“Mutual fund flow-performance dynamics under different market conditions in South Africa”

Questions regarding the specific factors that drive continuous cash allocations by investors into portfolios of actively managed funds, despite consistent underperformance, continue to remain an inexhaustive aspect of the literature that calls for further investigations. This study assesses the dynamic relationship between fund flow and performance of equity mutual funds in South Africa under different market conditions. The study employs a GMM technique to analyze the panel data of 52 South African equity mutual funds from 2006 to 2019. The analysis found that convexity is prevalent in the flow-performance relationship, where fund contributors in subsequent periods allocate recent underperforming and outperforming funds disproportionate cash. This finding is evident in the lack of significance in the past performance effects on subsequent fund flows. The study found that lagged fund flows, fund size, fund risk, and market risk drive subsequent fund flows under changing conditions of the general market and fund markets. Overall, it is posited that fund contributors and asset admin- istrators adapt to prevailing market dynamics relative to trading decisions. As a result, this affirms the normative guidelines of the Adaptive Markets Hypothesis, leading to the conclusion that exogenous factors drive fluctuations in fund flows in South Africa.


INTRODUCTION
The South African fund industry's trend statistics show that average returns earned across equity fund managers trailed the market by 34.01 percent in one year, trailed it by 84.66 percent in three years, and recorded a significant underperformance of 91.03 percent in five years (S&P, 2019). The evidence shows that within the five years under study (2014)(2015)(2016)(2017)(2018), only 8.97 percent of active managers managed to beat the market, reflecting substantial inconsistencies in the pattern of performance by the few fund managers who succeeded in earning superior returns compared to the market (S&P, 2019). Nonetheless, the volume of new cash inflows into South African equity mutual funds increased, with over R1.9 trillion assets under management at the end of the third quarter of 2018 trading year (Rangongo, 2018).
Prior studies on the flow-performance relationship in South Africa are conducted in the context of stable market conditions, and hence cannot explain the enigmatic circumstances behind the increased fund flows against persistent underperformance by equity mutual funds in South Africa (Tan, 2015; Thobejane et al., 2017;Arendse et al., 2018). However, the Adaptive Market Hypothesis explanations suggest that the relationship between fund flow and performance will not hold under different market conditions, as individual markets experience

Flow-performance dynamics under different market conditions
The flow-performance relationship is dynamic under varying conditions of the market, with the degree of responsiveness of cash flow to funds being more pronounced under bullish market conditions than under bearish markets (Gottesman et al., 2013;Jun et al., 2014). Also, investors generally exhibit a high degree of responsiveness to fund managers' recent outperformance compared to their reactions to underperformance (Rao et al., 2016). Furthermore, scholars have explained that the influence of contemporaneous performance is positively significant on cash flows after conditioning for endogeneity (Qian et al., 2014). However, evidence premised on fund contributors' reliance on past performance affirms a converse relationship between past performance and subsequent fund flows (Lou, 2012;Chen & Qin, 2014). A reverse interaction between flow and performance culminates in a situation where funds that benefit from increased cash flow levels can perform better than funds that secured limited cash flows in the past (Chen & Qin, 2014). This dynamic results from specific signaling indications embedded in past cash flows that are essential drivers of funds' future performance, notwithstanding its inadequacy to compensate for the cost involved in pursuing and utilizing such leading information (Lou, 2012;Nenninger & Rakowski, 2014).
Evidence has shown that factors that underpin periodic changeovers in market conditions and prices of financial securities are adequate predictors of investor cash allocation decisions on fund managers (Kurov, 2010 In the above context, changes in market conditions are more related to retail managers' cash flow than corporate managers, which indicates that differences in fund returns are often driven by irrational human tendencies of retail investors (Kim & Park, 2015;Ryu et al., 2017). Cognizant of the average investor's reactionary tendencies, the average fund manager is motivated to remain strategic in stock-picking under bull markets while exhibiting considerable caution in market timing under bearish market conditions to sustain fund flow (Huang et al., 2011). This behavior by fund managers is mainly associated with the view of the slow, sentimental response of equity investors to prevailing market trends under bullish market conditions compared to occasions of downturns in asset prices (Chalmers et al., 2013;Smales, 2017).
From the literature (Ippolito, 1992;Sirri & Tufano, 1998;Huang et al., 2007), issues that relate to marketing strategy and advertising, managerial skill, and strategy, investor search, investor participa-tion fees, and investor cognitive dissonance or disposition effects are associated with convexity in the flow-performance relationship under different conditions of the market (as cited in Jun et al., 2014, p. 2). Based on the discussed dynamics in the flow-performance sensitivities, it is postulated that the degree of responsiveness in the relationship between fund flow and performance is more evident under bullish states of the market than under bearish conditions.

The South African mutual fund industry in context
The mutual fund industry in South Africa remains an integral aspect of the national economic system. Total estimated wealth of investable fund assets in excess of R2.4 trillion were held in over 1,590 portfolios as of the end of the third quarter of 2019 (ASISA, 2019). The recent report attributes the momentous leap in investor cash flows to fund managers in South Africa from the previous year's figure of R1.9 to windfall benefits accruing from a robust national financial system (Rangongo, 2018;Glow, 2020). Additionally, the analysts link this state of flow-performance relationship to international trade conflict among major economies (for example, the trade war between the United States and China) and recent economic obstacles faced by major Western markets that gradually diminished the returns of most multinational corporations (Glow, 2020 Analysts project a significant rise in South African mutual fund assets due to a resurgence in stock investment in 2019 and beyond (Ziphethe-Makola, 2017; Glow, 2020). In this context, knowledge of the influencing dynamics in the flow-performance relationship under changing market conditions become an essential toolkit for fund contributors and industry players for optimal investment decision-making. Given the above context, this study aims to assess the dynamic relationship between fund flow and performance of equity mutual funds in South Africa under different market conditions.
Given the above context, this study aims to assess the dynamic relationship between fund flow and performance of equity mutual funds in South Africa under different market conditions.

Data and variable description
An unbalanced panel data for quarterly observations from 2006 to 2019 sourced from McGregor BFA Library, S&P Capital IQ, and the ASISA website is used to achieve the study's objective. A total of 52 actively equity mutual funds are included in the sample for analysis. For a fund to be included, it should have had a minimum of six years of data for analysis, and the sample is selected based on data availability. In calculating South African equity funds' performance, quarterly returns of the price index of funds are logarithmically computed. Following the literature, fund performance by raw returns is formulated as follows: where it R is the return on fund i in quarter , t it P denotes the current price of fund i in quarter at , t 1 it P − is the price of fund in the previous period 1, t − and ln is the natural logarithm of the price index (Brooks, 2014).
This study utilizes the Johannesburg Stock Exchange All Share Index (JSE ALSI) as a proxy for market performance during the preceding year for the sample period. Following Nenninger and Rakowski (2014), fund flow is computed as the net quarterly percentage of cash flows accruing to a fund resulting from investor stock purchasing and redemption activity. A fund flow is expressed as follows:

Modeling fund flow-performance dynamics under different market conditions
According to Helwege and Liang (2004), it is expected that the average fund investor would continue to pick more stocks with funds under conditions where they are more confident about expected returns because prior states of the market influence investor decisions on mutual funds (as cited in Lee et al., 2011, p. 12). In this context, it is conjectured that the relationship between mutual fund flows and performance is dissimilar across different states of the equity market, where it is more pronounced under bullish markets than under bearish market conditions.

Flow-performance with changing conditions from the fund market
As a further test of the relationship between flow and performance under different market conditions, the analysis is extended to estimate the degree of responsiveness between future fund flow and lagged performance under different conditions of the fund market. In this context, the dynamic relationship between fund flow and performance of active managers under different con-

Correlation analysis
According to Dormann et al. (2013), a correlation of 0.7 and beyond among independent variables suggests the existence of a multicollinearity problem. From Table 2, the highest correlation is 0.62, which is between STDFND (the annualized standard deviation of a fund's monthly return in the past year) and STDMKT (annualized standard deviation of daily equity market return in the past year). The rest of the values are lower than 0.7, eliminating the possibility of the prevalence of multicollinearity issues among the set of independent variables employed in the analysis.  From the table, lagged annualized standard deviations of daily market returns (proxy for market risk) report a negative and significant coefficient.

Flow-performance dynamics under different market conditions
This result suggests that fund contributors are sensitive to market returns dispersions, where an increase in market volatility adversely impacts fund flow. This finding is consistent with the findings of Barber et al. (2016), according to which increased market volatilities affect investor stock-picking decisions on mutual funds. In terms of fund risk, it can be observed from Table 3  Having discussed the difference in GMM estimation results, these are subsequently compared to results obtained through a two-stage system GMM approach. The two-stage process is more robust from the literature than the difference GMM estimator that simultaneously obtains all parameter estimates. A significant advantage of the two-stage system is that misspecified assumptions on the time-invariant regressors do not influence the estimation results for the time-varying variables' coefficients. According to scholars (Arellano & Bond, 1991;Blundell & Bond, 1998;Roodman, 2009), it allows for exploiting the advantages of estimators relying on transformations to eliminate the unit-specific heterogeneity (as cited in Kripfganz & Schwarz, 2015, pp. 1, 2).
From Table 3, lagged flow shows a highly significant coefficient via the two-stage approach, which is similar to the level of significance obtained by the difference GMM approach. However, it is pos-itive in this instance, compared to a negative coefficient obtained by the first approach. This result implied that the past flow of funds remains a key predictor of subsequent flows under changing market conditions. Consistent with results obtained through the difference GMM technique, past performance's coefficient is positive and insignificant through the two-stage approach. This result suggests that generally, investors' cash allocations to active managers are not driven by fund managers' past performance under changing market conditions. However, prior studies (Huang et al., 2012;Chou & Hardin, 2014) suggest that fund contributors generally pursue recent performance. Unlike results obtained through the difference GMM estimation, the market condition variable shows a negative and insignificant coefficient via the two-stage approach. From the table, the results obtained for fund size through the system GMM technique is similar to evidence obtained under the difference GMM estimation, where it reports a positive and significant coefficient. However, its effect on fund flow appears directionally divergent across the two estimation techniques.
In the context of the system GMM approach, the fund size variable's coefficient is negative compared to a positive coefficient obtained by the first estimation technique, the difference GMM. This finding suggests that the volume of a fund's total net assets influences the investor cash allocation decisions under changing market conditions. This result contrasts evidence obtained by Ferreira et al. (2012), which suggests that investors are more confident when fund managers' asset base is adequately large, as large funds have an advantage of more investment opportunities over smaller funds. They can withstand dynamic market fluctuations. However, the evidence obtained for fund size via the system GMM approach is consistent with the earlier finding by Chou and Hardin (2014), which suggests that an increase in fund size deteriorates the general performance of funds, including fund flow.
Evidence from the system GMM estimation shows a negative and insignificant co-efficient for fund age, which is similar to results obtained by the difference GMM approach. This finding implies that mutual fund investors generally do not consider the number of years a fund has been in existence when making investment decisions in the context of changing market conditions. This evidence contrasts the findings of prior studies (Pástor et al., 2015;Rao et al., 2016) that fund contributors' preference is influenced by fund age because older funds grow slower than younger funds. Given that this study is conducted in changing market conditions, the current results are expected as investors adapt to market dynamics with time, as explained under the AMH (Lo, 2012). From Table  3, the system GMM estimation reports a positive and significant coefficient for the standard deviation of daily fund market returns, which is similar to results obtained for the equal variable via the difference GMM. However, the coefficient of this variable is positive in this context compared to a negative value obtained under the difference GMM approach. This result implies that the mar-  Table 4 reports the results of GMM estimations for flow-performance with changing conditions from the fund market. From the difference GMM results, lagged performance reports a negative and insignificant coefficient. This finding suggests that active managers' past performance does not influence their subsequent cash flows under changing the fund market conditions. This result departs from the expected positive and significant relationship between fund performance and subsequent flows under bullish conditions of the market, as documented by Gottesman et al. (2013). Given that the current analysis is conducted under changing conditions from the fund market and not the general equity market, this result is not surprising. The lagged flows by difference GMM estimation report a negative and insignificant coefficient from the  Table 4, lagged annualized standard deviation of monthly fund returns obtained a negative and insignificant coefficient under difference GMM. This result suggests that fund contributors are less sensitive to fund returns dispersions under changing fund market conditions. This evidence contrasts the literature's position that investor preference relative to mutual fund investments is affected by fund portfolios' associated risk (Jun et al., 2014). This result is expected as the fund market's changing conditions may not be the same as changing conditions of the general market to engender generalized assumptions.

Flow-performance with changing conditions from the fund market
From Table 4, lagged annualized standard deviation of daily market returns reports a positive and significant coefficient under both difference and two-stage GMM approaches. This evidence implies that market risk generally positively affects fund flow as the variable for market risk. Intuitively, an increase in the dispersion of benchmark returns drives flows to active managers' portfolios under changing fund market conditions. Evidence by Barber Pástor (2015). As shown in Table 4, the coefficient of the market condition variable is positive and significant. This result affirms the prior assumption that fund contributors' stock-picking actions increase significantly during periods of high fund cross-sectional performance than during low cross-sectional fund performance periods. Comparing the results of the difference GMM to the two-stage system GMM estimations, it is evident that lagged fund flow, fund size, market risk, and fund risk have a significant effect on fund flow under the difference estimation technique. On the other hand, lagged flow and market risk exert a positive effect on future fund flows, while fund size and fund risk negatively affect fund flow via the twostage system GMM approach.

CONCLUSION AND POLICY IMPLICATIONS
This study presents original perspectives on the relationship between mutual fund flow and performance under different conditions of the general equity and fund markets in South Africa. It is found that factors other than past performance drive investor assets to equity fund managers. This finding is evident in the lack of significance in the effect of past performance on subsequent fund flows, which is linked to the prevalence of convexity in the flow-performance relationship. The study found that lagged fund flows, fund size, fund risk, and market risk drive future fund flow under changing conditions of the general market and fund markets. Moreover, this study's findings support the position of the literature that flow-performance sensitivity is more pronounced under bullish market conditions than bearish market conditions. In general, this study concludes that fund contributors and asset administrators adapt to prevailing market dynamics relative to trading decisions, and as a result, affirms the normative guidelines of the Adaptive Market Hypothesis. This makes fund managers rely on non-performance metrics such as advertising a superior means of engendering sustainable fund flows under changing market conditions, which should focus on future studies. This study contributes to the literature on mutual funds by being the first to provide novel perspectives on the relationship between fund flow and performance under different conditions of the general equity market and the fund market in South Africa.