Getting the online figures right: the academic struggle - 25 Nov. 2009

Harvard Law School Langdell Hall
Harvard Law School ZaNiaC via Flickr

The academic partners of the WageIndicator Foundation continue to address one of the key problems in looking at online surveys: how can we get them closer to the offline reality. 

In October we reported on a study by Stephanie Steinmetz and Kea Tijdens of the University of Amsterdam. That academic study is now also available online here.  The conclusion in October: while the current standard of online surveys is the best we can get, improvements are needed:

Steinmetz suggested several ways to improve the scientific value of the online surveys, including always having an offline reference survey at hand, although that would reduce the current cost and time advantages of online studies. Another ways was that we would have to live with the idea no survey would be really representative for the general population, including the online ones.

A second academic study has been produced by our partners at the Harvard Law School, written by Richard B. Freeman, Damian Raeses and Isabelle Ferreras, figures of our US website Paywizard. They call their study a "Preliminary Skirmish", equally suggesting that the academic debate is only starting.

Their conclusion: 

"Basic illustration of the tests on US data show significant bias in the non-representative Paywizard data, while techniques such as weighting and median quantile regression analysis (or any estimation that gives less weight to the extreme values) appears promising venues for reducing the selection bias. We hope that this and more elaborated tests yet to be defined will form the basis of a methods template that will be systematically applied by every national WageIndicator team on their data in order to generate a strong data set on wage and employment issues across the globe."

The Harvard study can be found here, the UvA AIAS study here. 

Reblog this post [with Zemanta]


News every year