Evaluating a new earnings indicator. Can we improve the timeliness of existing statistics on earnings by using salary information from online job adverts? (ESCoE DP 2020-19)

cube-no-animation-1

Evaluating a new earnings indicator. Can we improve the timeliness of existing statistics on earnings by using salary information from online job adverts? (ESCoE DP 2020-19)

By Jyldyz Djumalieva, Stef Garasto,

Go to next section

This paper examines how the salary information from online job adverts might be used to improve the timeliness of official statistics on earnings. The unique dataset underpinning the analysis contains over 51 million adverts for UK positions, collected between January 2012 and September 2018. The data was sourced from Burning Glass Technologies, a leading labour market intelligence company. We trial a mixture of forecasting approaches, including traditional econometric models and the relatively newer recurrent neural networks. For 2 out of 13 industries and for 5 out of 6 occupation groups, salaries from online job adverts are shown to improve the accuracy of earnings forecasts over and above official data on its own. More broadly, this paper provides a detailed methodology for evaluating a novel data source, such as salaries from job adverts, to inform an official statistical series, such as earnings.

Related publications

Research Projects