The Treasury

Global Navigation

Personal tools

Treasury
Publication

Review of the KiwiSaver Fund Manager Market Dynamics and Allocation of Assets

Appendix 3: Regression analysis of fees

As a further empirical test of the efficiency of the KiwiSaver fund management market, we analysed the determinants of fund fees. In a perfectly competitive market, one would expect that fees would be as low as possible to justify fund providers' participation in the market. However, perfect competition may not be an appropriate benchmark in this context given that providers may be, to some extent, differentiable based on the returns they generate.

In this sense, we might expect to observe the following relationships in an efficient KiwiSaver market:

  • Higher performing funds having higher fees (all else constant)
  • Larger funds having lower fees due to economies of scale, but with this effect diminishing as funds grow (all else constant)
  • Other fund characteristics, such as being a bank or a default provider are not positively related to fees (as this may suggest market power)

Methodology

We run similar regressions to fund flow regressions (see Appendix 2), except that we use TER as the dependent variable.

In line with other studies of fund fees, we use the natural logarithm of AUM and its square to test for a negative but convex relationship between fund size and fees - as would be expected under the theory of economies of scale. In an efficient market, we would expect to see TER follow a fund's long-run average cost curve due to competitive pressures around pricing. We also use provider AUM as an explanatory variable to account or the fact that a good part of a fund's costs will accrue at the provider level.

We use both fund returns and aggregate returns to account for the impact of returns of fees. The rationale for doing so is that we are mainly interested in the cross-sectional variation in returns affecting the fees funds can charge. Variations in market returns over time must therefore be controlled for help isolate this cross-sectional effect. There is a risk of reverse-causality if fees are contractually dependent on fund performance, though fortunately this is rare in practice. Reverse causality could also be a problem given we are using post-fee returns. However, this would suggest a negative relationship between fees and returns. Given that the variability in returns over time is significantly larger than the variability in fees (which tend to change relatively infrequently), this effect should not affect the results too much.

To test the potential impact of default status, default affiliation and bank status on pricing, we use these dummy variables as per the fund flow regressions. The key control variables are fund type dummies, given that fees vary fairly systematically across different types of funds depending on how active their management is. We also control for provider age.

Finally, we use a 2014 dummy and its interaction with fund returns to test the impact of the Fund Finder comparison tool on fee levels and the sensitivity of fees to returns respectively.

Results

Table 2 below presents the results from our regressions of fund fees against lagged explanatory variable across our three key specifications.

Table 2: Results from TER regressions
1 2 3
Intercept 1.31 (0.06) 1.28 (0.07) 1.32 (0.07)
Ln AUM ($m) 0.01 (0.02) -0.04 (0.02) 0.01 (0.02)
Ln AUM ^2 0.00 (0.00) 0.01 (0.00) 0.00 (0.00)
Ln Provider AUM ($m) -0.05 (0.01) 0.00 (0.01) -0.05 (0.01)
1 Yr Return (%) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00)
Aggregate KiwiSaver return (%) -0.01 (0.00) -0.02 (0.00) -0.01 (0.00)
Provider Age (Years) -0.01 (0.01) -0.03 (0.01) -0.01 (0.01)
Cash -0.25 (0.03) -0.28 (0.03) -0.25 (0.03)
Balanced 0.07 (0.02) 0.01 (0.02) 0.07 (0.02)
Growth 0.28 (0.03) 0.18 (0.03) 0.28 (0.03)
Other 0.22 (0.03) 0.21 (0.03) 0.22 (0.03)
Bank     -0.17 (0.02)    
Default     -0.49 (0.05)    
Default affiliated     -0.08 (0.02)    
2014         0.00 (0.00)
Return x 2014         -0.03 (0.04)
Adjusted R2 0.36 0.47   0.36

NB: Estimated coefficients are reported on the left with standard errors in brackets on the right. Coefficients significant at the10% level are highlighted.

Importantly, we do find slight evidence of economies of scale at the individual fund level. When controlling for default status, fees tend to fall marginally as fund size increases. Also, as suggested by economic theory, the observed rate of decline in fees does tend to diminish as a fund gets larger. In addition, there is some evidence of a negative relationship between provider size and TER, though this probably reflects the effect of lower-fee default providers as opposed to economies of scale.

Interestingly we find a statistically significant positive relationship between 1 year returns but a statistically significant negative relationship between market returns and fees. This suggests that fees do help drive fund performance, even when controlling for fund type, though reverse causality may also contribute to this result. That said, it does appear that relative performance not absolute performance is what drives fees in our sample.

Moreover, we find statically significant evidence that fees tend to fall over a fund's lifetime, at a rate of around 3bps per annum when controlling for other factors.

As expected, the fund type has a strong statistical impact of fees, with ‘cash' funds being consistently cheaper than conservative, whilst 'balanced', 'growth' and 'other' funds being more expensive.

We also see that default funds have significantly lower fees when controlling for other factors, which is expected as fees were a large component of the initial default provider appointment tender process. This observation also holds, albeit not as strongly, for those other funds offered by default providers.

Our evidence suggests that banks offer more competitive fees when holding other factors constant, with a statistically significant relationship between bank status and fees.

Finally, our results found no statistically significant effects of the introduction of Fund Finder comparison tool in late 2013 on fees, or on the sensitivity of fees to fund performance.

Page top