<![CDATA[Profitability in the Indonesian retail subsector reflects the joint operation of operating-cycle efficiency, capital-structure composition, and margin discipline. The DuPont identity (DuPont de Nemours, 1919; Soliman, 2008) expresses Return on Assets (ROA) as the product of Net Profit Margin (NPM) and Total Asset Turnover (TATO); the static trade-off theory (Modigliani & Miller, 1963; Myers, 1984) and the pecking-order hypothesis (Myers & Majluf, 1984) provide competing predictions on the sign of the leverage–profitability relationship. This study presents a pooled-OLS DuPont decomposition of ROA with the Debt-to-Asset Ratio (DAR) as an incremental control, using a balanced firm-year sample drawn from retail subsector firms listed on the Indonesia Stock Exchange (IDX) over 2022–2024. The study deliberately frames the OLS specification as an associational decomposition rather than as a strong determinant model, because two of the three predictors (NPM, TATO) are accounting-arithmetic components of the dependent variable (ROA). A purposive sampling procedure applied to a population of 29 listed retail firms produced a balanced sample of 19 firms with three years of complete audited financial statements, yielding 57 firm-year observations. The OLS regression was estimated with standard classical-assumption diagnostics (Kolmogorov-Smirnov normality; VIF multicollinearity; scatterplot and Glejser heteroscedasticity; Durbin-Watson autocorrelation), all satisfied at conventional thresholds. Results show that TATO and NPM are positively and significantly associated with ROA (TATO: B = 0.030; t(53) = 5.88; p < 0.001; NPM: B = 0.914; t(53) = 14.06; p < 0.001), while DAR is statistically indistinguishable from zero (B = 0.008; t(53) = 0.61; p = 0.542). The simultaneous test confirms joint significance (F(3, 53) = 69.68; p < 0.001) with an adjusted R² of 0.786 — a magnitude that is mathematically expected under the DuPont identity rather than a novel determinant finding. Findings are interpreted as initial, context-specific associational evidence; the modest sample, pooled-OLS specification, survivorship-related selection criterion, and DuPont arithmetic between the predictors and the dependent variable all constrain the strength of any determinant interpretation. The study identifies fixed-effects panel estimation with firm-clustered robust standard errors, omitted-variable controls (firm size, sales growth, year effects), short-term/long-term debt disaggregation, and non-linear DAR specifications as further-research priorities.]]>
