#Google product trends series#

The Google search data are what is often referred to as “unstructured” data. Google Trends and metro area initial UI claimsĪ fundamental hurdle with using Google Trends to explain UI take-up rates is distilling the “signal” from the “noise” of the data. Sources: Authors’ calculations based on data from the New York Department of Labor, Pennsylvania Department of Labor & Industry, Texas Workforce Commission, Washington State Employment Security Department, and U.S. County populations and weekly initial claims are aggregated to the DMA level to match the Google Trends search data. Darker shaded regions have higher initial UI claims per capita, while lighter shaded regions have lower initial UI claims per capita. Notes: Regions within states correspond to Nielsen designated media areas, or DMAs (see note 4 for more information). Initial unemployment insurance (UI) claims per 1,000 residents for metro areas in New York, Pennsylvania, Texas, and Washington These are the patterns in UI take-up that we aim to explain with the Google Trends data.įigure 1. The maps highlight two things: 1) the significant degree of heterogeneity in unemployment due to the Covid-19 crisis across metro areas, with metro areas in Washington and New York reporting much higher numbers of claims per capita than those in Texas and Pennsylvania (panel A), and 2) how labor market conditions have changed over the past few months in metro areas, with most outside of Texas seeing some improvement (panel B compared with panel A). 4įigure 1 shows maps of initial UI claims per 1,000 residents at two points in time during the Covid-19 pandemic: first, for the week ending March 28, 2020, which is approximately the peak week of initial UI claims, and second, for the week ending May 30, 2020, which is the most recent week that we have county-level claims data for. Finally, we aggregate the county-level UI clams data up to the DMA level to obtain a consistent metro-area-by-week panel of data. 3 We then average the daily indexes using the UI claim reference weeks. We normalize the daily Google Trends indexes for each metro area against the national index to obtain data that are comparable across jurisdictions. To these data, we then add Google Trends unemployment topic indexes at the designated market area (DMA) level, which is the smallest level of geographic disaggregation available that is consistent with the county-level UI claims data. The county-level initial UI claims data are for Pennsylvania, New York, Texas, and Washington, and we get them from their state governments (through the Haver Analytics database). We use two primary data sources in our analysis: weekly county-level initial UI claims and daily Google Trends indexes for the unemployment topic (see note 2). The Covid-19 pandemic’s impact on metro area UI claims Furthermore, it also appears likely that at least some of this correlation has been driven by UI policy changes and the news coverage surrounding them as opposed to search activity more directly related to filing a UI claim. metro areas within weeks (the cross-sectional dimension).

#Google product trends series#

metro areas across time (the time series dimension) and less so by variation across U.S. metro area level, the positive correlation between the Google Trends unemployment topic index 2 and the rate of unemployment insurance (UI) take-up during the Covid-19 pandemic is primarily driven by variation within U.S. during the early months of the pandemic here. 1 Just as an example, one can track how the number of Google searches with the term unemployment office has changed over the past week for the Chicago metro area or explore how unemployment became one of the top searched issues across the U.S. Since the onset of the pandemic, there has been growing interest in tracking labor market activity with “big data” sources like Google Trends.