PTWhat moves the value of a multimarket fund?
Multimarket funds can invest in interest rates, currencies, stocks and commodities at the same time. The research followed Brazil's 50 largest, day by day, from 2017 to mid-2023, measured how each one performed and looked for the economic indicators that move together with their units. Interest rates weighed the most. Explained in 12 screens.
A fund that can invest in almost anything
An investment fund pools the money of many people, the unitholders, and hands it to a professional manager. Investors buy units; the value of a unit rises or falls with what the manager did.
Sections 2.1 and 3.1. ANBIMA, Brazil's association of financial market entities, divides multimarket funds into subtypes by strategy (Table 1). On the right, how many of the 50 funds studied fall in each.
The 50 largest held about R$ 82 billion
The author chose the 50 open-ended multimarket funds with the largest net assets on 2 January 2017 and followed each one until 30 June 2023: before, during and after the pandemic. Each rectangle is a fund, sized by its net assets. Tap one.
Daily data from the ComDinheiro platform. The four closed funds (section 3.2) are left out of the year-by-year charts on the next screens, which use the 46 with complete data.
Earning more is not enough: you have to see how much risk was taken
The dissertation uses three classic ratios. All three divide the same gain above the CDI (Brazil's benchmark interbank rate) by a different measure of risk: the higher the ratio, the better the trade-off between gain and risk.
Made-up fund and numbers, just to show the calculation. In the dissertation's tables, the ratios come from ComDinheiro and the Sortino is on a daily basis, which is why its values are much smaller.
Over the whole period, 32 of the 46 funds beat the CDI
The CDI is the Brazilian investor's yardstick: what a conservative fixed-income investment earned. Over the whole period it added up to about 59%. Pick a year and see how the picture changes.
Our counts from Table 3 (the fund's return minus the CDI's, in percentage points). Accumulated CDI from Jan 2017 to Jun 2023 calculated by us as the fund's return (Table 7) minus the excess (Table 3): it comes to 59.1% for every fund. 2023 runs only to June.
Swinging more did not pay more
Each dot is a fund: how much its units swung per year (to the right, more risk) and how much they returned (up, more gain) over the 12, 24 or 36 months to June 2023. If risk were always rewarded, the dots would rise from left to right.
Correlation between swings and return our calculation: −0.35 over 12 months, −0.60 over 24 and −0.08 over 36. Accumulated return in the window; swings = annualised volatility.
The same fund can be 1st or 46th, depending on the yardstick
The dissertation calculates each ratio over four windows: 12, 24 and 36 months to June 2023 and the whole period since 2017. Pick a fund: the taller the column, the better its position among the 46.
Positions counted by us among the 46 funds with complete data (1st = highest value). In the Sortino, equal values share a position.
For funds that barely follow the stock market, the Treynor ratio goes haywire
In the 12- to 36-month windows, the funds' Treynor ranged from −2,991 to +6,482, while the Sharpe stayed between −3.9 and +8.9. The same fund that leads on the Sharpe comes last on the 12-month Treynor.
Our reading The dissertation does not publish the betas. The explanation comes from the formula (section 2.5): the Treynor divides by the beta, and the dissertation itself notes that relying on the beta is a limitation of the ratio. When the beta is close to zero, any gain becomes a huge number; if the beta is negative, the sign flips.
76,000 fund-days against 12 economic indicators
For each fund and each day, the model puts the unit value side by side with twelve indicators. Everything enters in logarithms, so the result reads in percentages: when the indicator is 1% higher, the units tend to be x% higher or lower.
Panel data regression with fund and year fixed effects and Driscoll-Kraay standard errors, which correct the autocorrelation flagged by the Wooldridge test; the Hausman test pointed to fixed effects. 76,434 observations (Table 12). Accumulated Selic: Brazil's policy rate added up day by day, an index that only grows.
Units moved with interest rates, almost one for one
In the accumulated Selic, 1% more came with about 1.04% more on the units, the model's biggest effect. The US stock market and the dollar moved the other way, with small effects. The Ibovespa, Brazil's main stock index, had no clear effect on its own.
Within-fund R². Our note Units and the accumulated Selic grow together over time, which helps explain such a high R²; the model shows association, not cause.
Funds with the best risk-return trade-off tracked the Selic more closely
The author split the funds into three groups by ratio and ran the model for the top third and the bottom third. At the top, the link with the Selic was about twice that at the bottom, and the Ibovespa weighed more in the opposite direction. Tap an indicator.
When choosing a fund, look at interest rates and at more than one ratio
For investors
Compare funds by risk-adjusted return, over more than one time window, and not only by how much they earned. The author reminds readers that past performance does not guarantee future results. This page is not investment advice.
For fund managers
Monitoring the indicators that weighed in the model, especially interest rates, inflation, commodities and foreign markets, helps anticipate how units react to changes in the economy.
For researchers
Include management and performance fees, each fund's liquidity and risk, test non-linear effects and use a longer horizon or high-frequency data.
What this research does not say
- Association, not cause. The model compares levels of units and of indices that grow over time; series like these tend to look linked. · our note
- Only the largest, and the survivors. 50 large funds from 2017; four were closed, and the results may not hold for other multimarket funds. · according to the author, and our note
- No costs. Management and performance fees were not in the model. · according to the author
- Numbers to double-check. In Table 12, some standard errors do not match the t-values and two coefficients have a misplaced decimal point; for gold we show only the direction. The bottom groups by Treynor repeat those by Sharpe and were left out. · our note
- Time changes things. Crises, rule changes and economic cycles can alter the relationships. · according to the author
Credits
BARBOSA, Harion Camargo. Determinantes de desempenho dos fundos multimercados no Brasil entre 2017 e 2023: uma análise de performance dos 50 maiores fundos do mercado brasileiro [Performance determinants of multimarket funds in Brazil between 2017 and 2023: a performance analysis of the 50 largest funds in the Brazilian market]. Master's dissertation (Business Administration) – Mackenzie Presbyterian University, São Paulo, 2024. Advisor: Prof. Dr. Denis Forte.
Research supported by CAPES, Brazil's federal agency for graduate education. Data from ComDinheiro; panel regression in Stata. This version simplifies the work for a general audience; the full tables are in the original text.