Automated analysis

This analysis was drafted by our research engine and has not been checked by a human editor. It may contain errors. It separates the paper’s own results from our tests, and any figures called ours come from our own backtest.

Our automated analysis

ESG per point: 0.124% in text, 1.7 points in Table 2

Across 26 Indonesian miners in 2019-2023, ESG is significant at p = 0.003; text and Table 2 disagree on size.

2026-10-22 · 5 min read · ESG factor investing · Indonesian mining equities

Reviewing: The Impact of Environmental, Social, and Governance (ESG) Disclosure and Mining Commodity Price Volatility on Stock Returns for Mining Sector Companies on the Indonesia Stock Exchange (2019–2023) · Nadila Dwi Ananda, Acep Komara and Mada Purwanto Wahyu Nugroho · Read it on openalex

The paper gives its ESG result a p-value without settling its size, the part a trader needs to use it.

What went into the regression

Ananda, Komara and Nugroho begin with signaling theory: disclosure tells investors something about a firm's risk and management. Mining faces heavy environmental scrutiny, so firms that reveal more about their environmental, social and governance practices should receive a lower perceived risk or a re-rating. The other variable comes from outside the firm. Mining revenues follow gold, coal and nickel prices, making instability in those markets a plausible driver of miners' stock returns.

The authors run a pooled OLS regression in SPSS 26. They regress annual stock return, calculated as price change plus dividends divided by the beginning-of-year price, on two variables. One is the Refinitiv composite ESG score, on a scale from 0 to 100. The other is commodity price volatility, measured as the standard deviation of monthly gold, coal and nickel prices.

The sample covers 26 miners on the Indonesia Stock Exchange (IDX) from 2019 to 2023, giving 130 firm-years. Inclusion requires continuous listing, complete Refinitiv coverage and no suspension lasting more than three months. As written, the equation includes neither controls nor firm or year effects.

ESG is positive and significant at p = 0.003; commodity volatility is insignificant. The model reports F = 4.231 (p = 0.017), with adjusted R² of 6.9%. The authors infer that IDX investors "are beginning to value sustainability transparency as an indicator of long-term performance, even though global commodity fluctuations are less relevant in the domestic market."

Does ESG carry the text's 0.124 or Table 2's 0.017?

In the text, β1 = 0.124 is interpreted as 0.124% of return for each ESG point. For that same variable, Table 2 reports B = 0.017, t = 3.051 and the same Sig of 0.003. Table 1 expresses returns as decimals: the mean is 0.02, the SD is 0.38 and the range is -0.80 to 0.92. In those units, 0.017 translates to 1.7 percentage points of return per ESG point.

Apply both readings to the ESG standard deviation of 9.54 in Table 1. The table's coefficient implies about 16 points of annual return for a one-SD change in ESG. The text's interpretation gives about 1.2 points. These are our calculations using the authors' figures, and they leave the two readings roughly fourteen times apart. The larger estimate amounts to about 0.43 of the pooled return SD of 0.38 over the 130 firm-years.

The descriptive statistics disagree too. The text reports a mean return of 7.45% and an SD of 18.32%. It gives an ESG mean of 52.67, ranging from 28.45 to 78.92, and a volatility mean of 12.54%. Table 1 instead reports 0.02, 0.38, an ESG mean of 72.81 ranging from 49.32 to 90.67, and a volatility mean of 38.54.

Neither set accommodates the constant of 25.887. Combining the Table 1 means with the Table 2 slopes produces a fitted mean return above 27, versus the reported 0.02. We could not find units that reconcile the constant with the slopes and means.

The volatility estimates also conflict. The text gives β2 = 0.056 and p = 0.312. Table 2 gives B = 0.000, t = -1.063 and Sig = 0.290. A positive coefficient in the text becomes a printed 0.000 with a negative t in Table 2. The p-values also change, from 0.312 to 0.290.

Same-year association, thin inference

Assume Table 2 contains the intended estimates. From the paper's description, a firm's ESG score is paired with its return for the same year. Nothing establishes that investors could see the score before that return window began. We did not find a description of when each Refinitiv score became observable relative to its matched return window. A score drawn partly from an annual report cannot be available at the start of the year it explains.

The trader needs the score first.

I can live with the 6.9% adjusted R². Annual returns are noisy; low R² alone is no reason to discard a signal. The t-stat is the weaker link. Those 130 observations belong to 26 firms. If ESG scores barely change within a firm's five years, the t of 3.051 treats as independent what is largely 26 firms observed five times.

We did not find clustered standard errors. On a stacked panel, the Durbin-Watson statistic of 1.987 depends on row order and tells us little about serial correlation within firms. There are five calendar years and no year effects. A common shock to mining returns can therefore load onto whichever regressor moves with the calendar. Requiring firms to remain listed throughout 2019 to 2023 also selects survivors. That is acceptable for an association; a backtest would need to include firms that left.

Three commodities squeezed into a firm-year measure

Commodity price volatility spans 9.64 to 95.51 over the 130 observations. Standard deviations calculated market-wide for three commodity series could yield at most fifteen values, five for each commodity. The paper calls its regressor a "mixed commodity index". A single index would have even less variation across firms.

Some assignment to firms must therefore have taken place. We could not find how the authors combined the three commodities or mapped them to firms. The method calls for a standard deviation of prices, the text expresses the variable in percent, and Table 1 supplies no unit. With a standard deviation of price levels, the commodity quoted in the largest units would dominate.

The authors explain the null as "likely because the mixed commodity index (gold, coal, nickel) reduces individual effects, or due to the dominance of internal ESG factors in the Indonesian market". Their mixed-index explanation leads straight back to measurement. The blend may capture no individual firm's exposure: coal producers and gold producers respond to different prices. Explaining 6.9% of returns with ESG and volatility gives no basis for declaring ESG dominant. And a t of -1.063 on this blend cannot support the claim that commodity fluctuations matter less to IDX miners.

Reconcile the estimates before a portfolio test

A portfolio test first needs a single set of estimates in consistent units, with a constant that agrees with the means. ESG scores need availability dates and should be lagged into the following return year. Firms should be matched to their own commodities. Inference also needs to account for the panel, alongside the size and leverage controls the authors propose for future work.

We could not run any of these tests. Our data lacks the IDX mining universe and historical Refinitiv ESG scores, as well as coal and nickel price series. The disagreement between text and tables also leaves us without a target estimate to replicate. Substituting US miners would examine another market. Using only gold volatility would examine another regressor, rather than the three-commodity blend. A lagged test would address a predictive extension, a separate claim from the association reported in the paper.

One result would change my view: a regression using lagged ESG for the same 26 firms, retaining a slope near Table 2's 0.017 with clustered errors.