Has the pandemic permanently changed job requirements?

·       Education and training

·       Cognitive and mental requirements

·       Physical demands

·       Environmental conditions

Job requirements information is currently being collected over a 5-year reference period, 2019–23. This article reports on research conducted to assess whether job requirements have been permanently changed as a result of the coronavirus disease 2019 pandemic.

How can one determine if the job requirements measured by ORS changed permanently? If one could measure requirements for employers before the start of the pandemic and also remeasure requirements for those same employers after the impact of the pandemic, then one could attribute the change from “before” to “after” to the pandemic. In general, however, employers are in the ORS sample only once. That is, a nationally representative sample is chosen for each of the 5 years, 2019–23, and only a small number of establishments are in the sample for more than 1 year. Another approach would be to ask employers whether the pandemic affected their job requirements. For those employers who said it had, one could try to obtain measures of the current requirements and measures from before the pandemic. Unfortunately, this type of approach would be quite burdensome on the respondents and also rely on them to recall how job requirements have evolved in recent years.

Instead of using these methods, BLS relied on two different approaches, which were less direct than the two just mentioned. In the first approach, during the third year (August 2020–July 2021) of the 5-year reference period, BLS field economists asked each employer whether any job requirements changed. The field economists recorded the employer’s response in one of four categories: (1) no change, (2) permanent change, (3) temporary change, and (4) other. To be clear, owing to response burden issues, this question was not asked for each requirement but for any requirement that might have changed. No information was collected on which requirement might have been involved.

The responses in the “other” category were excluded from the analysis. Prepandemic requirements were collected in the case of a temporary change. And those in the temporary-change category were grouped with those with no change. That is, a comparison of those in a combined no-change and temporary-change category with those in the permanent-change category could reveal which requirements were changing.

The second approach attempted to uncover the possible effects of the pandemic by comparing the requirements of jobs before the onset of the pandemic with the requirements of jobs after the impact of the pandemic. The ORS data that BLS economists collected between September 2018 and July 2021 were divided into the following three groups:

1. A baseline group composed of all jobs whose information was collected in the first year of the reference period (September 2018–August 2019)

2. A prepandemic group composed of all jobs in the second year of the reference period whose information was collected before March 1, 2020 (August 2019–February 2020)

3. A postpandemic group, which began after the start of the pandemic in the United States, composed of all jobs whose information was collected beginning on June 1, 2020, up to the end of the third year of the reference period (June 2020–July 2021)

A couple of clarifications are worth mentioning for the three groups. First, data collected in March, April, and May 2020 were not used. Employers were assumed to have adjusted to the pandemic’s impact in this period. Second, though the third group is called a postpandemic group for brevity, this name does not imply that the pandemic is not still with us. Rather, the name signifies that the data were collected after the employers had time to adjust to the impact of the pandemic.

Of the two approaches just mentioned, only the second one is discussed in this article. For the first approach, not reported in this article, only about 1 percent of the analysis sample was in the permanent change group. This finding suggests that any changes in job requirements that are measured using this survey would have affected only a small portion of the economy. Of the analysis sample, 77 percent reported no change in requirements and 22 percent indicated only temporary changes.

In the rest of this article, the first section reviews the ORS elements (job requirements) that might have been changed by the pandemic. The second section briefly discusses the methods for determining which requirements may have changed significantly. The third section provides, as background, the levels and changes in each of the elements during the period. The fourth section contains the analysis, which assesses whether statistically significant changes in elements remain after controlling for differences in occupational composition.

Job requirements

Has the pandemic changed job requirements permanently or at least those job requirements that are collected in the ORS? In general, the onset of the coronavirus has led to two easily seen impacts on jobs. The first is increased use of personal protective equipment and attention to personal hygiene in order to minimize exposure to the disease. The second is expanded efforts to increase distance between workers, which could affect how workers communicate with each other. To create a list of job requirements potentially changed by the pandemic, one must review the full list of ORS elements and select those that may have been affected.

As indicated earlier, ORS collects information in four areas: education and training, cognitive and mental requirements, physical demands, and environmental conditions. The first impact of the coronavirus (people minimizing exposure to the coronavirus) on jobs is equivalent to minimizing exposure to biohazards, which is excluded from ORS collection. Thus, the focus when considering ORS elements for analysis is primarily on physical distancing and its impact.

Did education and training requirements change because of the pandemic? Given that such job requirements change quite slowly,¹ they likely did not change noticeably. In the analysis, BLS examines this presumption with respect to on-the-job training and prior work experience requirements. Among the cognitive demands, several elements had potential to change. Efforts to increase physical distance, which could be permanent, may affect the prevalence of such ORS requirements as teleworking, working with the general public, working around crowds, the frequency of work being reviewed, and whether a supervisor is present. Increasing distancing may also limit verbal interactions and reduce the level of people skills required. Also, the physical demand of speaking may be affected. The reorganization of work to increase distance may result in greater keyboarding. Lastly among physical demands, reaching at or below the shoulder is also examined. An increased use of plexiglass and other barriers might heighten the need for reaching.

For the final ORS area environmental conditions, no elements were examined. As noted, the use of personal protective equipment and personal hygiene steps to minimize exposure are excluded from the ORS. In addition, it is not apparent that the other elements in this category would be subject to permanent change.

Table 1 displays the elements selected for the analysis. Defining a categorical variable to correspond to each element is necessary. Within the education and training area, one variable is defined for prior work experience and another variable for on-the-job training to determine whether each element appears. Thus, if the element equals “Yes,” the categorical variable will equal 1, while if the element equals “No,” the categorical variable will equal zero.

For the cognitive demand elements selected, teleworking, working with the general public, working around crowds, and a supervisor is present have response categories of “Yes” and “No.” Therefore, variables are defined on the basis of whether these elements are present. For the frequency of work being reviewed, the variable is 1 for more than once per day, zero otherwise. For verbal interactions, the variable is 1 for interactions of once per hour or fewer, zero otherwise. Finally, for people skills, it is 1 for more than basic, zero otherwise.² For the three physical demand elements, the variable is set equal to 1 if the element is present. If otherwise, the variable is set to zero.

Methods

BLS uses two methods to determine which job requirements may have been changed significantly by the pandemic. Each method has advantages and disadvantages. The first method uses what is known as a logit regression, which is used in multivariate analysis when the dependent variable is categorical, as is the case here. One can use such a regression to address the following question: Has the pandemic led to significant changes in the proportion of the economy with each of the elements? Two types of comparisons are tried. The first, using just the pre- and postpandemic groups defined earlier as the analysis sample, has as independent variables an indicator for whether or not the job is in the postpandemic group and indicators for the occupations of the jobs. Variables are used for occupations so as to control for differences by group in occupational composition. If the coefficient on the indicator for the postpandemic group is statistically significant, then one might believe that the pandemic has changed the job requirement. This is the case because jobs in the postpandemic group may have changed their requirements as a result of the pandemic, while those in the prepandemic group will not have.

The baseline group should be examined as well in order to assess if the assumed difference between the groups can be attributed to the pandemic. In the second comparison, one expands the analysis sample to include the baseline group, and one adds to the logit regression specification an indicator for whether the job is in the baseline group. If trends noted between the pre- and postpandemic groups are also present between the baseline and the prepandemic groups, then one might suspect that assuming that the entire difference between pre- and postpandemic should be attributed to the pandemic is too strong. Accordingly, at least part of the change between the pre- and postpandemic groups should probably be attributed to a continuation of preexisting trends. On the other hand, if the change between the pre- and postpandemic groups marks a reversal of that between the baseline and prepandemic groups, then one’s assurance that the more recent trend can be attributed to the pandemic will be enhanced.

The second method is occupation specific. For each element, the proportions in which the element is present for a given occupation is calculated for the pre- and postpandemic groups. A comparison of the two proportions, which uses a statistical hypothesis test, addresses the question of whether the proportion changed for an element for that occupation. One advantage of this method is that it facilitates making inferences for specific occupations. But a disadvantage is that assessing economywide changes is more difficult.

Levels and changes in selected elements

As background for the analysis of the next section, table 2 shows the percentage levels of each of the elements considered for the baseline, prepandemic, and postpandemic groups.³ The fourth and fifth columns present the difference in percentage points between the prepandemic and the baseline groups and the pre- and postpandemic groups, respectively. The final column displays the difference between the fourth and fifth columns to show whether any tendencies observed between the baseline and prepandemic groups were continued between the pre- and postpandemic groups.

Examining the difference between the pre- and postpandemic groups first (table 2, fifth column), we can see that three changes are statistically significant. The proportion of jobs with teleworking increased by 3.8 percentage points from 7.6 percent to 11.4 percent, significant at the 1-percent level. The proportion working around crowds declined by 1.0 percentage point to 2.7 percent. The share of employment requiring reaching at or below the shoulder fell by 3.1 percentage points to 76.6 percent. The latter two changes are significant at the 5-percent level.

It turns out, however, that of these three requirements, only the pre- to postpandemic change in teleworking remains statistically significant when compared with the baseline to prepandemic change. That is, the 3.8-percentage-point increase in teleworking following the start of the pandemic closely follows a 2.0-percentage-point decline in teleworking in the prior period. The tendency to work less around crowds was already evident in the baseline period so that the difference between the pre- and postpandemic groups and between the prepandemic and baseline groups is not statistically significant. For the element reaching at or below the shoulder, the difference between the two changes is not statistically significant, even though the movement from baseline to prepandemic was slightly in the opposite direction. On the other hand, the element working with the general public has emerged as statistically significant in the final column. This element increased by 5.4 percentage points from the baseline group to the prepandemic group, so the fact that it declined slightly in the prepandemic and postpandemic periods represents a reversal.

Accounting for differences by group in occupational composition

In this section, logit regressions and occupation-specific hypothesis tests are conducted to assess changes in job requirements after accounting for differences in occupation composition across the groups. The first set of logit regressions is run on prepandemic and postpandemic groups. The second set of logit regressions is run on baseline, prepandemic, and postpandemic groups. The first set of occupation-specific hypothesis tests examines six-digit occupational groups. The second set of occupation-specific hypothesis tests examines two-digit occupational groups.

Logit regressions

As noted earlier, the logit regressions that are run have as dependent variables a categorical variable for each element that is subject to analysis. The specifications of the logit regressions focus on comparing the three groups: baseline, prepandemic, and postpandemic. In the first set of specifications, only the pre- and postpandemic groups are included, and a dummy variable is used when the observation comes from the postpandemic group. In the second set of specifications, all three groups are included, and in addition to the dummy variable for the postpandemic group, a dummy variable is used for the prepandemic group. The other independent variables in the regressions are dummy variables for occupations. Occupations are specified in two ways, each having its advantages and disadvantages. First, six-digit occupational groups, which are detailed identifications of occupations, are used, but to facilitate convergence of the logit regressions only occupations containing at least 30 jobs are used.⁴ Second, two-digit major occupational groups are used, which allow the whole sample to be analyzed. The regressions on the two-digit occupations institute controls which impose the restriction that all six-digit occupations within two-digit occupations have the same impact on the dependent variable.

Table 3 displays, for both specifications, the estimated marginal effects applicable for each element’s regression. These effects are the average over the sample of moving from the prepandemic group to the postpandemic group, given the estimated coefficients. In other words, the marginal effects provide an estimate of how different the probability of an outcome would be, on average, if jobs were in the postpandemic group rather than in the prepandemic group. For most elements, the marginal effects are not statistically significant, suggesting that these requirements did not undergo major changes after the onset of the pandemic. Two notable exceptions are teleworking and working around crowds. Teleworking is 3 to 4 percentage points higher after the pandemic, depending on the specification, while working around crowds is 1 to 2 percentage points lower. Both these results are consistent with increased physical distancing. The only other element with statistical significance in either regression specification is people skills, in which the six-digit occupation specification suggests a reduced demand for advanced people skills.

As discussed in the methods section above, expanding the analysis sample to include the baseline group in the regressions is useful to determine if any differences between the pre- and postpandemic groups are merely the continuation of earlier changes. Table 4 displays, for both six-digit and two-digit occupation controls, two different columns of marginal effects and a column showing the difference between the two. In other words, the first column of each set of three columns provides the change between the baseline group and the prepandemic group, the second column of the three provides the change between the prepandemic group and the postpandemic group, and the third column of the three provides the difference between the two changes. A test of the hypothesis that this difference equals zero relies on an alternative scenario than did the earlier set of specifications (in table 3) about what job requirements may have been affected by the pandemic. The assumption is that the change noted between the baseline group and the prepandemic group would have continued were it not for the pandemic.

As shown in table 4, it is rare for the difference in the change between groups to be statistically significant. Only in the case of teleworking is this difference statistically significant for both occupational specifications. Because teleworking decreased between the baseline and prepandemic groups, the increase in teleworking between pre- and postpandemic registers as a reversal of a trend and thus is more likely attributable to the onset of the pandemic. A decrease in the baseline period for working around crowds rendered the difference in the change insignificant.

Comparisons by occupation

We divided the sample into six-digit and two-digit occupations and compared the proportion of a given element in the prepandemic group with the corresponding proportion in the postpandemic group. Restricting attention to six-digit occupations with at least 30 observations yields 252 occupations to assess. Table 5 summarizes the results of hypothesis tests for six-digit occupations. The hypothesis is that the proportion of a given element for each occupation in the prepandemic group is the same as the corresponding proportion in the postpandemic group. The result of each hypothesis test is distributed across four columns in table 5. A rejected hypothesis means that the difference between the periods is statistically significant. If the difference is statistically significant and the postpandemic group proportion is greater than the prepandemic category, then the first column is used. If it is less than the prepandemic group, then the second column is used. If the difference is not statistically significant, then the third column is used. In some cases, the statistic for the hypothesis test cannot be calculated, so the fourth column is used. This situation occurs if either both proportions equal 1 or both equal zero.

Examining table 5, we can see that for each element, either a high share of the hypothesis tests is not significant, implying that one cannot reject the hypothesis that the proportion is the same for the two groups or the test statistic cannot be calculated, which occurs more frequently when the proportions of an element are close to zero or to 1. The high share of insignificant hypothesis tests is not surprising. This is the case given that job requirements tend to change slowly over time under normal circumstances. Thus, there may be little reason to expect significant changes by occupation.

In the two columns that display significant results, the teleworking element stands out because of the high proportion of significant and positive cases, suggesting that the rate of teleworking had been increasing. This result is consistent with the finding of the logit regressions.

Table 6 is the same statistical test as conducted in table 5, except the hypothesis tests are done for major occupations instead of detailed ones. Because of the greater aggregation of the occupations, a test statistic is rarely not computed, although a high proportion of the tests are statistically insignificant. Teleworking has the highest number of statistically significant cases, and all these significant cases are consistent with growing rates of teleworking over time.

Conclusion

This article reports on an examination of whether the pandemic led to changes in job requirements measured in the ORS. Job-related data collected from the ORS between September 2018 and July 2021 are examined across three groups: baseline, prepandemic, and postpandemic. The methods rely on assumptions about what is behind job requirement differences across these three groups. That is, for much of the analysis, any difference between the prepandemic and the postpandemic groups are attributed to the pandemic. For part of the analysis, the assumption is that a change between the baseline and prepandemic groups would have continued had the pandemic not occurred. The results are generally consistent across the statistical approaches. Findings indicate that a pandemic-induced shift in a job requirement is rare. The exception to this conclusion is the element of teleworking, which moved toward increased use.