (1.47) 0.11 1.665 (4.00)** 3.319 (2.97)** 0.00 1.024 (4.52)** 1.440 (4.25)** 0.00 −2.382 (3.07)** −0.749 (0.53) 0.04 1.071 (2.95)** 0.015 (2.24)* 0.00 1.529 (3.54)** 0.197 (0.35) 0.02

0.156 (3.96)** −0.098 (1.83) 0.02 3.393 (3.72)** 1.141 (0.10) 0.03 −2.469 (1.26) −1.636 (0.27) 0.23 3.354 (4.85)** 3.471 (2.56)* 0.00 −1.436 (4.17)** −2.571 (1.59) 0.03 1.752 (2.56)* 3.692 (2.34)* 0.01 −2.290 (3.51)** −1.023 (3.76)** 0.00 1.460 (5.28)** 1.736 (1.60) 0.00

−0.098 (8.07)** −0.024 (0.44) 0.00 0.725 (2.83)** 1.597 (2.12)* 0.00 −0.240 (0.02) −3.604 (1.05) 0.10 2.099 (3.18)** 2.828 (2.12)* 0.00 1.395 (5.30)** 2.282 (1.65) 0.00 −4.066 (4.90)** −2.386 (1.52) 0.00 1.254 (2.44)* 0.572 (6.79)** 0.00 2.151 (3.04)** 0.552 (0.53) 0.00

Notes: The table reports the results of the panel VAR model estimated by GMM of net fund flows and business cycle variables. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

deficit and unemployment rate have a positive relationship with bond and balanced flows. This implies that equity and balanced flows (bond and money market flows) decrease (increase) their trading activities in times of poor macroeconomic conditions. This is expected, as a higher budget deficit ratio and an increased unemployment rate send negative vibes to the economy. Therefore, we find that investment by equity and balanced funds decreases in times of high economic and financial crises. However, we find a positive relationship between bond and money market flows with budget deficit and unemployment rate. We observe that fixed income funds increase their trading activities in deteriorating economic periods and are thus proved to be safe havens. A plausible explanation of this behavior is the risk-averse nature of investors who reallocate funds from risky to less risky securities and safe havens in the case of high economic risk. In addition, the lagged flows of all mutual fund classes are positively related to current GDP, exchange rate and investment. However, in the case of money supply, the lagged equity and balanced fund flows are positively, while the bond and balanced flows are negatively linked with it. This implies that equity and balanced flows signal positive changes in the expected monetary policy, i.e., an expected increase in money supply. Conversely, an increase in investment by fixed income funds (like bond flows) signals negative changes in the expected monetary policy, i.e. an expected decrease in money supply. Moreover, in the case of budget deficit and unemployment, the equity and balanced flows of the previous period have a negative influence, while the lagged bond and money market flows have a positive impact. This entails that equity and balanced flows decrease their trading with a higher expected budget deficit and unemployment situation in an economy whereas fixed income securities like bond and money market funds increase their trading with expected news of a high budget deficit and higher unemployment. Overall, we can conclude that good macroeconomic news reduces investments in bond and money market securities and leads to 143

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 12 Panel VAR model of total fund flows and business cycle variables.

Equity flows(t-1) Equity flows(t-2) Wald test p-value Bond flow(t-1) Bond flow(t-2) Wald test p-value Balanced flows(t-1) Balanced flows(t-2) Wald test p-value Money market Flow(t-1) Money market Flow(t-2) Wald test p-value

Flows

GDP

Inf

Ex

UE

MS

DG

Inv

1.129 (4.78)** 0.710 (4.82)** 0.00 0.211 (3.59)** 1.146 (4.52)** 0.00 1.998 (0.77) 1.483 (8.48)** 0.00 −0.001 (0.01) 0.118 (3.73)** 0.04

2.006 (2.97)** 2.001 (2.22)* 0.00 4.023 (3.98)** 3.124 (5.31)** 0.00 −0.738 (0.46) 4.382 (5.07)** 0.00 0.022 (0.14) 0.529 (3.54)** 0.04

0.101 (1.39) 0.156 (1.96) 0.09 −1.069 (1.52) −2.469 (1.26) 0.08 1.005 (0.74) 1.752 (1.56) 0.10 0.104 (0.58) −1.460 (0.28) 0.14

2.012 (2.57)* 2.098 (4.07)** 0.00 0.113 (6.49)** 0.240 (6.02)** 0.00 2.726 (1.44) 4.066 (5.90)** 0.00 0.592 (0.24) 2.151 (3.04)** 0.00

−0.014 (2.40)* −0.125 (2.83)** 0.00 0.004 (1.43) 2.099 (3.18)** 0.02 −0.156 (3.96)** −0.572 (6.79)** 0.00 1.152 (0.22) 2.395 (3.30)** 0.00

0.051 (4.67)** 45.231 (4.32)** 0.00 −0.190 (4.17)** −1.748 (6.86)** 0.00 2.297 (2.22)* 3.251 (3.66)** 0.00 −1.054 (0.39) −1.530 (2.50)* 0.05

−1.035 (4.01)** −3.562 (3.77)** 0.00 0.143 (3.00)** 1.665 (5.00)** 0.00 −0.396 (3.28)** −0.015 (0.24) 0.00 0.677 (4.66)** 4.024 (5.52)** 0.00

0.012 (2.92)** 3.393 (3.72)** 0.00 0.123 (2.93)** 1.354 (4.85)** 0.00 0.619 (4.56)** 1.023 (3.51)** 0.00 1.850 (2.26)* 2.436 (5.17)** 0.00

Notes: The table displays the result of the panel VAR model estimated by GMM of the net fund flows and business cycle variables. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

an increase in equity and equity-related investment in the stock market. This suggests that fixed income securities are safe havens in times of financial and economic crisis and that investors direct flows away from equity-based funds to fixed income-type funds in times of poor economic conditions. The results provide sufficient evidence to suggest that mutual funds not only incorporate economic information in their investment decisions but also help in predicting prospective economic conditions. This is consistent with the findings of Jank (2012) that mutual funds are forward-looking and predict expected real economic activity. The findings also support the study by Ferson and Kim (2012), who confirm that fund flows are linked to economic variables.26 4.3.6. Flow-investors’ expectations and the subprime mortgage crisis The analysis presented in the previous section is based on the complete sample from 1996Q1 to 2017Q3. However, it is possible that our sample is affected by the changes in monetary policy actions made by central banks to counter the subprime crisis.27 This period is reported to have a low interest rate, which motivated investors to search for better yield. Although the instability in financial markets in the pre- and post-subprime mortgage financial crisis periods has been captured in data estimation by introducing country and time fixed effects, we reproduce it for split sample periods with subsamples from January 1996 to June 2007 (pre-crisis) and from January 2010 to September 2017 (post-crisis) as a supplementary robustness check. The pre-crisis subsample ends in July 2007 because the subprime mortgage crisis started to show its explicit effects in August 2007. The sample has been isolated into two groups on the basis of the analysis carried out in previous studies (Kenourgios et al., 2011; Khan et al., 2016a,b). The estimation results for subsamples are reported in Tables 13–15. The analysis has been performed using all classes of fund flows and measures of investors’ expectations. A notable observation for this analysis is that the coefficients of fund flows and measures of investors’ expectations are found to be greater in the post-subprime mortgage crisis period. This indicates that the mutual funds and measures of investors’ expectations show a stronger relationship after the subprime mortgage crisis. This finding also supports the notion that after the period of unusually low interest rates offered by banks, investors prefer to have reasonable returns with safer investments such as mutual funds. Overall, the behavior of all other variables across all sample periods is similar to the results from the main estimation. Therefore, our findings are robust across different sample periods. 5. Conclusion and policy implications We extended the work of Jank (2012) to analyze the relationship of four mutual fund classes with measures of investors’ expectations and business cycle variables over the 1996Q1-2017Q3 period in BRICS markets. Additionally, we included additional variables such as money supply, the budget deficit to GDP ratio, the investment ratio and unemployment in Jank’s (2012) model to provide detailed insights into the relationship between funds flows and business cycle variables. The results suggest that fund flows are significantly related to alternative measures of investors’ expectations. An increase in the 26

IRF and FEVD of the flow-business cycle relationship are reported and discussed in Appendix A.1 and A.2. The correlation table shows that the money supply growth rate and the Treasury bill rate have the highest correlation with equity flows and thus the fund flows might be affected by the monetary policy indicator. We are grateful to an anonymous referee for highlighting this aspect. 27

144

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 13 Relationship between equity flows, bond flows and measures of investors’ expectations in the pre- and post-subprime mortgage crisis. Equity Flows

Flows t-1 Flows t-2 Wald test p-value TB t-1 TB t-2 Wald test p-value TS t-1 TS t-2 Wald test p-value DY t-1 DY t-2 Wald test p-value

Bond Flows

Subsample (pre-crisis)

Subsample (post-crisis)

Subsample (pre-crisis)

Subsample (post-crisis)

0.150 (0.99) 0.210 (2.29)* 0.03 0.006 (2.27)* 0.06 (2.58)* 0.00 −0.282 (2.20)* −0.016 (0.02) 0.05 −0.124 (2.34)* −0.021 (2.60)* 0.00

−0.220 (2.28)* −0.250 (2.81)** 0.00 0.011 (2.67)* 0.321 (2.86)** 0.00 −0.292 (2.99)** −0.027 (2.50)* 0.00 −0.162 (2.54)* −0.051 (2.78)* 0.00

0.019 (0.19) −0.011 (0.94) 0.09 −0.133 (2.48)* −0.014 (2.15)* 0.03 0.103 (2.24)* 0.172 (2.51)* 0.00 0.080 (2.78)* 0.071 (2.48)* 0.02

0.029 (2.19)* −0.013 (0.91) 0.05 −0.134 (2.67)* −0.024 (2.90)** 0.00 0.114 (2.74)* 0.192 (2.59)* 0.00 0.097 (2.98)** 0.075 (2.56)* 0.00

Notes: The table depicts the results of a panel VAR model estimated by GMM of net aggregate fund flows (equity and bond) and measures of investors’ expectations. The pre-crisis period consists of a subsample from January 1996 to June 2007 and the post-crisis period consists of a subsample from January 2010 to September 2017. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively. Table 14 Relationship between balanced flows, money market flows and measures of investors’ expectations in the pre- and post-subprime mortgage crisis. Balanced Flows

Flows t-1 Flows t-2 Wald test p-value TB t-1 TB t-2 Wald test p-value TS t-1 TS t-2 Wald test p-value DY t-1 DY t-2 Wald test p-value

Money Market Flows

Subsample (pre-crisis)

Subsample (post-crisis)

Subsample (pre-crisis)

Subsample (post-crisis)

0.710 (2.72)* 0.150 (2.12)* 0.05 0.004 (2.94)** 0.034 (2.14)** 0.03 −0.007 (2.18)* −0.037 (3.18)** 0.02 −0.210 (2.81)** −0.202 (2.09)* 0.00

0.921 (3.60)** 0.230 (2.97)** 0.00 0.021 (3.60)** 0.124 (2.94)** 0.00 −0.021 (4.60)** −0.154 (4.20)** 0.00 −0.260 (3.22)** −0.216 (2.66)* 0.00

0.201 (0.22) 0.134 (0.32) 0.13 −0.015 (2.19)* −0.065 (2.09)* 0.02 0.09 (2.96)** 0.023 (2.06)* 0.00 0.218 (0.78) 0.270 (2.18)* 0.05

−0.274 (3.27)** 0.170 (2.65)* 0.00 −0.216 (2.84)** −0.216 (2.64)* 0.00 0.097 (3.35)** 0.030 (2.92)** 0.00 0.261 (2.69)* 0.281 (2.23)* 0.00

Notes: The table depicts the results of a panel VAR model estimated by GMM of net aggregate fund flows (balanced and money market) and measures of investors’ expectations. The pre-crisis period consists of a subsample from January 1996 to June 2007 and the post-crisis period consists of a subsample from January 2010 to September 2017. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are provided in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

TB ratio indicates better expected economic conditions and thus has a positive (negative) effect on equity and balanced flows (bond and money market flows). Conversely, an increase in the dividend yield and term spread signals a poor state of the economy, thus reducing (increasing) equity and balanced flows (bond and money market flows). The results suggest that fund flows carry economic information in themselves. Fund flows are forward-looking and assist in forecasting real economic conditions. We also tested the flow-business cycle relationship as a robustness check and found that increases in GDP, domestic real 145

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Table 15 Panel VAR model of total fund flows and measures of investors’ future expectations in the pre- and post-subprime mortgage crisis. Subsample (pre-crisis)

Equity flows(t-1) Equity flows(t-2) Wald test p-value Bond flow(t-1) Bond flow(t-2) Wald test p-value Balanced flows(t-1) Balanced flows(t-2) Wald test p-value Money market Flow(t-1) Money market flow(t-2) Wald test p-value

Subsample (post-crisis)

TB

TS

DY

TB

TS

DY

1.443 (2.56)* 1.800 (2.92)** 0.00 −0.412 (2.16)* −0.102 (2.01)* 0.05 0.013 (2.13)* 0.201 (2.20)* 0.02 −0.021 (2.09)* −0.052 (2.29)* 0.05

−0.901 (2.06)* −0.093 (3.35)** 0.00 0.912 (2.06)* 0.699 (2.49)* 0.03 −0.010 (2.08)* −0.031 (2.29)* 0.03 −0.051 (0.28) 0.030 (2.21)* 0.05

−0.404 (1.67) −1.721 (2.51)* 0.03 0.403 (2.22)* 0.298 (2.14)* 0.05 −0.009 (3.01)** −0.011 (0.10) 0.04 0.076 (1.72) 0.143 (2.94)** 0.02

1.475 (2.86)** 1.867 (4.91)** 0.00 −0.475 (2.56)* −0.107 (2.30)* 0.00 0.023 (2.93)** 0.216 (2.80)** 0.00 −0.028 (2.39)* −0.086 (2.99)** 0.00

−0.938 (2.26)* −0.103 (4.32)** 0.00 0.938 (2.26)* 0.706 (2.99)** 0.00 −0.015 (2.28)* −0.071 (2.89)** 0.00 −0.025 (0.68) 0.031 (2.91)** 0.00

−0.412 (1.73) −1.791 (3.51)** 0.00 0.412 (2.83)** 0.301 (2.54)* 0.00 −0.010 (3.29)** −0.013 (0.11) 0.03 0.092 (2.02)* 0.150 (3.54)** 0.00

Notes: The table displays the results of the panel VAR model estimated by GMM of the net fund flows and measures of investors’ future expectations. The pre-crisis period consists of a subsample from January 1996-June 2007 and the post-crisis period consists of a subsample from January 2010 to September 2017. The reported numbers display the coefficients of regressing column variables on lags of row variables. The T-statistics are in parentheses. ** and * indicate significance at the 1% and 5% level, respectively.

investment and capital formation in the economy all boost fund flows and have a positive effect on both risky securities (e.g. equity and balanced flows) and less risky securities (e.g. bond and money market flows). In addition, changes in monetary and fiscal policy have a direct impact on fund flows. An increase in the money supply indicates better than expected economic conditions, whereas an increase in the budget deficit and unemployment signals a poor state of the economy. The findings of the study suggest that a lower money supply curtails the purchasing power of investors and switches their preference to investing in secure avenues. On the other hand, an increase in the budget deficit and a rise in unemployment tend to reduce equity-related investments and increase fixedincome investments. Our findings suggest that risky funds, i.e. equity and balanced funds, are more correlated to those macroeconomic variables that contain good economic news, whereas less risky securities like bond and money market fund flows are more reactive to variables containing poor economic information. These findings differ from Jank’s (2012) study, which states that riskier funds are also more related to all macroeconomic news proxies than less risky funds. We infer that investors switch their investment from risky to less risky funds in times of economic crises. The findings of this study could be of help to investors and portfolio managers in making efficient investment and asset allocation decisions at the international level, particularly in developing regional countries like the BRICS. Professional managers require a detailed understanding, sufficient experience, knowledge, evaluation and assessment of the financial security market and business sector of the economy. The findings provide significant information for portfolio managers concerning flight to quality, since investors make flight to quality allocating decisions and increase their portfolio returns by shifting investment from equity to fixed income securities in case of economic downturns and vice versa. Moreover, determining the predictive ability of mutual fund flows may facilitate forecasting and planning the future state of economic health for policymakers and investors. Economic conditions influence investors’ investment decisions and help them transfer their investments to safe havens in case of poor economic prospects. For international diversified portfolio investors, it is imperative to study economic conditions, as they affect security flows and investment. Investment and asset allocation decisions by mutual funds are beneficial to the performance of financial markets and the economy of developing countries. It would be interesting to see the relationship between mutual fund flows and the macroeconomy by applying wavelet-based techniques to isolate components that capture information across different time scales/frequencies and determine at which horizon/ frequencies the relationship is more dominant. In addition, there might be the possibility of market timing by mutual funds, which can be investigated further by taking daily data. Moreover, future research can determine the relationship between exchanges in traded funds and the business cycle using daily or monthly data.

146

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Appendix A A.1 Impulse Response Function (IRF) Before the estimation of IRF and FEVD, the stability condition of the estimated PVAR is checked. Fig. A2 of eigenvalues confirms that the estimates are stable (Fig. A1). All the eigenvalues lie inside the unit circle. PVAR satisfies the stability condition. 28 The graph of ORIFs is presented in Fig. A3. The 5% error bands are estimated using the Gaussian approximation generated by

Fig. A1. PVAR stability check of flow-measure of investors’ expectations model.

Fig. A2. PVAR stability check of flow-business cycle model.

28 The result for stability is the same for all PVAR models reported in Tables 11 and 12. All PVAR models are stable. The figures are not shown for brevity purposes.

147

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Fig. A3. Mutual fund flows and business cycle variables. Notes: The X-axis shows the number of time periods and the Y-axis shows the unit shock standard deviations. The blue line shows the orthogonalized shock and response of one variable on another variable. The red lines represent the plus and minus two standard deviation bands.

Monte Carlo simulation with 1000 reps.29 Fig. A3 depicts graphs of the impulses and responses of MF flows and macroeconomic variables. The response of equity flows to GDP shocks is positive in the estimated coefficients. This is expected, as equity flows and GDP are positively correlated. Similar patterns can be observed in the response of flows to money supply, investment and exchange rate shocks, since these variables signal good economic conditions. However, inflation shocks have an insignificant effect on equity flows. The response of equity flows is negative to unemployment rate shocks and deficit to GDP rate shocks. This is obvious, as the unemployment and deficit to GDP ratio signals poor economic conditions. Similar behavior can be witnessed in response to balanced flows to shocks of macroeconomic variables. The response of bond flows to macroeconomic shocks is positive with the exception of money supply. The response of money market flows to macroeconomic shocks is positive with the exception of money supply and inflation. Inflation shocks have almost insignificant effects on all classes of fund flows due to its being captured mostly by GDP and money supply. Moreover, the response of macroeconomic variables to shocks of equity flows is mixed, with a negative reaction toward unemployment and budget deficits, and a positive response of GDP, money supply and exchange rates. The responses of all macroeconomic variables to shocks are significant except for inflation. The response of unemployment and budget deficit to shocks of bond flows is positive. Overall, the results are consistent with the PVAR estimates in Section 4.3.5. A.2 Error Variance Factor Decomposition (FEVD) The FEVD for the PVAR model is presented in Table A1. Market returns and macroeconomic variables explain more of the fund flows’ variations, ten periods ahead. It is observed that equity flows explain 50% of total variations by flows themselves, bond flows Table A1 FEVD of fund flows and business cycle variables. Flows

Flows

GDP

Inf

Ex

UE

MS

DG

Inv

Equity Bond Balanced Money Market

0.50 0.70 0.49 0.60

0.18 0.089 0.19 0.16

0.001 0.001 0.002 0.001

0.10 0.05 0.119 0.05

0.03 0.07 0.04 0.099

0.09 0.02 0.03 0.02

0.03 0.05 0.04 0.05

0.069 0.02 0.079 0.02

Notes: The percent of variation in the row variable (10 periods ahead) is explained by the column variable. FEVD standard errors and confidence intervals are based on Monte Carlo simulations.

29

The study follows the procedure of generating impulse response function from Love and Zicchino (2006) and Abrigo and Love (2016). 148

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

70%, balanced 49% and money market 60%. GDP explains 18% and 19% of equity and balanced flows. The exchange rate and the money supply rate explain greater variation in equity flows of about 15% and 9% compared to other flows. This is due to the fact that the exchange rate and money supply encompass positive economic news and equity flows increase with better economic news. Inflation has a very small impact on variations in fund flows. The unemployment rate has a greater impact on bond and money market flows at about 7% and 9.9%, respectively. In addition, deficit to GDP explains 5% of bond and money market flows each. The unemployment rate and deficit to GDP signal negative news about economy and bond flows, and money market flows increase in times of expected worse economic situations. Investment explains 8% and 7% of the variations in balanced and equity flows. Overall, the findings corroborate the reported PVAR results.

References Abrigo, M.R., Love, I., 2016. Estimation of panel vector autoregression in stata. Stata J. 16 (3), 778–804. Alexakis, C., Dasilas, A., Grose, C., 2013. Asymmetric dynamic relations between stock prices and mutual fund units in Japan. An application of hidden cointegration technique. Int. Rev. Financ. Anal. 28, 1–8. Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. J. Econom. 68 (1), 29–51. Assenmacher, K., Gerlach, S., 2008. Monetary Policy, Asset Prices and Macroeconomic Conditions: a Panel-var Study. National Bank of Belgium Working Paper (149). Aydogan, B., Vardar, G., Tunç, G., 2014. The interaction of mutual fund flows and stock returns: Evidence from the turkish capital market/yatirim fonlari ile hisse senedi piyasasinin etkilesimi: Türk sermaye piyasasindan bulgular. Ege Akad. Bakis 14 (2), 163. Bali, T.G., Brown, S.J., Caglayan, M.O., 2014. Macroeconomic risk and hedge fund returns. J. Financ. Econ. 114 (1), 1–19. Barro, R.J., 1990. The stock market and investment. Rev. Financ. Stud. 3 (1), 115–131. Ben-Rephael, A., Kandel, S., Wohl, A., 2011. The price pressure of aggregate mutual fund flows. J. Financ. Quant. Anal. 46 (02), 585–603. Binswanger, M., 2000. Stock returns and real activity: Is there still a connection? Appl. Financ. Econ. 10 (4), 379–387. Binswanger, M., 2004. Stock returns and real activity in the G-7 countries: Did the relationship change during the 1980s? Q. Rev. Econ. Finance 44 (2), 237–252. Bloom, N., 2009. The impact of uncertainty shocks. Econometrica 77 (3), 623–685. Bloom, N., Bond, S., Van Reenen, J., 2007. Uncertainty and investment dynamics. Rev. Econ. Stud. 74 (2), 391–415. Bodie, Z., 1976. Common stocks as a hedge against inflation. J. Finance 31 (2), 459–470. Campbell, J.Y., 1991. A Variance Decomposition for Stock Returns. Retrieved from. http://www.nber.org/papers/w3246. Campbell, J.Y., Shiller, R.J., 1988. The dividend-price ratio and expectations of future dividends and discount factors. Rev. Financ. Stud. 1 (3), 195–228. Campbell, J.Y., Thompson, S.B., 2008. Predicting excess stock returns out of sample: can anything beat the historical average? Rev. Financ. Stud. 21 (4), 1509–1531. Campbell, J.Y., Viceira, L.M., 1996. Consumption and portfolio decisions when expected returns are time varying. Q. J. Econ. 114 (2), 433–495. Cao, C., Chang, E.C., Wang, Y., 2008. An empirical analysis of the dynamic relationship between mutual fund flow and market return volatility. J. Bank. Financ. 32 (10), 2111–2123. Chatziantoniou, I., Duffy, D., Filis, G., 2013. Stock market response to monetary and fiscal policy shocks: multi-country evidence. Econ. Model. 30, 754–769. Chen, N.-F., Roll, R., Ross, S.A., 1986. Economic forces and the stock market. J. Bus. 383–403. Choi, J.J., Hauser, S., Kopecky, K.J., 1999. Does the stock market predict real activity? Time series evidence from the g-7 countries. J. Bank. Financ. 23 (12), 1771–1792. Click, R.W., Plummer, M.G., 2005. Stock market integration in Asean after the Asian financial crisis. J. Asian Econ. 16 (1), 5–28. Du, D., 2006. Monetary policy, stock returns and inflation. J. Econ. Bus. 58 (1), 36–54. Du, D., Denning, K., Zhao, X., 2012. Real aggregate activity and stock returns. J. Econ. Bus. 64 (5), 323–337. Edelen, R.M., 1999. Investor flows and the assessed performance of open-end mutual funds. J. Financ. Econ. 53 (3), 439–466. Edelen, R.M., Warner, J.B., 2001. Aggregate price effects of institutional trading: a study of mutual fund flow and market returns. J. Financ. Econ. 59 (2), 195–220. Fama, E.F., 1981. Stock returns, real activity, inflation, and money. Am. Econ. Rev. 545–565. Fama, E.F., 1990. Stock returns, expected returns, and real activity. J. Finance 45 (4), 1089–1108. Fama, E.F., French, K.R., 1988. Dividend yields and expected stock returns. J. Financ. Econ. 22 (1), 3–25. Fama, E.F., French, K.R., 1989. Business conditions and expected returns on stocks and bonds. J. Financ. Econ. 25 (1), 23–49. Fama, E.F., Schwert, G.W., 1977. Asset returns and inflation. J. Financ. Econ. 5 (2), 115–146. Ferreira, M.A., Keswani, A., Miguel, A.F., Ramos, S.B., 2012. The flow-performance relationship around the world. J. Bank. Financ. 36 (6), 1759–1780. Ferson, W.E., Kim, M.S., 2012. The factor structure of mutual fund flows. Int. J. Portf. Anal. Manag. 1 (2), 112–143. Ferson, W.E., Schadt, R.W., 1996. Measuring fund strategy and performance in changing economic conditions. J. Finance 51 (2), 425–461. Flannery, M.J., Protopapadakis, A.A., 2002. Macroeconomic factors do influence aggregate stock returns. Rev. Financ. Stud. 15 (3), 751–782. Galeotti, M., Schiantarelli, F., 1994. Stock market volatility and investment: do only fundamentals matter? Economica 147–165. Geske, R., Roll, R., 1983. The fiscal and monetary linkage between stock returns and inflation. J. Finance 38 (1), 1–33. Goetzmann, W.N., Massa, M., 1999. Index Funds and Stock Market Growth. Retrieved from. http://www.nber.org/papers/w7033. Haddad, M., Lim, J.J., Pancaro, C., Saborowski, C., 2013. Trade openness reduces growth volatility when countries are well diversified. Can. J. Econ./Revue canadienne d’économique 46 (2), 765–790. Halim, L., 2007. United States Emerging Markets Mutual Funds Index and the Growth of Emerging Market Economies an Empirical Study on Asian Economies (Doctor of Philosophy). The City University of New York. Hansen, L.P., 1982. Large sample properties of generalized method of moments estimators. Econometrica 1029–1054. Hodrick, R.J., 1992. Dividend yields and expected stock returns: alternative procedures for inference and measurement. Rev. Financ. Stud. 5 (3), 357–386. Holtz-Eakin, D., Newey, W., Rosen, H.S., 1988. Estimating vector autoregressions with panel data. Econometrica 1371–1395. Jank, S., 2012. Mutual fund flows, expected returns, and the real economy. J. Bank. Financ. 36 (11), 3060–3070. Kacperczyk, M., Van Nieuwerburgh, S., Veldkamp, L., 2013. Time‐varying fund manager skill. J. Finance 69 (4), 1455–1484. Kaul, G., 1987. Stock returns and inflation: the role of the monetary sector. J. Financ. Econ. 18 (2), 253–276. Kaul, A., Phillips, B., 2008. Economic Conditions, Flight to Quality and Mutual Fund Flows. Unpublished Manuscript. University of Alberta. Keim, D.B., Stambaugh, R.F., 1986. Predicting returns in the stock and bond markets. J. Financ. Econ. 17 (2), 357–390. Kenourgios, D., Samitas, A., Paltalidis, N., 2011. Financial crises and stock market contagion in a multivariate time-varying asymmetric framework. J. Int. Financ. Mark. Inst. Money 21 (1), 92–106. Khan, H.H., Ahmad, R.B., Gee, C.S., 2016a. Market structure, financial dependence and industrial growth: evidence from the banking industry in emerging Asian economies. PLoS One 11 (8), e0160452. Khan, H.H., Ahmed, R.B., Gee, C.S., 2016b. Bank competition and monetary policy transmission through the bank lending channel: evidence from Asean. Int. Rev. Econ. Financ. 44, 19–39. Khorana, A., Servaes, H., Tufano, P., 2005. Explaining the size of the mutual fund industry around the world. J. Financ. Econ. 78 (1), 145–185. Klapper, L., Sulla, Vc., Vittas, D., 2004. The development of mutual funds around the world. Emerg. Mark. Rev. 5 (1), 1–38. Kopsch, F., Song, H.-S., Wilhelmsson, M., Johnson, D., 2015. Determinants of mutual fund flows. Manag. Financ. 41 (1).

149

Economic Systems 43 (2019) 130–150

F. Qureshi, et al.

Lamont, O., 1998. Earnings and expected returns. J. Finance 53 (5), 1563–1587. Laopodis, N.T., 2009. Fiscal policy and stock market efficiency: evidence for the united states. Q. Rev. Econ. Financ. 49 (2), 633–650. Lettau, M., Ludvigson, S., 2001. Consumption, aggregate wealth, and expected stock returns. J. Finance 56 (3), 815–849. Lettau, M., Ludvigson, S.C., 2005. Expected returns and expected dividend growth. J. Financ. Econ. 76 (3), 583–626. Love, I., Zicchino, L., 2006. Financial development and dynamic investment behavior: evidence from panel var. Q. Rev. Econ. Financ. 46 (2), 190–210. Mensi, W., Hammoudeh, S., Reboredo, J.C., Nguyen, D.K., 2014. Do global factors impact brics stock markets? A quantile regression approach. Emerg. Mark. Rev. 19, 1–17. Merton, R.C., 1973. An intertemporal capital asset pricing model. Econometrica 867–887. Narayan, P.K., Narayan, S., Thuraisamy, K.S., 2014. Can institutions and macroeconomic factors predict stock returns in emerging markets? Emerg. Mark. Rev. 19, 77–95. O’Neill, J., 2011. The Growth Map: Economic Opportunity in the Brics and Beyond. Penguin, UK. Oh, N.Y., Parwada, J.T., 2007. Relations between mutual fund flows and stock market returns in korea. J. Int. Financ. Mark. Inst. Money 17 (2), 140–151. Roodman, D., 2006. How to Do xtabond2: an Introduction to Difference and System Gmm in Stata. Center for Global Development Working Paper. pp. 103. Samargandi, N., Kutan, A.M., 2016a. Private credit spillovers and economic growth: evidence from brics countries. J. Int. Financ. Mark. Inst. Money 44, 56–84. Samargandi, N., Kutan, A.M., 2016b. Private credit spillovers and economic growth: evidence from brics countries. J. Int. Financ. Mark. Inst. Money. Schwert, G.W., 1990. Stock returns and real activity: a century of evidence. J. Finance 45 (4), 1237–1257. Shanken, J., 1990. Intertemporal asset pricing: An empirical investigation. J. Econom. 45 (1), 99–120. Shiller, R.J., Fischer, S., Friedman, B.M., 1984. Stock prices and social dynamics. Brookings Pap. Econ. Act. 457–510. Sirri, E.R., Tufano, P., 1998. Costly search and mutual fund flows. J. Finance 53 (5), 1589–1622. Thomas, A., Spataro, L., Mathew, N., 2014. Pension funds and stock market volatility: An empirical analysis of oecd countries. J. Financ. Stab. 11, 92–103. Verardi, V., Dehon, C., 2010. Multivariate outlier detection in Stata. Stata J. 10 (2), 259–266. Warther, V.A., 1995. Aggregate mutual fund flows and security returns. J. Financ. Econ. 39 (2), 209–235. Westerlund, J., Narayan, P.K., Zheng, X., 2015. Testing for stock return predictability in a large chinese panel. Emerg. Mark. Rev. 24, 81–100. Wooldridge, J.M., 2001. Applications of generalized method of moments estimation. J. Econ. Perspect. 87–100.

150