operationalise 'education' in cross-national survey

[Guest]

Hi there,
I'd like to include education as a control in my analysis. I am using the ESS, so my data is cross-national. For better comparison across different educational systems I'd like to include a continuous variable based on 'years in education'. I like the following operationalisation:

"Education is assessed continuously to enhance the comparability of educational systems between countries (Berigan and Irwin, 2011). The impact of outliers is minimized by equalizing all scores higher than the country-year-specific mean plus 3 standard deviations."

Could anyone tell me how to code it ?
My first thought was using a foreach-loop somehow but I am a little overwhelmed on how to do it, especially, on how to limit the values to 3 sd from the mean....
Thanks a lot in advance!!

[Maarten Buis]

Years of education cannot be used to increase comparabiltiy across countries. The idea of comparability is that the same number means (roughly) the same thing regardless of the country. The problem is that in heavily tracked countries the same number of years of education can mean wildly different things depending on what track they are in, and whatever track you are in, that number will be completely uncomparable with the same number of years of education in a less tracked system. So years of education does not contain the necessary information (in which track were these years obtained) to make them comparable. No amount of statistical trickery can create information that was not present in the inputs.

Instead you could look at a measure designed for being internationally comparable, like the ISLED: http://www.harryganzeboom.nl/ISLED/index.htm The advantage is that the ISLED was actually designed for the ESS, so there is code ready to use (SPSS code, but the translation to Stata is straightforward) The link on that site to the article does not seem to be working, but here it is: https://doi.org/10.1093/esr/jct026

[Guest]

Hello Marten,

thanks a lot for your response. Looks interesting, I'll definitely take a look at it. However, after skimming through the article, it seems that it is based on the ISCED too. Around 70.000 cases were not possible to harmonize into the ES-ISCED. It would be a shame to loose all these cases. I will check the article anyways!

Second idea would be to create a relational variable for education. Like working out arithmetically standardised values around local means and defining local means by country and broad birth cohort.
Conceptually, I know how this could work. Practically unfortunately I do not...

Thanks!

[Maarten Buis]

your second idea would not work if it is based on years of schooling, as it does not use the tracking information.

I would instead start by looking at what is going on with these 70,000 cases, are they from particular waves or particular countries in which necessary information somehow was not asked? Or are they using a different harmonization scheme. If the latter is the case then there is usually a translation available form one scheme to the other.