Abstract:We present the Stata commands probitfe and logitfe, which estimate probit and logit panel data models with individual and/or time unobserved effects. Fixed effect panel data methods that estimate the unobserved effects can be severely biased because of the incidental parameter problem (Neyman and Scott, 1948). We tackle this problem by using the analytical and jackknife bias corrections derived in Fernandez-Val and Weidner (2016) for panels where the two dimensions ($N$ and $T$) are moderately large. We illustrate the commands with an empirical application to international trade and a Monte Carlo simulation calibrated to this application.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the bias in the fixed - effect estimator in non - linear models (such as binary response models) in panel data models due to the incidental parameter problem. Specifically, when the model contains individual effects or time effects, with the sample size \(N\) fixed, as the time dimension \(T\) increases, or with the time dimension \(T\) fixed, as the sample size \(N\) increases, the estimates of the fixed - effect estimator \(\hat{\beta}\) and the Average Partial Effects (APEs) will become inconsistent. This inconsistency stems from the fact that each unobserved effect has only a finite number of observations for estimation, leading to estimation bias.
To solve this problem, the author introduces two bias - correction methods: Analytical Bias Correction and Jackknife Bias Correction. These methods are applicable to panel data sets or other pseudo - panel data structures where both dimensions (\(N\) and \(T\)) are relatively large. Through these correction methods, more accurate estimates of model coefficients and average partial effects can be obtained.
### Main contributions
1. **Analytical Bias Correction**: By estimating the main terms of the bias and subtracting these bias terms from the fixed - effect estimate, the bias is reduced.
2. **Jackknife Bias Correction**: By estimating on different sub - samples and combining these estimation results, the bias is reduced. The paper introduces six different Jackknife correction methods, including ss1, ss2, js, sj, jj and double.
### Application examples
The paper demonstrates the effectiveness of these methods through an empirical application of international trade flows and a Monte Carlo simulation experiment. These examples show that by using these bias - correction methods, the accuracy and reliability of the estimates can be significantly improved.
### Key technical details
- **Model specification**: The paper considers probit and logit models with individual effects and time effects.
- **Bias correction**: Analytical Bias Correction reduces bias by estimating the bias terms \(B_\beta\) and \(D_\beta\) and subtracting these bias terms from the fixed - effect estimate. Jackknife Bias Correction reduces bias by estimating on different sub - samples and combining these estimation results.
- **Scope of application**: These methods are applicable to panel data sets of moderate length, such as traditional micro - economic panel surveys, international cross - national panels, and US state - level panel data.
### Conclusion
By introducing these bias - correction methods, the paper provides an effective way to deal with the bias problem of the fixed - effect estimator in panel data, especially when both \(N\) and \(T\) are relatively large. The implementation of these methods enables researchers to estimate model parameters and average partial effects more accurately, thereby improving the reliability and accuracy of empirical research.