The Treasury

Global Navigation

Personal tools


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

Appendix 2: Regression analysis of fund flow

The aim of this analysis is to help uncover what factors determine the ‘success' of KiwiSaver funds, measured by their ability to increase membership and therefore the volume of savings the fund has under management on which to receive fees. One possible measure of success would be member growth, but unfortunately we have not been able to access suitable data on this measure. Instead, we measure success using funds' net inflows of funds from members in a given period, also known as “fund flow”.

In an efficient market, one would expect the flow of funds to be determined by members seeking to maximise their after-fee returns for a given risk-preference. This would place competitive pressures on providers to minimise fees, maximise returns, and would drive high-cost, poor-performing funds out of the market. In this regard, we would expect to observe the following factors:

  • Funds with lower fees experience higher fund flow for (all else constant)
  • Funds with higher returns experience higher fund flow for (all else constant)
  • Other fund characteristics such as size or being a default provider or a bank do not case funds to experience higher fund flow (all else constant)


We run standard OLS regressions using quarterly fund flow as a percentage of start-of-quarter assets under management (AUM) as the dependent variable. Fund flow is considered in percentage terms as opposed to dollar terms to account for the strong relationship between size and fund flow resulting from the inflows of funds from existing members.[79] Flows from existing members tend to be much larger than flows from members shifting funds. However, customer inertia would suggest that the decision to continue contributing to your existing fund is not made actively. By contrast, shift of savings between funds and initial entry into a fund does entail an active decision (even if that decision is to stay with a default fund.

One shortfall of using percentage fund flow is that, in the context of growing AUM, percentage fund flow will tend to decline over time as relatively constant contributions represent a falling proportion of existing AUM. However, we address this shortfall by controlling for provider age in years (unfortunately data on individual fund age was not available).

In terms of key explanatory variables are as follows:

  • Total expense ratio (TER), which measures total fees as a percentage of AUM
  • Past returns (after fees) from the 12 months
  • Natural logarithm of AUM in millions of dollars
  • Dummy variables for whether a fund is operated by a bank, whether it is a default fund, and whether it is a fund offered by a provider that also offers default funds[80]

To control for the potential variability in contributions over time in response to rates of return, we control for aggregate KiwiSaver market returns. This is calculated as the AUM-weighted average of fund returns for the 12 months before the start of the quarter. We also control for potential variability across fund 'types' using dummy variables for ‘cash' funds, ‘balanced' funds, ‘growth' funds, and ‘other' funds, using ‘conservative' as a reference. A September quarter dummy variable is also included to control for the spike in fund flow that tends to occur annually, presumably due to the self-employed making contributions after filing their year-end tax returns.

Finally, we use a 2014 dummy to assess the impact of the introduction of the Fund Finder online comparison tool in November 2013. This dummy is interacted with TER and returns to give an indication of the effect of Fund Finder on the relationship between these variables and fund flow.


We use quarterly data over the 15 quarters between Q1 2011 to Q2 2014. Data on fund flow, fees, fund performance, and fund type (ie, conservative, growth, etc) is from FundSource. This is merged with data on AUM from Morningstar. Default status variable, bank dummy, and provider age are added manually using information from the KiwiSaver website. The dataset is trimmed to exclude observations with missing data points,[81] and observations for fund-quarters where funds have been merged (given that fund flow is abnormally high or low for these observations).

Our final dataset comprises 1226 observations. It is based on a sample of 134 individual KiwiSaver funds across 17 providers. This includes all major fund providers, with Kiwibank, KiwiWealth (previously Gareth Morgan Investments) and Craigs Investment Partners the notable exceptions.


  • [79]AUM explains 64% of the variation in dollar-value fund flow in a bivariate regression.
  • [80]Note that in some cases, entities that operate default funds offer KiwiSaver funds under multiple brands. For example, in the dataset ASB (a default provider) also offers KiwiSaver products under both its ASB and Firstchoice brands. In this case all the funds offered by both ASB and Firstchoice would be considered as “default affiliated”.
  • [81]In practice, we excluded observations where either fees or fund flows were zero.
Page top