The estimation relies on ordinary least squares (OLS), which build on several assumptions, including linearity, homoscedasticity and no multicollinearity (Wooldridge, 2010[1]). The plot of residuals (Figure A F.1) does not show any clear patterns or systematic deviations from the horizontal line at 0, suggesting that the linearity assumption is met. However, the residuals exhibit wider dispersion as the fitted values increase, indicating that the variance of the residuals is not constant. This pattern suggests the presence of heteroskedasticity. To address the issue of heteroskedasticity and ensure reliable hypothesis tests for variable significance and confidence intervals, robust standard errors are employed.
Promoting the Development of the Semiconductor Ecosystem in Panama
Another key assumption of OLS estimation is the independence of the regressors, or no multicollinearity. The presence of multicollinearity can complicate the interpretation of the model, as it becomes difficult to isolate the effect of each predictor. The variance inflation factor (VIF) is a useful metric for detecting multicollinearity. In this analysis (Table A F.1), VIF values are close to 1, indicating low correlation among the variables. The estimates obtained may be biased due to potential endogeneity arising from simultaneity or reverse causality. For example, wages and hours may be jointly determined. Additionally, the model may suffer from omitted variable bias which could impact the reliability of the findings. Future research, conditional on the availability of data, should explore methods to address endogeneity, such as using instrumental variables or dynamic panel data models in order to improve the robustness of the findings.
References
[1] Wooldridge, J. (2010), Introductory Econometrics: A Modern Approach, South-Western Cengage Learning, https://www.cengage.com/c/introductory-econometrics-a-modern-approach-7e-wooldridge/9781337558860PF/?app=cmp (accessed on 6 May 2026).