Hi all!
I investigate the effects of certain legislative conditions (c.law (year and industry specific value)) on labour market outcomes (wages). I have yearly micro-level panel data, with observations on the level of individuals (xtset id year, yearly) and some added controls on higher levels.
Additionally, I want to look at the changes of total log employment levels, calculated in-sample as sum of all individuals employed in a given industry and year.
My question now is about the model that I need to use for this micro-to-macro level analysis. With the approach to use the individual level regression, as depicted above, one finds more relationships significant (due to the high number of observations, even though the outcome does not change within industry-year groups).The use of aggregated group means is also not suitable in my understanding, as a lot of information is lost using the transformation, it yields biased estimates and one might wrongly reject the hypothesis of interest (see Becker, Dominik, Wiebke Breustedt, and Christina Isabel Zuber. 2018. “Surpassing Simple Aggregation: Advanced Strategies for Analyzing Contextual-Level Outcomes in Multilevel Models.” 31 Pages / methods, data, analyses, Online First. https://doi.org/10.12758/mda.2017.05.).
So am I right to assume, that the best model to use would be a multilevel model (or mixed model) via mixed or do I miss out on other models that would fit better with my data?
Thanks a lot in advance for your feedback!
Kind regards, Paula
I investigate the effects of certain legislative conditions (c.law (year and industry specific value)) on labour market outcomes (wages). I have yearly micro-level panel data, with observations on the level of individuals (xtset id year, yearly) and some added controls on higher levels.
Code:
xtreg lnwage i.year i.industry c.law $Personal_Controls $Industry_Controls, fe vce(robust)
My question now is about the model that I need to use for this micro-to-macro level analysis. With the approach to use the individual level regression, as depicted above, one finds more relationships significant (due to the high number of observations, even though the outcome does not change within industry-year groups).The use of aggregated group means is also not suitable in my understanding, as a lot of information is lost using the transformation, it yields biased estimates and one might wrongly reject the hypothesis of interest (see Becker, Dominik, Wiebke Breustedt, and Christina Isabel Zuber. 2018. “Surpassing Simple Aggregation: Advanced Strategies for Analyzing Contextual-Level Outcomes in Multilevel Models.” 31 Pages / methods, data, analyses, Online First. https://doi.org/10.12758/mda.2017.05.).
So am I right to assume, that the best model to use would be a multilevel model (or mixed model) via mixed or do I miss out on other models that would fit better with my data?
Thanks a lot in advance for your feedback!
Kind regards, Paula
