Do past performance and past cash flows explain current cash flows into retail superannuation funds in Australia?

Abstract:

This paper examines the link between current-quarter cash flows and both past performance and past cash flows using a sample of Australian retail superannuation fund data (managed growth and managed stable) drawn from the period 1994 to 2000. This is a rapidly growing sector within the superannuation industry and it reflects investment behaviour of smaller investors rather than institutions and large corporations. Using both the Gruber (1996) approach and panel-data analysis we find a positive relationship between past performance and current-quarter cash flows as well as evidence of persistence in cash flows over time. Panel-data analysis also identifies a positive relationship between current net cash flows and past performance and cash inflows as well as a negative relationship between current net cash flows and past outflows. Market-wide growth in the retail superannuation sector over the study period does not appear to be driving these results.

Keywords:

RETAIL SUPERANNUATION FUND; FUND PERFORMANCE; FUND CASH FLOWS.

1. Introduction

This paper concentrates on the link between current-period cash flows and past performance and between current-period cash flows and past cash flows. The investment fund literature has tended to focus on fund performance and, more recently, performance persistence. Studies on fund performance find that managed funds in the USA under-perform the market, even before expenses. (1) Studies in Australia (2) note similar results, although Bilson, Frino and Heaney (2004) find some evidence of superior performance in small retail funds and, to a limited extent, in retail growth funds.

The performance persistence literature is less clear-cut. While there is support for short-term performance persistence over one to three years (3), both Grinblatt and Titman (1992) and Malkiel (1995) show that past performance can predict future performance over longer periods, up to five years. These results have been questioned in the literature, with the suggestion that persistence is explained by factors other than performance. (4) There is little conclusive Australian evidence to support performance persistence. For example, Drew, Stanford and Veeraraghavan (2001) find no evidence of persistence over a one-year period when using retail superannuation fund raw returns or Sharpe index values. Hallahan (1999) finds similar results using Rollover funds. Further, Bilson, Frino and Heaney (forthcoming) show that retail fund performance persistence is sensitive to performance measure choice.

Performance persistence has important implications for investors. If performance persists, then fund track record should influence investor selection of managed funds, and so it is possible that strong performance in one period could attract cash flows into the fund in following periods and poor performance could lead to cash outflows. There is evidence of a positive relationship between various measures of performance and net cash flows in the literature for US mutual funds (Ippolito 1992; Gruber 1996; Sirri & Tufano 1998) and for Australian wholesale funds (Sawicki 2000, 2001).

Further analysis of this question in an Australian context is important, particularly with respect to Australian retail funds. The Australian managed funds industry has experienced substantial growth over the past decade, with much of this growth attributable to the growth in assets held by superannuation funds. In Australia, superannuation is paid into the funds by employer contributions under the Superannuation Guarantee Charge and by individuals as voluntary contributions. In the financial year 2000/2001, approximately 64% of all member contributions were made into retail superannuation funds and at 31 December 2001 retail funds held 32% of the total superannuation assets. Thus, retail funds form a substantial part of the total assets under management.

This study focuses on the question of whether cash flows follow past performance and whether there is persistence in cash flows over time. Sawicki (2000, 2001), in the most recent Australian research based on wholesale funds, notes a positive relationship between current cash flows and prior-period performance. While we also observe a positive correlation between net cash flows in the current-period and prior-period fund performance, we extend this analysis with the inclusion of separate prior-period cash inflow and outflow variables as well as prior-period performance. This more complex model clearly identifies the relative importance of prior-period cash flows and prior-period performance in determination of current-period net cash flows. An important contribution of this paper is the finding that, while prior-period performance has an impact, past cash inflows and outflows play an important role in explaining current-period net cash flows into retail funds.

We use a survivorship bias-free sample of 398 Australian managed-growth and managed-stable retail superannuation funds spanning the period September 1994 to June 2000. While our sample is limited to quarterly observations it does provide access to separate cash inflow and cash outflow data as well as net cash flow data. This finer cash flow information is not evident in prior Australian research. The remainder of this study is organised as follows. Section 2 discusses the literature. Section 3 outlines the managed fund and benchmark data used. Section 4 presents the analysis and we finish with a summary in section 5.

2. Cash Flows and Performance

The prior performance of funds is generally reported in the Australian press in the form of raw returns or rankings. Even though there is often a disclaimer attached that states 'historical performance is not indicative of future performance' (MLC MasterKey Booklet 2002), prior performance can have an impact on investors' evaluation of funds and their subsequent fund selection choice.

There's a disclaimer in most funds' advertising that past performance is no guide to future performance. The axiom that industry players mutter, however, is that it is an indicator of where funds will flow. (White 2002, p. 25)

This relationship between performance and net cash flows (5) has been analysed by Ippolito (1992), Gruber (1996), Sirri and Tufano (1998) and Sawicki (2000, 2001). While these results are not sensitive to performance measure choice there is evidence of asymmetry in the reaction of investors to prior performance in the USA. It is expected that rational investors will invest in winners and disinvest in losers. However, both Ippolito (1992) and Sirri and Tufano (1998) find that investors appear to be attracted to good past performers without exiting from poor past performers. Ippolito's (1992) study hypothesises that the differential response of investors may be associated with the high costs of selling units in the funds. In Australia, Sawicki's (2000, 2001) studies on Australian wholesale funds between 1980-1995 also indicate a significant positive relationship between net cash flow and various performance metrics, however, Sawicki finds no evidence to support asymmetry in investor response to past performance for her sample of Australian wholesale funds.

The impact of prior cash flows on the level of current cash flows into funds is rarely dealt with in the literature and there is no evidence of analysis of this question using Australia data. Bergstresser and Poterba (2002) note the existence of persistence in net cash inflows and, although their analysis deals with the determinants of cash inflows, they do not include previous period cash flows as explanatory variables in their analysis. Sirri and Tuffano (1998) also analyse fund cash flows though they use current-period cash flows for funds drawn from the same fund objective category in the same time period.

Analysis of the relationship between past performance and cash flows for Australian funds is limited to analysis of wholesale funds and there is little evidence of analysis of the impact of past cash flows on present fund cash flows. In light of this, further study appears warranted and this study contributes to the Australian literature by examining both the relationship between past performance and current cash flows as well as the relationship between past cash flows and current cash flows, focusing on the important retail sector of the superannuation industry.

3. Data

Plan For Life supplies our Australian retail fund data. (6) The data is drawn from a larger data set consisting of 808 funds with 417 managed growth and managed stable funds covering the period September 1991 to June 2000. Separate cash inflow and cash outflow data is only available after June 1994 and this restriction reduces the sample to 399 funds existing after June 1994. One fund is dropped from this group due to a lack of separate cash inflow and outflow information giving a final sample of 398 funds in existence after June 1994 with full cash flow details. Managed growth and managed stable funds are included in the analysis to facilitate comparisons with overseas research given the equity focus of these funds. Managed growth funds typically have larger exposure to equities and smaller exposure to fixed interest securities, relative to managed stable funds though both have substantial equities exposure.

We use all available funds to avoid survivorship bias as we have access to data from all funds that operated during the period. For each fund the database identifies the investment management company, product name, fund investment category or style, year and quarter, funds under management (FUM) at the beginning and end of the quarter, investment earnings, total cash inflows and total cash outflows. This provides the opportunity to disaggregate cash flows rather than limit analysis to net cash flows as occurs in Sawicki (2000).

Investment earnings represent the change in funds under management after adjustment for net cash flows and are net of management fees, taxes and exit fees but gross of entry fees. (7) All earnings in the form of dividends and capital gains are assumed to be reinvested to purchase new units in the fund. The investment performance, or realised return, of the fund is measured using quarterly earnings expressed as a percentage of the beginning of quarter funds under management.

The cash inflows reported by Plan For Life represent voluntary superannuation contributions made into the fund, rollovers from other superannuation funds or eligible termination payments. The cash outflows are customer-initiated withdrawals, either made at retirement, death, or transfers to other funds within the superannuation system.

A number of filters are applied to the data. Amalgamation of funds occurs in some quarters, generating dramatic changes in the level of funds under management as assets are withdrawn from one fund and added to another fund. It is impossible, with the current data, to disentangle the various effects of these changes and so funds with dramatic changes in funds under management, or the level of cash inflows, are excluded for the quarter in which the change occurs. Similarly, funds with zero funds under management (FUM) are also excluded from the data set. Return data is not available for some of the funds and a zero return appears in the data set in these cases. These zero return observations are deleted from the data set. Thus a quarterly observation is included in the study where: the funds under management at the beginning of the quarter are greater than zero; the growth in the funds managed over the quarter is within three standard deviations of the mean for that fund; inflows at the end of the quarter represent 50% or less of the funds managed at the beginning of the quarter (also known as normalised inflows); and investment return is provided (non-zero). (8)

Table 1 provides the descriptive statistics for the sample. While average returns are around 2.8% per quarter, the average quarterly inflows (8.5%) are somewhat larger than the outflows (7.9%). Average fund size is round $18 million though there is considerable variation in the size of the funds included in the analysis with fund size ranging from $90,000 to $575 million.

4. Analysis

If investors follow past performance there will be a positive relationship between past performance and current-quarter cash flows. Yet, there is mixed evidence as to the sensitivity of cash flows to poor past performance. US research suggests that investors focus on good performers but do not disinvest poor performers, however the Australian evidence dealing with wholesale funds (Sawicki 2000, 2001) does not support this asymmetry argument. We apply the Gruber (1996) approach to the separate analysis of cash inflows to the fund and cash outflows from the fund in section 4.1. While there is evidence of sector wide growth in the retail superannuation sector (section 4.2), panel-data analysis, reported in section 4.3, shows that there is a relationship existing between current net cash flows and past performance, past cash inflows and past cash outflows even after controlling for sector wide effects.

4.1 Current-Quarter Cash Flows and Past Performance

Table 2 depicts the average realised quarterly normalised net cash flows, inflows and outflows for deciles over a one-year holding period. Following Gruber (1996), funds are placed into ranked deciles using past one-year average raw return or past one year Sharpe index. Given the normalised net cash flows, reported in table 2, it is apparent that the past top performing funds experience greater net cash inflows than past poor performers and the positive relationship between past performance and cash flows holds when funds are ranked on either raw returns or their Sharpe index. This is confirmed by the positive and significant Spearman rank coefficient between performance ranking and normalised cash flows.

The statistical tests for variation across the decile based portfolio performance are not strong though the difference between the top decile and bottom decile is statistically significant at the 10% level for net cash flows, providing support for a positive correlation reported between past performance and net cash flows. These results are consistent with those of Gruber (1996), though not as statistically convincing. (9)

Table 2 also reports the break down of net cash flows into separate cash inflows and cash outflows. This provides further insight into the impact of past performance on cash flows with both past year raw returns and Sharpe measures correlated with cash inflows and cash outflows. Research into this question has generally focused on net cash flows (Gruber 1996; Sawicki 2000) and the separation of net cash flows into the separate components, cash inflows and cash outflows, is unusual in the literature.

Positive and significant correlation between past performance and average realised inflows indicates that past good performers attract greater inflows than past poor performers. The negative correlation between past performance and average outflows shows that funds with poor past performance have greater outflows relative to good performers. As expected the differences between deciles are positive when using inflows, however these results are not statistically significant. When using outflows, the difference between deciles is negative, but the only statistically significant difference between the top and bottom deciles is observed for the Sharpe index.

The finer definition of cash flows adds further weight to the argument that strongly performing funds attract cash inflows more than poorly performing funds and that poorly performing funds face greater cash outflows than strongly performing funds. While the symmetry in our results is consistent with Sawicki's finding for Australian wholesale superannuation funds they are not consistent with the USA findings of Ippolito (1992) and Sirri and Tufano (1998). For example the spearman rank correlations in our analysis between past performance, using raw returns, and cash inflows is 0.564 while the correlation between past performance and cash outflows is 0.552. These correlations suggest a fairly symmetric response to performance in terms of both cash flows into the fund and cash flows out of the fund. The symmetric response of Australian investors to performance at the retail level, as well as the wholesale level (Sawicki 2000), provides an important question for future research, particularly given the asymmetric response observed in the USA.

4.2 Current-Period and Prior-Period Cash Flows

There has been little research into the relationship between market wide cash inflows and cash outflows over time and we focus on this question in this section. Table 3 reports on the relationship between past cash flows and future cash flows using the Gruber (1996) approach. In all three cases, past inflows, past outflows and past net cash flows are statistically significantly correlated with future inflows, future outflows and future net cash flows, respectively. This is shown by the positive and statistically significant correlation between past cash flow ranked deciles and the average realised cash flows over the holding period. There are also significant differences between deciles subgroups for all but some of the outflow deciles. Thus funds with high cash inflow (outflow) in one period exhibit high cash inflow (outflow) in the following period.

4.3 Retail Sector Cash Flows

While the impact of previous period performance on cash flows (section 4.1) is discussed in the literature, the impact of prior-period cash flows (section 4.2) is not commonly noted in the literature. It is possible that cash flow persistence could be explained by other factors such as the general growth in the industry (see figure 1). If the general growth in the industry is important then it is expected that there will be a positive correlation between current inflows and past inflows. Further, if this were the only determinant, there would be no relationship between past cash outflows and either present cash outflows or present cash inflows as outflows would be essentially a random event.

[FIGURE 1 OMITTED]

If growth in funds under management is responsible for the positive correlation between cash inflows from one quarter to the next at the individual fund level then this should be evident in average cash flow data for the total sample. We use total sample average quarterly cash inflows and outflows to test for possible market wide movements in cash inflows and cash outflows. Vector auto regression (VAR) is applied to the sample average cash inflows and outflows to model the time series behaviour of these variables.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (1)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (2)

Where [SCFI.sub.t] refers to the sample average of cash inflows to total funds under management in the quarter t while [SCFO.sub.t] refers to the sample average of cash outflows to total funds under management in the quarter t. We use VAR lag exclusion Wald tests, as well as Akaike and Schwarz information criterion values in deciding upon the final choice of two lags for the VAR. The results from estimating this model are reported in table 4. (10)

The model includes time trend and constant term as well as the variables, SCFI = cash inflows as a percentage of beginning assets, SCFI(n) = cash inflows lagged n quarters, SCFO = cash outflows as a percentage of beginning assets, SCFO(n) = cash outflows lagged n quarters. ** (*) is statistically significant at the 5% (10%) level of significance.

Figure 1 shows that the cash under management in our sample has grown over the study period though there is also some evidence of a two-quarter effects in the data. The statistically significant positive relationship (see table 4) between current-quarter SCFI and one quarter lagged SCFI is consistent with continued growth over the period, though the time trend variables show that the rate of cash inflow as a percentage of funds under management is decreasing over the study period. This growth in the industry funds under management over the study period could help to explain the cash flow persistence noted in table 3.

[FIGURE 1 OMITTED]

The relationship between current cash flows and past cash outflows presents a more complex picture. There is a statistically significant negative relationship observed between past cash outflows (SCFO) and both current SCFI and current SCFO. Apparently, increases (decreases) in cash outflows in one quarter are likely to be followed by decreases (increases) in the level of cash inflows into the sector and decreases in the level of cash outflows out of the sector in following quarters. Further, the negatively signed SCFO parameters suggest the existence of mean reversion in outflows over the study period at a market wide level.

Thus general growth in the sector may provide an alternative explanation for the relationship observed in table 3 but it does not explain the more complex interrelationships that exist between cash out flows and cash inflows identified in this analysis. While we leave further analysis of this rather complex time series behaviour to future research it is important to reiterate the finding of positive autocorrelation in the average retail fund cash inflows at the total sample level.

4.4 Panel-Data Analysis of Net Cash Flows

In order to gain further insight into the determinants of net cash flows into the funds, panel-data analysis is conducted using quarterly data from September 1994 to June 2000. This extends on the Sawicki (2000) model with inclusion of both lagged cash inflows and outflows, at both the fund and the total sample level. We use lagged fund returns, included as a proxy for prior quarter performance. (11) We do not replicate the Sawicki (2000) analysis because there are only 66, out of the maximum of 398 funds, in the sample with observations available in each of the 28 available quarters. We apply panel-data analysis to all available observations rather that just focus on those funds with a full set of observations. Further, we extend the Sawicki (2000) model to include the impact of both the fund specific cash flow variables as well as total sample average cash flow variables as described in the section 4.3. The possibility that sector wide movements in cash flows could explain the observed cash flow persistence was raised in the previous section and this model provides an attempt to assess the importance of sector wide effects at an individual fund level. The net cash flow model used in this analysis is:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (3)

Where [NCF.sub.t] is the net cash flow for the fund as a percentage of the funds under management at time t, [CFI.sub.t-i] is the cash inflow for the fund as a percentage of the funds under management at time t-i, [CFO.sub.t-j] is the cash outflow for the fund as a percentage of the funds under management at time t-j, [RET.sub.t-k] is the fund return on assets invested (performance) as a percentage of the funds under management at time t-k, SIZE, is the natural log of the funds under management at the start of the quarter at time t. The results of this analysis are reported in table 5.

A random effects pooled regression model is used in analysis with CFI(n) = cash inflows as a percentage of beginning assets lagged n quarters, CFO(n) = cash outflows as a percentage of beginning assets lagged n quarters, RET(n) = percentage return earned by the fund lagged n quarters, SIZE is the natural log of assets under management at the end of the quarter, SCFI(n) = average cash inflows percentage of total funds for the sample lagged n quarters, SCFO(n) = average cash outflows percentage of total funds for the sample lagged n quarters. ** (*) is statistically significant at the 5% (10%) level of significance.

Lag choice can be problematic with these models and so we report a model chosen using the general to specific approach. The procedure starts with a regression that includes 6 lags for cash inflows, cash outflows and return variables. Statistically insignificant parameters are excluded on the basis of F-tests and t-tests until that stage where further exclusion has a statistically significant impact on the model (Johnston & Di Nardo 1997, p. 248-50).

At the individual fund level, cash inflows up to five quarters ago are positively related with net cash flows today, suggesting that cash inflows are quite persistence at the individual fund level. The existence of persistence in cash inflows is consistent with Bergstresser and Poterba (2002) and Sirri and Tuffano (1998) though we find that this persistence lasts for a considerable period of time in our Australian data. The two cash outflow parameters, for lags two and three, are also statistically significant and negative. Net cash flows observed today are negatively correlated with lagged cash outflows out to three lags with the larger and more statistically important parameter observed at the third quarter. Once cash outflows begin they tend to persist though there is not the same level of persistence that is observed for cash inflows with significant lags extending only to the third quarter. The arguments for symmetry (Sawicki 2000) seem reasonable with cash outflows following cash outflows and cash inflows following cash inflows.

These results provide an important insight into the behaviour of the investing public at the retail level because investors who invest in retail funds tend to follow investment advice provided by professional financial advisers. It has been suggested, in discussions with practitioners, that financial advisers use investment lists when advising their clients. These are lists of funds that are believed to be sound investments and the funds are generally placed on an investment list after some period of time over which the fund has performed reasonably well. It has been suggested that this list is rather 'sticky' and there is some reticence to drop funds once they are included on an investment list. Further, it may take some time before new funds are actually added to the list as it takes time for new funds to establish a track record. The observed persistence in cash flows may reflect this practise.

Prior quarter performance is also important in determining current-quarter net cash flows, with statistically significant positive fund return parameters for the one-quarter lag variable and for two quarters lag variable. Consistent with the literature and with previous results reported in table 2, prior quarter performance is positively related with current-quarter net cash flows for both lags two and three.

It should be noted that the results discussed above are estimated using a model that also includes the impact of sample wide movements in cash inflows and cash outflows at various lags. These variables are included to control for the possibility that market wide movements are driving the fund specific results reported in tables 2 and 3. Although there are statistically significant parameters at lags two and five for the sample wide average cash inflows, there are also four statistically significant parameters for the sample wide average cash outflows at lags, two, three, four and five. The lagged market wide cash outflow parameters are positive, suggesting that market wide cash outflows from two through five quarters back are positively related with net cash inflows today. Further, the market wide cash inflow parameter is positive at the two-quarter lag and negative at the five-quarter lag. It is apparent that, while sample wide cash flow effects have an impact at the individual fund level, they are not driving the results reported in tables 2 and 3.

4.5 Summary

Overall, we find further support for a positive relationship between past performance and cash flows. After separating net cash flows into inflows and outflows, there is also evidence of a positive relationship between past performance and inflows and a negative relationship between past performance and outflows though these results are not statistically strong. In addition, there is a significant positive correlation between past cash flows and subsequent cash flows, whether they are expressed as net cash flows, inflows or outflows. Vector autoregression over the sample wide average cash inflows and cash outflows identify a link between current-quarter and previous quarter fund cash flows.

We use panel-data analysis at the individual fund level to gain further insight into the relationship between cash flow over time and between past performance and current cash flows after adjust for market wide effects. This analysis provides further support for the importance of past fund performance, past fund cash inflows and outflows as well as sector wide cash movements in the determination of net cash inflows into retail funds in Australia.

5. Conclusions

Analysis of cash flows and performance shows that past performance is positively correlated with future net cash flows, consistent with the findings of Gruber (1996), Sirri and Tufano (1998) and Sawicki (2000, 2001), although, consistent with Sawicki, we do not observe asymmetry in the cash flow response to past performance that is noted in the USA studies.

Cash flow movements appear to persist over time, with cash inflows into (outflows out of) a fund in one quarter followed by cash inflows into (outflows out of) the fund in the following quarter. This result is evident both with simple correlations over average cash data as well as in more comprehensive panel-data analysis. There appears to be 'stickiness' in cash flows over time and this is not driven solely by market-wide factors.

The authors would like to thank Plan for Life Pty Ltd for providing the fund data used in the analysis.

(Date of receipt of final transcript: September 11, 2005. Accepted by Garry Twite, Area Editor.)

References

Australian Prudential Regulation Authority, 2001, Superannuation Trends December Quarter, Australian Prudential Regulation Authority, Canberra.

Bergstresser, D.B. & Poterba, J. 2002, 'Do after-tax returns affect mutual fund inflows?', Journal of Financial Economics, vol. 63, no. 3, pp. 381-414.

Bilson, C., Frino, A. & Heaney, R.A. 2004, 'Australian retail superannuation fund performance', Accounting Research Journal, vol. 17, pp. 43-56.

Bilson, C., Frino, A. & Heaney, R.A. 2005, 'Australian retail superannuation fund performance persistence', Accounting and Finance, vol. 45, no. 1, pp. 25-42.

Bird, R., Chin, H. & McCrae, M. 1983, 'The performance of Australian superannuation funds', Australian Journal of Management, vol. 8, pp. 49-69.

Brown, S.J., Goetzmann, W.N., Ibbotson, R.G. & Ross, S.A. 1992, 'Survivorship bias in performance studies, Review of Financial Studies, vol. 5, no. 4, pp. 553-80.

Carhart, M.M. 1997, On persistence in mutual fund performance, Journal of Finance, vol. 52, no. 1, pp. 57-82.

Drew, M.E. & Stanford, J.D. 2001, 'Asset selection and superannuation fund performance: A note to trustees', Economic Papers, vol. 20, no. 1, pp. 57-55.

Drew, M.E. & Stanford, J.D. 2003, 'Returns from investing in Australian equity superannuation funds, 1991-1999', Services Industries Journal, vol. 23, no. 3, pp. 12-24.

Drew, M.E., Stanford, J.D. & Veeraraghavan, M. 2001, 'Selecting Australian equity superannuation funds: A retail investor's perspective', Working paper, Queensland University of Technology, Australia.

Droms, W.G. & Walker, D.A. 2001, 'Persistence of mutual fund operating characteristics: Returns, turnover rates, and expense ratios', Applied Financial Economics, vol. 11, pp. 457-56.

Elton, E.J., Gruber, M.J., Das, S. & Hlavka, M. 1993, 'Efficiency with costly information: A reinterpretation of evidence from managed portfolio' The Review of Financial Studies, vol. 6, no. 1, pp. 1-22.

Goetzmann, W.N. & Ibbotson, R.G. 1994, 'Do winners repeat? Patterns in mutual fund behaviour', Journal of Portfolio Management, vol. 20, no. 2, pp. 9-18.

Grinblatt, M. & Titman, S. 1992, 'The persistence of mutual fund performance', Journal of Finance, vol. 47, no. 5, pp. 1977-84.

Gruber, M.J. 1996, 'Another puzzle: The growth in actively managed mutual funds', Journal of Finance, vol. 51, no. 3, pp. 783-810.

Hallahan, T.A. 1999, 'The information content of portfolio performance history and persistence in fund performance: An examination of rollover funds' Accounting and Finance, vol. 39, no. 3, pp. 255-74.

Hendricks, D., Patel, J. & Zeckhauser, R. 1993, 'Hot hands in mutual funds: Short-run persistence of relative performance, 1974-1988', Journal of Finance, vol. 48, no. 1, pp. 93-130.

Ippolito, R.A. 1992, 'Consumer reaction to measures of poor quality: Evidence from the mutual fund industry', Journal of Law & Economics, vol. 35, pp. 45-70.

Jensen, M.C. 1968, 'The performance of mutual funds in the period 1945 1964' Journal of Finance, vol. 23, no. 2, pp. 389-416.

Johnston, J. & Di Nardo, J. 1997, Econometric Methods, 4'h Edition, McGraw Hill Companies Inc, New York.

Malkiel, B.G. 1995, 'Returns from investing in equity mutual funds 1971 1991', Journal of Finance, vol. 50, no. 2, pp. 549-72.

MLC Limited, 2002, MLC MasterKey Superannuation Customer Information for Individuals and Small Business, MLC.

Robson, G.N. 1986, 'The investment performance of unit trusts and mutual funds in Australia for the period 1969 to 1978' Accounting and Finance, vol. 26, no. 2, pp. 55-79.

Sauer, D.A. 1997, 'Information content of prior period mutual fund performance rankings', Journal of Economics and Business, vol. 49, pp. 549-67.

Sawicki, J. 2000, 'Investors' response to the performance of professional fund managers evidence from the Australian wholesale funds market', Australian Journal of Management, vol. 25, no. 1, pp. 47-66.

Sawicki, J. 2001, 'Investors' differential response to managed fund performance', The Journal of Financial Research, vol. 24, no. 3, pp. 367-84.

Sharpe, W.F. 1966, 'Mutual fund performance', Journal of Business, vol. 39, no. 1, pp. 119-38.

Sirri, E.R. & Tufano, P. 1998, 'Costly search and mutual funds flows', Journal of Finance, vol. 53, no. 5, pp. 1589-622.

Treynor, J.L. 1965, 'How to rate management of investment funds', Harvard Business Review, vol. 43, pp. 63-75.

White, A. 2002, 'Business survey series 2--Managed funds', The Australian, March 7, pp. 1-13.

(1.) Carhart (1997), Elton, Gruber, Das and Hlavka (1993), Gruber (1996), Jensen (1968), Malkiel (1995), Sharpe (1966) and Treynor (1965).

(2.) Bird, Chin and McCrae (1983), Robson (1986) and Drew and Stanford (2001a, 2003).

(3.) Hendricks, Patel and Zeckhauser (1993), Goetzmann and Ibbotson (1994), Gruber (1996) and Droms and Walker (2001).

(4.) Carhart (1997), Sauer (1997) and Brown, Goetzmann, Ibbotson and Ross (1992).

(5.) Where net cash flows are defined as flows not attributable to capital gains or dividends

(6.) This is a firm of actuaries and researchers who monitor the retail managed fund industry and provide statistical information and analysis.

(7.) The funds are all complying superannuation funds under the s42.5 of the Superannuation (Industry) Supervision Act 1993 (Cth) therefore the reported earnings are net of taxes.

(8.) Application of the various filters to the initial data set creates gaps in the time series for some funds. Deleting observations is preferable to deleting entire funds, which would result in a sample that is biased towards surviving funds. The approach followed here ensures that, for any quarter, the maximum number of funds is represented.

(9.) Gruber (1996) found that past one and three year raw returns and past one and three year four factor alphas are good predictors of net cash flows with similar correlations to those reported in this paper.

(10.) There is no evidence of serial correlation in the residuals and the null of jointly normally distributed residuals cannot be rejected at the 5% level of significance.

(11.) Sawicki (2000) shows that the actual choice of performance measure has little impact on analysis.

Angela Frino ([dagger]) Richard Heaney ([section]) David Service ([dagger])

([dagger]) Faculty of Economics and Commerce, School of Finance and Applied Statistics, ANU, ACT, 0200.

([section]) School of Economics and Finance, Business, RMIT University, Level 12, 239 Bourke Street, Melbourne, VIC, 3000. Email: [e-mail]

Table 1
Descriptive Statistics

This table provides descriptive statistics for the sample for the
period, September 1994 to June 2000. The indicated variable is first
averaged across all quarterly observations for a particular fund and
statistics are presented on these mean quarterly values. The sample
includes all managed growth and managed stable funds. Return is net
of expenses, taxes, exit fees but gross of entry fees. All cash flows
are normalised by dividing the dollar cash flow by the total funds
under management at the beginning of the quarter. FUM is the assets
managed at the beginning of the quarter.

                         Mean    Median   Std. Dev.     Max.    Min.

Full Sample (N = 398)
  Return                 2.79%    2.48%      1.64%     14.64%   -3.02%
  Inflows                8.47%    6.19%      7.07%     39.34%    0.00%
  Outflows               7.86%    6.29%      7.07%     90.20%    0.00%
  FUM ($million)        18.66     5.04      47.84     574.82     0.09

Table 2
Average Quarterly Cash Flows for Portfolios of Funds Formed on
Past Average Raw Return or Sharpe Index

The table shows the average realised quarterly net cash flows,
inflows and outflows for decile portfolios formed on the basis
of past year average raw returns and past year Sharpe index.
Funds are sorted at the end of the June quarter from 1994 to
1999 into equally weighted decile portfolios based upon their
average raw return or Sharpe index from the past year. Each
decile portfolio was held for 1 year and the average cash
flows measured. The average cash flow of the six time series
observations is reported below. The SI is calculated as the
average excess return ([R.sub.it]-[R.sub.ft]) per unit of total
risk ([sigma]), where [R.sub.it] is the return on portfolio i
in quarter t and [R.sub.ft] the return on the 90-day Australian
Treasury Note in quarter t. All cash flows are normalised by
dividing the dollar cash flow by the total funds under management
at the beginning of the quarter. Funds with the highest average
raw return and Sharpe index comprise decile 10 and funds with the
lowest comprise decile 1. Funds are required to survive over the
entire two year ranking and holding period. The Spearman Rank
Coefficient provides a test of significance of the correlation
across the deciles in the ranking and holding periods. The table
also shows the difference between deciles and their statistical
significance is determined using a one-sample t-test. Fund returns
are measured net of expenses and taxes, cash inflows are gross of
entry fees and cash outflows are net of exit fees.

Past Year Ranking         Normalised Net Cash Flows

                           Raw
Decile                    Return        SI

Worst 1                   -0.032      -0.032
2                         -0.021      -0.030
3                         -0.013      -0.014
4                         -0.001       0.003
5                         -0.005       0.001
6                         -0.019      -0.013
7                         -0.004      -0.007
8                         -0.011      -0.007
9                          0.010      -0.005
Best 10                    0.013       0.018
Spearman Rank              0.782 **    0.685  **
Correlation Coefficient
Top decile-bottom          0.046 *     0.051  *
  decile
Top decile-average of      0.024       0.030
  bottom 9 deciles
Top 5 deciles-bottom       0.012       0.012
  5 deciles

Past Year Ranking         Normalised Inflows

                            Raw
Decile                    Return       SI

Worst 1                    0.043     0.046
2                          0.054     0.041
3                          0.049     0.043
4                          0.058     0.067
5                          0.047     0.051
6                          0.033     0.044
7                          0.045     0.049
8                          0.058     0.056
9                          0.066     0.053
Best 10                    0.070     0.072
Spearman Rank              0.564 *   0.673 **
Correlation Coefficient
Top decile-bottom          0.027     0.026
  decile
Top decile-average of      0.020     0.022
  bottom 9 deciles
Top 5 deciles-bottom       0.004     0.005
  5 deciles

Past Year Ranking         Normalised Outflows

                            Raw
Decile                    Return        SI

Worst 1                    0.076      0.078
2                          0.074      0.071
3                          0.062      0.058
4                          0.059      0.063
5                          0.053      0.051
6                          0.052      0.057
7                          0.049      0.055
8                          0.069      0.064
9                          0.056      0.058
Best 10                    0.057      0.054
Spearman Rank             -0.552 *   -0.527
Correlation Coefficient
Top decile-bottom         -0.019     -0.024 **
  decile
Top decile-average of     -0.004     -0.008
  bottom 9 deciles
Top 5 deciles-bottom      -0.008     -0.007
  5 deciles

Note: ** (*) is statistically significant at the 5% (10%)
level of significance.

Table 3
Average Quarterly Cash Flows for Portfolios of Funds Formed on
Past Average Cash Flows

The table shows the average realised quarterly net cash flows,
inflows and outflows for decile portfolios formed on the basis
of past year average net cash flows, inflows and outflows,
respectively. Funds are sorted at the end of the June quarter
from 1994 to 1999 into equally weighted decile portfolios based
upon their average cash flow from the past year. Each decile
portfolio was held for 1 year and the average cash flows measured.
The average cash flow of the six time series observations is
reported below. All cash flows are normalised by dividing the
dollar cash flow by the total funds under management at the
beginning of the quarter. Funds with the largest average past
net cash flow, inflow or outflow comprise decile 10 and funds
with the smallest comprise decile 1. Funds are required to
survive over the entire two year ranking and holding period. The
Spearman Rank Coefficient provides a test of significance of the
correlation across the deciles in the ranking and holding periods.
The table also shows the difference between deciles and their
statistical significance is determined using a one-sample t-test.

                            Normalised
                             Net Cash    Normalised   Normalised
                              Flows        Inflows      Outflows

Smallest 1                  -0.050         0.021        0.048
2                           -0.046         0.020        0.048
3                           -0.037         0.021        0.048
4                           -0.027         0.033        0.057
5                           -0.019         0.027        0.052
6                           -0.017         0.041        0.061
7                           -0.009         0.048        0.065
8                            0.010         0.078        0.075
9                            0.044         0.102        0.072
Largest 10                   0.067         0.128        0.078
Spearman Rank Correlation    1.00  **      0.976 **     0.927  **
  Coefficient
Top decile--bottom decile    0.117 **      0.107 **     0.030
Top decile--average of       0.083 **      0.084 **     0.020
  bottom 9 deciles
Top 5 deciles--bottom 5      0.055 **      0.055 **     0.020  **
  deciles

Note: ** (*) is statistically significant at the 5% (10%0 level
of significance.

Table 4
Vector Auto Regression Using Sample Average Cash Inflow and
Cash Outflow Variables

               SCFI      t-statistic     SCFO       t-statistic

SCF1(1)      0.6337 **       3.57      -0.0617         -0.46
SCFI(2)     -0.1954         -1.06       0.0045          0.03
SCFO(1)     -0.4918 *       -1.76      -0.0542         -0.26
SCFO(2)     -0.8002 **      -2.89      -0.4268 **      -2.04
CONSTANT     0.1426 **       4.42       0.1102 **       4.54
TIMETREND   -0.0009 **      -2.16      -0.0001         -0.47
R-squared    0.70                       0.22
N           28                         28

Table 5
Panel Analysis of the Relation Between Current-Quarter Net Cash
Flows and Prior Quarter Cash Inflows, Cash Outflows and Returns

Variable            Coefficient   t-Statistic

Constant                -0.3752      -7.38 **
CFI(1)                   0.1023       3.66 **
CFI(2)                   0.1788       6.39 **
CFI(3)                   0.0650        2.5 **
CFI(4)                   0.1021       4.39 **
CFI(5)                   0.0794       3.87 **
CFO(2)                  -0.0549      -1.93 *
CFO(3)                  -0.0943      -3.65 **
RET(1)                   0.1780       3.48 **
RET(2)                   0.0958       1.86 *
SIZE                     0.0154       10.1 **
SCFI(2)                  0.4181       3.62 **
SCFI(5)                 -0.2645      -2.66 **
SCFO(2)                  0.4763       1.96 **
SCFO(3)                  1.0343       4.52 **
SCFO(4)                  1.2711       5.07 **
SCFO(5)                  1.2797       5.87 **
Adjusted R square        0.3284
N                      3279