2018
DOI: 10.31235/osf.io/8xb4z
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

How to Deal With Reverse Causality Using Panel Data? Recommendations for Researchers Based on a Simulation Study

Abstract: Does X affect Y? Answering this question is particularly difficult if reverse causality is present. Many social scientists turn to panel data to address such questions of causal ordering. Yet even in longitudinal analyses reverse causality threatens causal inference based on conventional panel models. Whereas the methodological literature has suggested various alternative solutions, these approaches face many criticisms, chief among them to be sensitive to the correct specification of temporal lags. Applied re… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

1
58
0
1

Year Published

2019
2019
2023
2023

Publication Types

Select...
8

Relationship

0
8

Authors

Journals

citations
Cited by 54 publications
(60 citation statements)
references
References 20 publications
1
58
0
1
Order By: Relevance
“…In order to estimate whether a causal relation exists between two variables, a study design is needed that is able to prevent bias in the estimates due to reverse causation, incorrect specification of the lag of the effect, and confounding. Methods that can take all these issues into account require at least three measurements of both dependent and independent variables (Allison et al 2017;Hamaker et al 2015;Leszczensky and Wolbring 2019), which were not available in TRAILS. Future research that has at least three measures of all variables at study would be able to overcome these issues with estimating how characteristics induce change in one another.…”
Section: Discussionmentioning
confidence: 99%
“…In order to estimate whether a causal relation exists between two variables, a study design is needed that is able to prevent bias in the estimates due to reverse causation, incorrect specification of the lag of the effect, and confounding. Methods that can take all these issues into account require at least three measurements of both dependent and independent variables (Allison et al 2017;Hamaker et al 2015;Leszczensky and Wolbring 2019), which were not available in TRAILS. Future research that has at least three measures of all variables at study would be able to overcome these issues with estimating how characteristics induce change in one another.…”
Section: Discussionmentioning
confidence: 99%
“…However, we must acknowledge that these alternative explanations cannot be entirely dismissed given that there seems to be no bulletproof solution, especially regarding reverse causality, which is particularly true for cross-sectional data but in many regards also extends to longitudinal data (see e.g., Vaisey and Miles 2017). With longitudinal data, more sophisticated methods are available (cf., Leszczensky and Wolbring 2018), but these methods are also not without their own problems. Therefore, further research is needed to corroborate our results.…”
Section: Discussionmentioning
confidence: 99%
“…Each structural equation contains an autoregressive term, a cross‐lagged term, a time trend, nine lagged covariates, and the idiosyncratic error. The cross‐lagged models address endoegenity due to reverse‐causality through the cross‐lagged terms, endogeneity due to omiited variables by imposing cross‐equation error covariance, compensation persistence through the autoregressive term, acquisition rate persistence through the autoregressive term (Leszczensky & Wolbring, ). We used the lagged covariates that we have used in the previous models.…”
Section: Resultsmentioning
confidence: 99%