Panel data of the 28th and 29th rounds of the Russian Longitudinal Monitoring Survey - Higher School of Economics (RLMS-HSE) in 2019 and 2020 served as empirical basis of the study. The RLMS-HSE survey is an annual survey of the population conducted on the basis of a probability stratified multi-stage territorial sampling¹. The RLMS-HSE data contain detailed information about socio-demographic and economic characteristics of the respondent and their household. In particular, big blocks of the survey questions are devoted to the respondent’s employment in the labor market and their household income. An important factor for choosing the empirical basis of the study was panel structure of the RLMS-HSE survey. The panel data structure makes it possible to analyze changes in employment and economic situation at the micro level, and identify socio-demographic variation of these changes. Additionally, the study uses aggregated indicators published by Rosstat (the cost of a fixed basket of consumer goods and services, subsistence minimum, unemployment rate) to control regional characteristics (Rosstat 2022).

When studying changes in employment characteristics, the main focus of the study is people of working age as of 2019 (females aged 18-54 and males aged 18-59). In order to assess the pandemic impact on employment, we have limited the sample to those individuals who had paid employment in the labor market in 2019. The sample included only those respondents whose information was present in the data for both 2019 and 2020. The total sample size of the study equaled to 5784 individuals. Tables P1 and P2 (Annex) present distribution of the respondents by main socio-demographic group, as well as characteristics of their employment in 2019.

The paper has also analyzed changes in household income. In the course of the analysis, the RLMS-HSE data on households were combined with data on individuals, making it possible to combine information on household income and employment of household members. The study object in this case is a household. The sample selection criteria included availability of household income data in 2019 and 2020, as well as values of independent variables included in the regression analysis. The final sample size equaled to 6,198 households. Table P3 (Annex) outlines main socio-demographic characteristics of households.

• A separate objective of the study is to test the hypothesis about the impact of vulnerable employment on the probability of deteriorated position of a person in the labor market and their reduced household income during the COVID-19 pandemic, made by the authors in their previous paper (Kartseva, Kuznetsova 2020). In the context of the pandemic, workers of the following three groups are considered vulnerable:
• self-employed and individual entrepreneurs;
• employees of non-governmental enterprises of the industries most affected by the pandemic (transport, culture, leisure and entertainment, physical culture and sports, hospitality and tourism, trade, catering, household services and education²;
• workers with precarious employment (Kartseva, Kuznetsova 2020). Precarious employment is determined according to the approach proposed in the papers (Bobkov 2017; Bobkov et al. 2017). An employment is considered precarious if at least one of the following signs is present: low wages (less than the regional subsistence minimum); unofficial wages; work without official contract; salary delay; salary reduction, involuntary reduction of working hours; forced unpaid leave.

The study constructs an indicator of vulnerable employment, which takes the value 1 if the employee is attributed to at least one of the above groups and 0 – otherwise. In 2019, half of the employees (50.4%) had employment recognized as vulnerable (see Table. P2, Annex).

The basis of the analysis of changes in household income in the context of the pandemic is the indicator of per capita household income, defined as the ratio between the total nominal household income and the number of its members. We compare per capita household income in 2019 and 2020. By analogy with wages, in order to ensure intertemporal and inter-regional comparisons, per capita incomes are adjusted to the cost of a fixed basket of consumer goods and services in the region and to the 2019 prices.

It should be noted here that specifics of income data collection under the Russian Longitudinal Monitoring Survey may influence assessment of the pandemic consequences. The fact is that within the framework of the Survey information about income is collected at the end of the year as of 30 days prior to the Survey. In this regard, comparing 2019 and 2020, we cannot take into account financial assistance that was provided in the midst of the pandemic and quarantine to various categories of households, including families with children. Thus, our comparisons are more focused on assessing longer-term consequences of the epidemiological crisis.

In general, according to the RLMS-HSE data, 53% of the households experienced per capita household income reduction in 2020 compared to 2019. The average reduction equaled to 20%. The paper has considered the following three variants of the per capita household income reduction indicator: a 10% and over reduction, 20% and over, and 30% and over reduction. An expanded range of income reduction indicators allows us to get an idea of sustainability of impacts of various factors, as well as how they are associated with weaker or, vice versa, stronger income shocks.

In 2020, 31% of the households faced a 10% and over reduction in income, income of 20% of the households was reduced by 20% and over, and 13% of the households experienced a 30% and over reduction in income (Table P6, Annex). Families with children under the age of 18, households in megapolises and regional centers, as well as households with vulnerable employees more often experienced reduction in income. (Table P6, Annex).

During the COVID period families with high income were affected the most. We have considered quintiles of per capita household income in 2019 as income groups. It turned out that in 2020, about 17% of the poorest households and more than half of the households from the highest income group experienced a rather insignificant reduction in income - by 10% and over. A significant reduction in income is also much more often observed among households of the highest income quintile: about 30% of the households from the fifth group experienced a 30% and over reduction in per capita income over a year. For comparison: among households of the 1st and 2nd quintile, such a significant reduction in income is much less common, registered in 7-8% of cases. Therefore, income mainly decreased among those who had something to lose.

The study has undertaken an econometric analysis of socio-demographic and economic factors determining the probability of job loss in 2020. Let’s remember that the analysis is carried out for those individuals with paid employment in 2019. The following logit model has been evaluated (M1):

(1)

lost_i is employment indicator of a person in 2020 (1 — employed, 0 — otherwise), dem_i is a vector of socio—demographic characteristics of an individual and their household that potentially affect their employment (including such indicators as gender and age of the agent, education, place of residence, marital status, paternal status with children of different ages), work_i is a vector of indicators characterizing employment of a person in 2019 (including the employment type according to the Russian Classification of Occupations 010-2014), indicators of industry classification of employment, indicator of self-employment and low wages (below the regional subsistence level for the third quarter of 2019) and unreg_i is the 2020 regional unemployment rate. The regional unemployment rate is included in the set of explanatory variables in the analysis of employment characteristics as a factor characterizing local labor market strains, which significantly vary across the Russian regions.

To determine significance and magnitude of the vulnerable employment effect on the actual probability of job loss in the short term during the pandemic, the following logit model has been evaluated(M2):

(2)

lost_i is employment indicator of a person in 2020 (1 — employed, 0 — otherwise), dem_i is a vector of socio—demographic characteristics of an individual and their household that potentially affect their employment (including such indicators as gender and age of the agent, education, place of residence, marital status, paternal status with children of different ages), vulemp_i is an indicator of vulnerable employment in 2019, and unreg_i is the 2020 regional unemployment rate. The model (2) does not include employment characteristics of an individual, since the vulnerable employment indicator is largely based on them.

For those people who remained employed in 2020, determinants of the probability of wage reductions has been evaluated. The study used logit models (M3 and M4) similar to those used to assess the risk of job loss:

(3)

(4)

wage_loss_i is an indicator of a 10% and more reduction in wages in 2020 (compared to 2019), dem_i is a vector of socio-demographic characteristics of a person and their household, work_i is a vector of employment indicators of a person in 2019) , vulemp_i — a vulnerable employment indicator in 2019, and unreg_i is the 2020 regional unemployment rate .

The study also used a regression logit model (M5) to assess factors influencing the risks of household income reduction:

(5)

hhinc_loss_i is the indicator of a 10%, 20% and 30% reduction in per capita household income in 2020 (compared to 2019), hh_i is a vector of household characteristics in 2019, including the number of children under the age of 18, family members of retirement age, members with vulnerable employment, members with other employment, type of household place of residence, one of the five quintile income groups the household is affiliated with in 2019.

Table 1 presents evaluation results of the job loss logit models during the coronavirus pandemic (M1 and M2 models).

The conducted econometric analysis shows statistically significant gender variation in the probability of job loss – the probability of becoming unemployed is 2 percentage points (p.p.) higher among females compared to males. Distribution by age shows that young people (aged 16 – 29) experience the highest risk of job loss. The risk of job loss is 3 p.p. higher among young employees compared to employees of the middle age. For the older age group (45 years and over), the probability of job loss does not statistically differ between older employees (45 and over) and employees of the middle age.

The study results failed to identify any significant impact of being married/having a partner on the risk of job loss. The presence of one or two children under the age of 18 is also insignificant in terms of the probability of job loss. However, if an employee has three or more children, then the probability of job loss is statistically significantly higher compared to the one without children. Further analysis showed that this effect is registered among females only –three and more children increase the probability of job loss by 3 p.p. (evaluation results of M1 and M2 models on subsamples of men and women are available upon request).

Education is the most important factor for maintaining employment during the pandemic. Employees without professional education were more likely to lose their job by 2-4 p.p. (depending on the model) compared to employees with higher education. A similar effect is registered among workers with basic vocational education (3 p.p.).

In the short term, the pandemic-related crisis has had a greater impact on the employment of residents of metropolises (Moscow, St. Petersburg). For employees living in Moscow and St. Petersburg, the risk of job loss is significantly higher on average compared to those living in regional centers. This is largely due to the employment structure of metropolises – there is a high share of services out there, which were affected the most during the pandemic. Employees in cities that are not regional centers, on the contrary, experienced a slightly lower risk of job loss, but this effect is statistically insignificant.

The employment characteristics prior to the pandemic have a significant impact on the probability of job loss during the crisis period. Those employed in trade, hospitality and tourism, household services, and catering had significantly higher risks of job loss. Unskilled and mid-skilled workers were more likely to lose their job compared to highly skilled workers (by 5 p.p. and 3 p.p, respectively). In addition, a higher risk of job loss was registered among self-employed and workers with low earned income.

The study has confirmed the hypothesis earlier made by the authors (Kartseva, Kuznetsova 2020) that vulnerable employment prior to the pandemic has a statistically significant negative impact on the probability of employment during the pandemic (M2 model). For an employee with vulnerable employment in 2019 the probability of job loss in 2020 is 5 p.p. and more higher compared to other employees.

Table 1 also presents results of econometric assessment of the wage reduction risk factors during the pandemic (M3 and M4 models).

The analysis failed to identify any significant variation in the probability of wage reduction by gender and age. The risk of pay cuts during the pandemic hardly differs among workers with different levels of education. A slightly lower probability of wage reduction is registered among workers with secondary vocational education; however, this result is statistically insignificant. Being married/having a partner is not a significant factor in terms of the probability of wage reduction. In contrast to the risk of job loss, the probability of wage reduction is not statistically different among workers (both men and women) with and without children.

Compared to employees living in regional centers, workers from other cities (which are not regional centers) were more likely to experience wage reduction. On the contrary, wage reduction among those employed in rural areas, was less frequently reported.

Certain characteristics of employment prior to the pandemic significantly affect the risk of wage reduction. A relatively high probability of earned income reduction was registered among those employed in such industries as trade, hospitality and tourism, household services, catering, construction, transport and communications. Employees with low wages are significantly less likely to experience reduction in earned income compared to employees with higher earnings. The employment type and qualification of the employee have no statistically significant impact on the risk of wage reduction.

Vulnerable employment is the most important significant factor of earnings reduction. The probability of wage reduction among those with vulnerable employment prior to the pandemic is 4 p.p. higher compared to other workers.

The paper has analyzed factors of household income reduction in 2020 compared to 2019. As a dependent variable, three indicators of per capita household income reduction have been considered (a 10%, 20%, and 30% reduction). The inclusion of several variants of the indicator in the analysis makes it possible to identify factors associated with both insignificant and more significant changes in income.

Table 2 outlines results of the regression analysis of the relationship between the probability of household income reduction and socio-demographic and economic characteristics of the household and its members. It is shown that children under the age of 18 in the household significantly increase the risks of income reduction – the higher the number of children, the higher the probability of income reduction. For example, for families with one child, the probability of a 30% reduction in income is 4 p.p. higher compared to families without children, while for families with 3 or more children – by 14 p.p. Having pensioners in the family, on the contrary, had a small deterrent effect, reducing the probability of income reduction by 2-3 p.p. for families with one pensioner and by 4-5 p.p. for families with two or more pensioners.

Household income on the eve of the pandemic turned out to be the most important factor of income reduction in 2020. As it was already mentioned in the discussion of descriptive statistics (Section II and Table P6, Annex), more well-off families were more likely to lose income. The regression analysis results confirm this finding. In particular, the risks of a 30% reduction in household income among the poorest quintile were 2 p.p. lower compared to the third (middle) income group, and 26 p.p. higher among the quintile with the highest income. For a 20% reduction in income, similar values equal to 7 p.p. and 28 p.p., respectively, and 15 p.p. and 26p.p. for a 10% reduction. A possible explanation for a higher impact of the pandemic on income of more well-off households is that such households are more integrated into the labor market and focused on earned income, and they are also more likely to include employees of the most quarantine-affected industries. Our results are indirectly confirmed by decreased income inequality recorded by official statistics. Thus, according to Rosstat data, the Gini coefficient (a measure of income inequality in a population) decreased from 0.412 in 2019 to 0.406 in 2020, and the coefficient of funds, measuring the differences between the decile groups with maximum and minimum income - from 15.5 to 14.8 (Rosstat 2021), that is, the reduction in inequality was largely due to a relative decrease in income of the richest groups.

Higher risks of income reduction are observed for households living in the metropolises (significant for small income shocks) and in rural areas. It should be particularly noted here that at the level of descriptive analysis, higher risks of income reduction in rural households were not identified: the share of those who faced income reduction is slightly lower compared to the sample average (see Table P6, Annex). This is due to the fact that households with lower income in 2019 were less likely to experience income reduction during the pandemic. Inclusion of information about income on the eve of the pandemic in the model makes it possible to better understand what has happened to poorer households, which, in particular, include families living in rural areas.

The impact of the number of employed household members on the risks of income reduction during the pandemic depends upon the employment type. Thus, workers with vulnerable employment in the household significantly increase the risk of a 10% reduction in income. At the same time, other workers in the household (without vulnerable employment) significantly reduce the risks of income shock – by 2-3 p.p. for families with 1 non-vulnerable worker and by 4-5 p.p. for families with 2 or more members with non-vulnerable employment. It should be noted here that the latter fact, as in the case of the risk of income reduction among rural households, is not distinguishable when analyzing descriptive statistics (on average, 29% of the households in the sample and 32-34% of the households with vulnerable employment experienced a 10% or more reduction in income, see Table P6, Annex) and is defined only when information about household income on the eve of the pandemic is included in the regression model.