The size effect in Nurfitriana's regression of 146 recent IPOs gives traders a selection hypothesis they cannot yet trade. Its two significant coefficients explain a return that starts at the offer price, available only to allocated buyers. Size deserves further research. The rest of the regression table warrants more doubt than its significance levels suggest.

What is being explained?

The paper builds its argument around information asymmetry and signaling. An issuer publishes a prospectus, where favorable figures on assets, profitability, capital structure and growth send a positive signal. First-day demand then lifts the closing price above the offer. The paper measures this underpricing as (first-day close minus offer price) divided by offer price. The issuer leaves that money on the table; an investor allocated shares earns it.

The author regresses the first-day return on five variables:

The sample covers every company listed between 2023 and 2025. Described as a "saturation" sample of 146, it was assembled from websites and estimated by OLS in SPSS. The paper does not name the exchange. The author's Indonesian affiliation and references to studies of the Indonesia Stock Exchange suggest that market. First-day gains were common: the paper reports underpricing for 68%, 90% and 85% of listings in 2023, 2024 and 2025.

Size and proceeds are the only variables that survive. The size coefficient is positive: B = 0.0754, standardized beta 0.728, t = 6.897 and p =.000. For proceeds, the coefficient is negative: B = -0.2937, standardized beta -0.293, t = -2.980 and p =.003. Sales growth (p =.352), ROA (p =.928) and DER (p =.989) amount to noise. The abstract advises investors to assess this information "to improve their prospects of earning returns." That claim needs testing.

Access to the offer price

An investor buying at the open on day one captures none of the measured move from offer to first-day close. Capturing it requires an allocation in the book. The author acknowledges this in the introduction: underpricing "actually benefits investors because they can purchase shares at a lower price on the primary market." We nevertheless found no allocation, oversubscription or fill data anywhere in the paper. The abstract still promises investors better returns. Size's standardized beta of 0.728 ranks issuers by their first-day gains, while the data say nothing about how much an outside buyer could earn.

The year-end figures sharpen the timing problem. Among the 2023 listings underpriced on day one, 67% were lower at year-end. The corresponding figure for 2024 was 61%. The pattern flipped in 2025, when 68% of underpriced listings finished higher and 32% lower. In 2023 and 2024, the money shown by these figures was on day one, available to investors who received shares at the offer.

Allocation is the missing link.

Large balance sheets, relatively small raises

The significant coefficients describe conditional relationships. Holding proceeds fixed, size is positive; holding size fixed, proceeds are negative. Together they suggest that large balance sheets raising relatively little had the biggest first-day gains. That makes a coherent selection hypothesis. The proceeds result also runs against the author's prediction.

H5 predicted that higher proceeds would mean higher underpricing, reasoning that large raises imply high market capitalization and stable prices. Instead, B = -0.2937 with t = -2.980. The discussion explains the negative coefficient through dilution and external-funding risk, a story introduced only after the result. H2 correctly predicted the positive size coefficient: "Companies with high total assets will demonstrate stable performance, which will result in higher underpricing." But this conflicts with the paper's information-asymmetry argument. If underpricing compensates investors for information asymmetry, larger, more transparent issuers should require a smaller discount. Size instead has the table's largest standardized beta, 0.728.

The hypotheses clash elsewhere too. The text says high ROA "indicates low underpricing," then H3 predicts that higher ROA accompanies higher underpricing. At p =.928, the contradiction goes untested.

Bigger firms tend to raise more, so log assets and log proceeds should be strongly correlated. For its five-regressor model, the paper supplies just one tolerance value, 0.985. That single figure cannot establish separate identification of the size and proceeds coefficients. Those are the only results carrying the argument.

Annual listings total 144; the stated sample is 146

Annual listing counts are 78, 40 and 26. Their sum is 144, against a stated sample of 146 that supposedly covers the entire population. Two extra rows will not shift a t of 6.897. They do leave the census unresolved, especially when the source description goes no further than websites.

The diagnostics offer even less reassurance. At 0.725, the Durbin-Watson statistic is far below 2 and signals strong positive correlation between neighboring residuals. Row order determines what that means in a cross-section. If listings are ordered by date, 0.725 suggests time clustering; equation (1) contains no year dummies to absorb it. The reported proportion underpriced ranges from 68% to 90% across the three years. A regime shift of that kind is exactly what the residuals could retain. The author reports the statistic, then continues as though the test had passed.

The reported heteroscedasticity value, 0.928, matches the ROA p-value exactly. We did not find an R-squared, an F-statistic or descriptive statistics. Without them, interpreting the coefficients is difficult. Size's B = 0.0754 appears beside a constant of 171.5476 and a sales-growth B of -130.7582. Missing means and standard deviations leave us unable to tell whether underpricing is expressed in percent or fractions. The introduction also says "nearly 50%" of shares were underpriced over 2020 to 2025. Reconciling that with 68% to 90% in each of 2023 to 2025 would require much lower figures in 2020 to 2022.

We could not backtest the result. The universe is Indonesian; our data covers US equities. We also lack a source for offer prices, proceeds and point-in-time prospectus financials. Even those inputs would leave the allocation problem unresolved. Secondary-market prices cannot establish fills at the offer, and assuming such fills would manufacture the return being measured.

We have encountered this pattern before: a single-table Indonesian regression describes prices without producing a return signal. It appeared in our review of a bank-sector study. This paper offers more, because the size result at t = 6.897 is a large effect worth re-estimating with year controls and an allocation-weighted return. Evidence that the size premium survives among deals a non-anchor investor could actually fill would change our view.