Early UI Data Reveals Differential Impacts of the Economic Shutdown

Philip Garboden, Justin Tyndall, Blogs, COVID-19, Economy


By Justin Tyndall and Philip Garboden

Two weeks ago the Hawaii Department of Labor and Industrial Relations released its monthly snapshot of individuals receiving unemployment insurance (UI) benefits. The data represents only the 69,729 individuals receiving benefits on April 12th, 2020. 

Data is also presented at the county level, allowing for comparisons across our islands. Since the story is relatively similar across counties, here we discuss only the statewide impact.

Comparisons between the April data and earlier months are interesting, but should be interpreted with an understanding of how unemployment insurance operates in normal times. Most people think of UI as protecting employees who work for companies that are going out of business or downsizing. It does this, but it also fills a vital role in the construction industry (as well as other contract-heavy industries like administrative services). Construction work is highly volatile, with companies needing large crews for some periods, with delays in between. Unemployment insurance covers those gaps.

For this reason, a plurality of the roughly 6,000 monthly claimants pre-COVID were in construction and waste services. Two thirds were male, and the largest racial groups were Filipino, Hawaiian, and White.

Much of this has changed, of course, during the shutdown. UI now serves a very different role in our social safety net, covering a broad range of industries, but focused on those where the tourism shutdown and social distancing have had the greatest impact. Indeed Construction is the only industry where UI claims didn’t at least double from the pre-COVID period (although they did increase by nearly 50%).

Industry Impacts

Not surprisingly, the Accommodations/Food Service industry was hit the hardest, with 27,000 of the 70,000 early claims. During a typical year, these jobs make up 21% of our non-government workforce, but they represent 46% of the claims. Administration and Waste Management jobs (perhaps the most confusing category, including a wide range of support services for businesses) are also disproportionately high. The sector, which normally has 8% of the non-government jobs, now has 12% of the claims.

Retail and Transportation also have large numbers of UI claims — over 4,000 each in this early data, although Retail claims are proportionally lower than the number of retail jobs. Healthcare too has many claims, 3,700, but these claims are also proportionality lower than what we would expect, were all industries equally affected.

In summary, all major industries had made significant cutbacks as early as April 12th (except perhaps for Finance and Insurance), but these are all dwarfed by reductions in the tourism-focused industries.

The below figures indicate the relative number of claims across industries. The blue bars represent the industry’s share of the pre-COVID workforce, while the orange bars represent the share of claims made within the industry in April. When the orange bar exceeds the blue bar, this group is overrepresented among the unemployment claims.

Unfortunately the data on what occupations within those industries [1] were most impacted is highly limited. 47,000 of the 70,000 claimants are listed as “information not available.” We doubt that this missing data is distributed randomly, and therefore urge against putting much weight on data in the “OCCUPATION” table. The fact that a plurality of job losses are associated with “Food Preparation/Serving” is expected, but beyond that better data is needed.

Gender, Race, and Age

As noted above, because construction is a male-dominated profession, the majority of UI claims in normal times are from men even though the State’s workforce is split almost equally between men (51%) and women (49%) [2]. The shutdown, at least in the early periods, seems to have had a disproportionate impact on women, who make up 57% of the claims, likely because they are heavily represented in service sector jobs.

The race data is, unfortunately, also somewhat hard to interpret because it does not align with any other racial/ethnic categories. The number of claimants classified as “Others” (roughly 40%) likely reflects the high number of individuals in Hawaii who identify as more than one race. The fact that this number is higher than standard statistics on individuals with mixed-race heritage in Hawaii could suggest that this group is disproportionately impacted or that there is a missing data problem similar to the data on occupation. Without knowing how to allocate the Others category, it is difficult to derive insight. 

In terms of age, the filers align fairly closely with the distribution of workers in the state. Individuals under 22 are slightly under-represented in the UI claims, and folks 25-54 slightly over-represented, but the differences amount to a few percentage points [3].

The stories these data tell align with other data we have brought to bear on this topic. The consequences of the shutdown span all industries, but hits hardest in industries most tied to the visitor and food service industries. Without quality occupational data, we cannot be definitive, but what we know about these industries is that they not only hire a disproportionate number of women (something we can see in the data) but also higher numbers of folks without a college or graduate degree (a finding that is prevalent in national analyses). As the State works to develop policies to help weather what promises to be an extended recession, it will be important to keep these differential impacts in mind.

Note: We will continue to update this analysis of UI as new data become available over the next few months.

Industry breakdown by county:

[1] The difference between industry and occupation statistics is subtle but important. One can work, for example, in the Art occupation but work for a company in the Accommodations industry. One can work in information technology in the Real Estate Industry.

[2] According to data from the 2018 American Community Survey.

[3] ibid.