Indonesian property study's Market Value Added result depends on firm dummies

This paper gives a trader no reason to buy an Indonesian developer with high Market Value Added. The measure subtracts book equity from market capitalization, tying it to the same share price that determines the return.

Putri and Dumilah examine whether accounting ratios and a market-based measure explain annual stock returns for property and real estate companies on the Indonesia Stock Exchange. Earnings per share divides net profit by shares outstanding; return on equity divides net profit by total equity. Their third measure, Market Value Added, is share price times shares outstanding minus book equity. It measures the premium investors pay over the capital shareholders contributed. The authors reason that higher earnings on equity, or a market premium over book, should attract buyers. They calculate return as the annual price change plus dividends, divided by the prior price.

The population covers 92 subsector companies from 2020 to 2024. Requiring continuous listing and complete data reduces it to six companies and 30 firm-years. The authors estimate pooled, fixed-effects and random-effects regressions in EViews 13. Fixed effects assigns a dummy to each firm; the Chow and Hausman tests select that model.

There, MVA records t = 3.431978 (Prob. 0.0025). ROE's coefficient is 3.116689, with t = 2.603972 (Prob. 0.0166), while EPS contributes little at t = -0.637467 (Prob. 0.5307). The overall F has Prob. 0.061987. The authors interpret adjusted R-squared of 0.261440 as the three indicators accounting for 26.14% of return variation. Across the sample, the average firm-year return was -0.037349 and its standard deviation was 0.181937.

Six survivors out of 92

Only 6.5% of the population makes the sample. Keeping firms that remained listed through 2024 selects on the outcome period. Survivorship therefore enters by construction: developers that delisted or stopped filing during the COVID years disappear. A constant, three slopes and five firm dummies leave 21 residual degrees of freedom.

What happens when the firm dummies go?

Little survives. Pooled MVA has t = 0.578322 (Prob. 0.5680), while ROE has t = 1.542470 (Prob. 0.1350). The model's adjusted R-squared is 0.008872; its F-test Prob. is 0.372244. Random effects produces much the same picture, with ROE at Prob. 0.0961 and MVA at Prob. 0.4473. Weighted adjusted R-squared is 0.005557.

The model-selection tests do carry information. Chow reports a cross-section F of 2.778259 on (5, 21), Prob. 0.0445. Hausman reports 11.237835 on 3 d.f., Prob. 0.0105. These results establish that firm intercepts matter and correlate with the regressors. They do not establish that a trader could earn the coefficient. Fixed effects identifies the slopes from five annual observations per firm after demeaning. If the model is right, Hausman rejection means the correlation between firm intercepts and regressors biases the pooled and random-effects slopes.

The share price enters both sides

MVA subtracts equity from market capitalization. We could not find the date of the price used to calculate it. If the authors use the year-end close, the same Pt appears inside MVA and in the return numerator. A rally within a firm will tend to raise MVA in that year as well. The t of 3.43 on this relationship is partly mechanical.

ROE brings a separate timing problem. Annual net profit and equity become available when the report is filed, after the return year has ended. We did not find a lagged specification. A trader therefore could not have opened a position using these figures at the start of the window they explain.

The reporting adds uncertainty. The fixed-effects table gives the MVA coefficient as 7.77E-14, whereas both the t-test section and the discussion give 0.000777. The descriptive text reports a minimum MVA of 5.17 trillion; the table prints -5.17E+12. For property stocks, that sign matters: negative MVA means the shares trade below book. The t-test section also calls H₃ "rejected" in the very sentence that describes ROE as significant.

Five firm dummies enter the joint F-test

The authors themselves are careful on this point. They write that "the significant individual effects of MVA and ROE should not be interpreted as evidence of a significant simultaneous effect of all three variables."

EViews' overall F for the fixed-effects regression is the reported 2.283198 with Prob. 0.061987. It tests the three indicators together with the five firm dummies. Using R-squared of 0.465181, eight regressors and 21 residual d.f. in the standard formula reproduces F ≈ 2.28. The paper thus never reports a joint test of EPS, MVA and ROE conditional on firm effects. The printed output does not let us calculate one.

The dummies dilute that reported test. Given firm effects, a joint test confined to the three indicators would very likely reject at 5%: MVA has t = 3.431978 and ROE has t = 2.603972 on 21 d.f. This reading strengthens the authors' within-firm result. The case against it then rests on the mechanical connection between MVA and price. Adjusted R-squared presents the same attribution problem. Its 0.261440 includes the variation explained by firm intercepts, whereas the pooled model, with only the three indicators and a constant, reaches 0.008872.

Firm dummies lift adjusted R-squared from 0.008872 to 0.261440; the printout cannot divide that gain between the dummies and the three indicators.

No trade to assess yet

We found neither a portfolio nor a holdout period. There is no formed trading edge against which to measure costs or out-of-sample decay. We could not run our own test. Our data covers US equities; we lack Indonesian price and fundamental histories as well as IDX subsector membership needed to reconstruct the 92-name universe. Substituting US homebuilders or REITs would test a different market.

One result would change my mind. Rank all 92 names, including those that delisted, on prior-year MVA scaled by equity and on prior-year ROE. Use only reports already filed, then hold those ranks over the following twelve months. If the ROE spread remains positive, Putri and Dumilah have found something. With five annual returns, a mean of -0.037349 and fixed effects carrying the result, the current paper describes how six survivors' prices moved alongside their own balance sheets.