Difference-in-Differences and Treatment x Pretest Interactions: A Critique of Dual-Centered ANCOVA
DOI:
https://doi.org/10.35566/jbds0208Keywords:
Difference-in-Differences, Dual-Centered ANCOVA, Parallel Trends Assumption, Effect Heterogeneity, Causal InferenceAbstract
This paper critically examines dual-centered ANCOVA, proposed by Lin and Larzelere (2020) and Larzelere and Lin (2025), which is claimed to accommodate a Treatment X Pretest interaction while preserving the difference-in-differences (DiD) estimate. We take issue with their claims on four grounds. First, DiD without an interaction term does not assume constant treatment effects; under the parallel trends assumption, it identifies the average treatment effect on the treated even when treatment effects vary across pretest scores. Second, ANCOVA requires correctly modeling the conditional expectation of the outcome in order to identify causal effects, whereas DiD does not, so the concern about omitting an interaction term applies to ANCOVA but not to DiD. Third, dual-centered ANCOVA is not a distinct ANCOVA-type method but merely a re-expression of DiD, sharing the same estimand and identification conditions. Fourth, the claimed innovation of presenting both estimates---the DiD estimate of the treatment effect and the ANCOVA estimate of the interaction term---within a single model offers little practical advantage, because obtaining correct standard errors requires a second analysis. Overall, we argue that dual-centered ANCOVA is based on a misunderstanding of what DiD assumes and is better understood as a re-expression of existing analytic methods rather than a methodological innovation.
References
Allison, P. D. (1990). Change Scores as Dependent Variables in Regression Analysis. Sociological Methodology, 20, 93–114. doi: https://doi.org/10.2307/271083
Angrist, J. D. (1998). Estimating the Labor Market Impact of Voluntary Military Service Using Social Security Data on Military Applicants. Econometrica, 66(2), 249–288. doi: https://doi.org/10.2307/2998558
Brorsen, B. W., Lin, H., & Larzelere, R. E. (2025). Critique of enhanced power claimed for Quasi-ANCOVA and Dual-Centered ANCOVA. PLOS ONE, 20(1), e0317860. doi: https://doi.org/10.1371/journal.pone.0317860
Callaway, B., & Sant’Anna, P. H. C. (2021). Difference-in-Differences with Multiple Time Periods. Journal of Econometrics, 225(2), 200–230. doi: https://doi.org/10.1016/j.jeconom.2020.12.001
Dahabreh, I. J., & Bibbins-Domingo, K. (2024). Causal Inference About the Effects of Interventions From Observational Studies in Medical Journals. JAMA, 331(21), 1845–1853. doi: https://doi.org/10.1001/jama.2024.7741
Hernán, M. A., & Robins, J. M. (2020). Causal Inference: What If. Boca Raton: Chapman & Hall/CRC.
Holland, P. W. (1986). Statistics and Causal Inference. Journal of the American Statistical Association, 81(396), 945–960. doi: https://doi.org/10.1080/01621459.1986.10478354
Holland, P. W., & Rubin, D. B. (1983). On Lord’s Paradox. In H. Wainer & S. Messick (Eds.), Principals of Modern Psychological Measurement: A Festschrift for Frederic M. Lord (pp. 3–25). Hillsdale, NJ: L. Erlbaum Associates.
Huitema, B. E. (2011). The Analysis of Covariance and Alternatives: Statistical Methods for Experiments, Quasi-Experiments, and Single-Case Studies (2nd ed.). Hoboken, NJ: Wiley. doi: https://doi.org/10.1002/9781118067475
Jamieson, J. (2004). Analysis of Covariance (ANCOVA) with Difference Scores. International Journal of Psychophysiology, 52(3), 277–283. doi: https://doi.org/10.1016/j.ijpsycho.2003.12.009
Kim, Y., & Steiner, P. M. (2021). Gain scores revisited: A graphical models perspective. Sociological Methods & Research, 50(3), 1353–1375. doi: https://doi.org/10.1177/0049124119826155
Larzelere, R., & Lin, H. (2025). An Innovation to Test Treatment X Pretest Interactions within Difference-in-Differences. Journal of Behavioral Data Science, 5(1), 51–66. doi: https://doi.org/10.35566/jbds/larzelere
Lechner, M. (2011). The Estimation of Causal Effects by Difference-in-Difference Methods. Foundations and Trends in Econometrics, 4(3), 165–224. doi: https://doi.org/10.1561/0800000014
Lin, H., & Larzelere, R. E. (2020). Dual-centered ANCOVA: Resolving contradictory results from Lord’s paradox with implications for reducing bias in longitudinal analyses. Journal of Adolescence, 85, 135–147. doi: https://doi.org/10.1016/j.adolescence.2020.11.001
Lord, F. M. (1967). A Paradox in the Interpretation of Group Comparisons. Psychological Bulletin, 68(5), 304–305. doi: https://doi.org/10.1037/h0025105
Lüdtke, O., & Robitzsch, A. (2025). ANCOVA versus Change Score for the Analysis of Two-Wave Data. The Journal of Experimental Education, 93(2), 363–395. doi: https://doi.org/10.1080/00220973.2023.2246187
Maris, E. (1998). Covariance Adjustment versus Gain Scores—Revisited. Psychological Methods, 3(3), 309–327. doi: https://doi.org/10.1037/1082-989X.3.3.309
Morgan, S. L., & Winship, C. (2015). Counterfactuals and Causal Inference: Methods and Principles for Social Research (2nd ed.). Cambridge: Cambridge University Press. doi: https://doi.org/10.1017/CBO9781107587991
Petersen, M. L., & van der Laan, M. J. (2014). Causal Models and Learning from Data: Integrating Causal Modeling and Statistical Estimation. Epidemiology, 25(3), 418–426. doi: https://doi.org/10.1097/EDE.0000000000000078
Robins, J. M. (1986). A New Approach to Causal Inference in Mortality Studies with a Sustained Exposure Period—Application to Control of the Healthy Worker Survivor Effect. Mathematical Modelling, 7(9–12), 1393–1512. doi: https://doi.org/10.1016/0270-0255(86)90088-6
Rubin, D. B. (1974). Estimating Causal Effects of Treatments in Randomized and Nonrandomized Studies. Journal of Educational Psychology, 66(5), 688–701. doi: https://doi.org/10.1037/h0037350
Senn, S. J. (2006). Change from Baseline and Analysis of Covariance Revisited. Statistics in Medicine, 25(24), 4334–4344. doi: https://doi.org/10.1002/sim.2682
Słoczyński, T. (2022). Interpreting OLS Estimands When Treatment Effects Are Heterogeneous: Smaller Groups Get Larger Weights. The Review of Economics and Statistics, 104(3), 501–509. doi: https://doi.org/10.1162/rest_a_00953
Snowden, J. M., Rose, S., & Mortimer, K. M. (2011). Implementation of G-Computation on a Simulated Data Set: Demonstration of a Causal Inference Technique. American Journal of Epidemiology, 173(7), 731–738. doi: https://doi.org/10.1093/aje/kwq472
van Breukelen, G. J. P. (2006). ANCOVA versus Change from Baseline Had More Power in Randomized Studies and More Bias in Nonrandomized Studies. Journal of Clinical Epidemiology, 59(9), 920–925. doi: https://doi.org/10.1016/j.jclinepi.2006.02.007
van Breukelen, G. J. P. (2013). ANCOVA versus CHANGE from baseline in nonrandomized studies: The difference. Multivariate Behavioral Research, 48(6), 895–922. doi: https://doi.org/10.1080/00273171.2013.831743
