Four calendar years identify the only result here that a fund-selection desk might trade. Inflation itself is unavailable to buy.

Wulandari, Suwarno and Triharyati begin with a puzzle in the Indonesian fund industry: investor numbers rose while industry NAV shrank. NAV declined from Rp578.44 trillion at the end of 2021 to Rp496.84 trillion at the end of 2024. The largest part of the decline came in 2022, which alone erased about 12.58%. The authors ask what drives fund performance under those conditions.

Their proposed mechanism follows the textbook. Inflation reduces real returns, while a higher policy rate draws money into low-risk fixed income. Both should damage fund performance.

They test this by pooling fund-year Sharpe ratios over 2021-2024 and estimating OLS in SPSS with five candidate regressors:

The funds are purposively sampled from a stated population of 107 conventional products. Across fund-years, the Sharpe averages -0.072 and ranges from -0.6998 to 1.0180.

Inflation and the BI Rate reach significance. Inflation's coefficient is B = -0.046 (t = -2.819, Sig. 0.006); for the BI Rate, B = +0.061 (t = 2.365, Sig. 0.020). The p-values for size, age and allocation all exceed 0.80. A joint F of 2.382 has Sig. 0.046. These results lead the authors to conclude that macroeconomic factors proved more important than fund characteristics during the period.

The BI Rate sign clashes with the cited research

The authors' literature review points in the opposite direction from their BI Rate estimate. Their background sections cite Mishkin on higher rates depressing stock and bond prices. They also cite Rahmawati et al. (2022) for a negative BI Rate effect and Saputri and Yudiantoro (2022) for negative rate and inflation effects on Indonesian funds. Yet the reported coefficient is positive 0.061, and the paper describes the rate as making a real contribution to performance. We did not find an explanation reconciling that sign with the cited work.

Timing creates another problem. The inflation mean, 2.8900, exactly matches the average of the four annual prints cited in the paper: 1.87%, 5.51%, 2.61% and 1.57%. Each fund-year Sharpe is therefore paired with inflation from the same year. The regression explains the year retrospectively. An investor trying to trade the result would have needed the 5.51% print for 2022 before 2022 began. The paper does invite that use: it advises investors to monitor the BI Rate and inflation ("BI Rate, inflasi") when selecting funds.

How many observations identify the macro slopes?

Four.

Within a year, every fund receives an identical BI Rate and inflation figure. The descriptive table's 92 observations correspond to 23 funds times 4 years. The macro coefficients consequently draw on four year-level points repeated across funds, with 92 rows in the descriptives and 84 in the regression's ANOVA.

Of those four points, the model uses three parameters for the two macro slopes and the constant. Only one degree of freedom remains for year-level variation. Other differences between 2022 and 2024 are absorbed by those variables, including that year's equity and bond market performance.

The pooled OLS t-stats treat all 84 regression rows (total df 83) as independent draws. We did not find fund or year fixed effects or clustered standard errors in the paper. Year clustering would offer little help: four clusters are too few to trust the correction. The Durbin-Watson statistic of 2.098 tests serial ordering in a stacked file. It cannot tell us about 23 funds exposed to the same macro shock.

Samples that refuse to line up

The paper gives conflicting counts. Its Methods section names 23 funds; the abstract names 32. Descriptives use N = 92, whereas the normality test uses N = 120. The ANOVA's total df 83 implies 84 observations (residual df 78, five regressors).

The fit measures diverge as well. R Square is 0.132 in the text, consistent with the ANOVA calculation of 0.602 / 4.548. In Table 4, however, R Square is 0.287 and adjusted 0.195. Those figures must belong to a different run.

The residuals used for the normality test average -0.0361. With an intercept, OLS residuals average exactly zero over the estimation sample. The tested residuals therefore come from something other than the reported fit on the reported sample.

Asset allocation is harder still to follow. According to the abstract, it "could not be further analyzed because it was constant and excluded from the regression model." Yet Table 5 supplies a coefficient of 0.000, t 0.156, Sig. 0.876. The descriptive statistics show a range of 0 to 99 and an SD of 30.360. Allocation is absent from the VIF table, while the discussion uses that same t-test to reject its hypothesis. An SD of 30.36 is incompatible with a constant variable.

Even the citations are unstable. The paper cites Rahmawati et al. (2022) for BI Rate effects on mutual funds, although the listed title concerns the composite stock index.

An inflation sign worth keeping, with limits

Inflation's sign matches the prior work cited by the authors. Size and age also produce tight nulls under the paper's own unclustered standard errors: 0.014 for log AUM and 0.005 for age. Clustering by fund would widen them.

The abstract takes more care than the conclusion. It describes macro and fund variables as having "showed varying effects," then closes by giving macroeconomic conditions and internal fund characteristics equal weight: both "play an important role." The authors also request a longer period and a larger, more varied sample. They say a longer period would capture economic change over the long run. That reason goes directly to the four-year objection, yet the conclusion still places macro factors above fund characteristics.

The ranking rests on two slopes estimated over four years, with an R Square reported as either 0.132 or 0.287. Its joint F passes 0.05 with Sig. 0.046. Those results do not support committing capital.

We could not test the finding ourselves. We lack histories of Indonesian fund NAVs and AUM, along with BI Rate and Indonesian CPI series. Substituting US ETFs and US macro data would test another economy.

A credible check would require monthly NAVs for every conventional fund registered during 2021-2024, including closed funds, together with fund-type labels. Coverage would have to extend beyond the 107 funds with complete data. The question would be whether inflation retains its sign when standard errors are clustered across 48 months rather than four years. If the sign survived that test, we would take the inflation finding seriously.