Modeling the employment rate in Russia: a spatial-econometric approach Текст научной статьи по специальности «Социальная и экономическая география»

For citation: Demidova, O. A., Daddi, P., Medvedeva, E. V. & Signorelli, M. (2018). Modeling the Employment Rate in Russia: a Spatial-Econometric Approach. Ekonomika Regiona [Economy of Region], 14(4), 1383-1398 doi 10.17059/2018-4-25 UDC: 331

O. A. Demidovaa), P. Daddib), E. V. Medvedevac), M. Signorellib)

a) National Research University Higher School of Economics (Moscow, Russian Federation; e-mail: demidova@hse.ru)

b) University of Perugia, Department of Economics (Perugia, Italy) c) PJSC "Post Bank" (Moscow, Russian Federation)

MODELING THE EMPLOYMENT RATE IN RUSSIA: A SPATIAL-ECONOMETRIC APPROACH 1

The purpose of this study is to identify factors that affect the level of employment in Russian regions. However, Russia is not a homogeneous country, and this effect may not be the same for all regions. That is why we split the regions of Russia into three groups, depending on the state of the labor market in this and neighboring regions. The HH (high-high) group comprises regions with a favorable situation in their labor markets, and which are also surrounded mostly by prosperous regions. Two groups of regions with a less favorable situation are located respectively in the south of Russia (LL1, low-low group 1) and southern Siberia and Zabaikalye (LL2, low-low group 2). We considered the twelve-year period from 2005 to 2016. As explanatory variables, we used variables for the attractiveness of the region, demographic characteristics of the region, and the degree of diversity of employees by economic activities. We tested hypotheses about differences in 1) the spatial effects and 2) the impact of the various explanatory variables for these groups of variables. To test our main hypotheses, we used spatial regression dynamic models estimated with the help of the generalized method of moments. Both main hypotheses received empirical confirmation. Spatial effects were different. The regions of the LL2 group are not affected by the situation in other local markets; regions of LL1 and HH groups are affected by the rest of Russia's regions, and the extent of this influence decreases with the increase in geographical distance between regions. Moreover, the regions of the LL1 group compete with neighboring regions: if the situation in one of them improves, then it draws on the resources of the others. Regarding the impact of the explanatory variables, the "group effect" was revealed for the variables: share of urban population, net migration rate, shares of people below and above working age, share of people with higher education. Our results can favor the better design of national and regional policies to improve labor market performance in Russia based on the heterogeneity of the Russian regions.

Keywords: employment, labor market, regional data, spatial effects, spatial models, labor policies, development policies

1. Introduction

The socio-economic development plan for the Russian Federation until 2020 states that the priorities of the state regional policy are (i) balanced socio-economic regional development, and (ii) the reduction of interregional disparities.

So, knowing how regions are distributed into high or low employment groups/clusters is a key empirical issue with significant policy implications. However, according to [1] Oschepkov and Kapelyushnikov (2015), there is no single joint Russian labor market; instead, there is a system of rather weak interrelated territorial/local labor markets. The reasons are mainly the low mobility of the Russian population and significant differences among regions located in different parts of Russia. The authors also note that there are

1 © Demidova O. A., Daddi P., Medvedeva E. V, Signorelli M. Text. 2018

two groups of regions that were quite stable in the time interval 2000-2014: "leaders" (with high employment, low unemployment, and high wages) and "outsiders" (with low employment, high unemployment, and low wages).

In our research, we sought to determine, by means of a dynamic spatial econometric approach, which factors affect one of the most important indicators of labor markets: the level of employment of the population. The focus was on the mutual influence of regions on each other. If we did not take account of such influence in the models used, we might have encountered the problem of shifting estimates due to the omission of an essential variable. At the same time, it was difficult to take account of the influence of regions on each other; in this case, the number of model parameters that would have to be estimated would exceed the number of observations. However, there are spatial-econometric models that enable the use of

1384 социально-демографический потенциал регионального развития

several parameters to take the mutual influence of regions into account.

In the next section, we provide a brief overview of the key theoretical aspects and the main literature on regional aspects of the Russian labor market. In the third section, we describe the distinction of Russian regions into groups using Moran plots and the "leader-outsider" approach, present our data sources, discuss the choice of explanatory variables, and state the main research hypotheses. In the fourth section, we describe the methodology of econometric modeling. The sixth section sets out the results of the estimation and their interpretation. The last section contains some concluding remarks and policy implications.

2. The Theoretical Aspects and Literature Review

An original feature of this paper is its use of the employment rate and not the more traditional unemployment rate indicator. Here we briefly provide a short theoretical explanation for this choice.

Although the unemployment rate is still adopted in both theoretical and empirical studies, a growing number of economists have shown the key importance of the employment rate and its relative advantages with respect to the unemployment rate, especially because of the difficulties with this latter indicator in clearly defining the "active search for a job" as the crucial feature distinguishing unemployed people from inactive ones. In addition, some international institutions have started to define key policy objectives in terms of employment rates; the main example is the European Union that, within the framework of the European Employment Strategy, in 2000 at the Lisbon Council defined total and female employment rates as the labor market's performance objectives, and a similar employment rate index has been confirmed with the "Europe 2020" strategy launched in 2010.

It should be stressed that the level and dynamic of the employment rate cannot be simply derived from the level and dynamic of the unemployment rate. Let us consider the relationship between the two labor market performance indicators. First, we define the unemployment rate (UR) as the percentage ratio between the number of unemployed person U (i.e. unemployed people actively searching for a job) and the labor force LF (i.e. employed plus unemployed persons); second, we define the employment rate (ER) as the percentage ratio between the employed persons (E) and the working age population P15-72; third, we define the participation rate (PR) as the percentage ratio between

labor force (LF) and the working age population P

15-72"

UR =

ER =

PR =

U x100 LF ' E x 100 P '

15-72

LF x 100

P

(1) (2) (3)

15-72

Starting from equations (1), (2) and (3), the employment rate can be redefined as the complement to one of the unemployment rate (divided by 100) multiplied by the participation rate:

ER =

E x 100 LF - U LF x 100

LF j

=11 - ur v pr. 100 j

(4)

From equation (4), we can derive the unemployment rate (UR) as the complement to one of the ratios between employment rate (ER) and participation rate (PR) (the result multiplied by 100).

UR = | 1 - — |x100. PR

(5)

Considering equation (5), a complex relationship emerges between the unemployment and employment rates; in fact, for example, a reduction in the unemployment rate is compatible with a reduction in the employment rate if the absolute value of this latter is lower than the absolute value of the reduction in the participation rate. Hence, it is not surprising that the unemployment rate can have different dynamics over time with respect to the employment rate.

For the reasons above mentioned, we adopted the employment rate as the key indicator of labor market performance in our study applied to Russian regions.

A growing body of literature investigates labor market performance at regional (sub-national) level, especially in large countries. In particular, the mutual influence of regions on each other in modeling the unemployment rates of regions in one or several countries is more often taken into account with the help of spatial-econometric models ([2] Caroleo and Pastore, 2010; [3] Mussida and Pastore, 2015, [4] Dolton et al., 2015; [5] Vega and Elhorst, 2016; [6] Manning and Petrongolo, 2017). There are several studies on the European regions, like [7] Head and Thierry (2006), [8] Ketterer and Rodriguez-Pose (2016). Some authors note the heterogeneity of the labor market and often identify clusters of regions, or they use 'core-periph-

ery' models. However, there are fewer papers devoted to the regional labor market in transition countries. A review can be found in [9] Huber (2007) and [10] Bah and Brada (2014). Russia is an example of such a country.

According to [3] Pastore and Missuda (2015, introduction), "the Russian case seems to be specific and interesting not only among other transition countries but also in the European perspective". [11] Vakulenko and Gurvich (2016) highlight that "high wage flexibility is an important salient feature of the Russian labor market"; [12] Kapelyushnikov et al. (2012) argue that the current model of labor relations in Russia is a combination of very formal rules embodied in the Labor Code and a great variety of informal arrangements that make it feasible to relax those rules. This is a substantially flexible system.

In the case of Russia, from a regional perspective, issues related to economic growth have been studied more extensively ([13] Solanko, 2008; [14] Ledyaeva et al., 2008, [15] Kholodilin et al., 2012; [16] Akhmedjonov et al., 2013; [17] Lehmann and Silvagni, 2013; [18] Dolinskaya, 2002).

Almost all the available studies on regional labor markets examine unemployment rates ([19] Demidova and Signorelli, 2012, [20] Demidova et al., 2013, [21] Demidova et al., 2015, [22] Blinova et al., 2015, [23] Blinova et al., 2016, [24] Rusanovskiy and Markov, 2016). Hence, there is a lack of studies modeling regional employment rates. The exception is the work of [1] Oschepkov and Kapelyushnikov (2015) mentioned above, in which the authors come to the following conclusions: (i) regional labor markets in Russia converge; (ii) both regions-"leaders" and re-gions-"outsiders" tend to form clusters of nearby or adjacent regions. Thus, it makes sense to try to identify such clusters. [25] Danilenko et al. (2017) attempt to reveal clubs for the unemployment rate using Moran plots. In this paper, in addition to Moran plots, the "leader-outsider" approach was used ([1] Oschepkov and Kapelyushnikov (2015). Considering the above-discussed relationship between unemployment rate and employment rate it is not surprising that in this paper, we obtain empirical results different from those of other studies focused on unemployment rates (e.g., [25] Danilenko et al., 2017).

3. Data and Variables 3.1 Data

Our sample consisted of 80 regions during the twelve-year period from 2005 to 2016. The majority of the data used in the research were availa-

Table 1

United subjects of the Russian Federation

Data Merging regions Incorporated as

01.01.2007 Taymyr Autonomous Okrug Krasnoyarsk Territory

Evenk Autonomous Okrug

Krasnoyarsk territory

01.07.2007 Kamchatka oblast Kamchatka territory

Koryak Autonomous Okrug

01.01.2008 Ust-Orda Buryat Autonomous Okrug Irkutsk region

Irkutsk region

01.03.2008 Chita region Zabaykalsky Territory

Aginsky Buryatsky Autonomous Okrug

01.07.2012 Moscow, Moscow oblast Moscow

ble for public access via the website of the Federal State Statistics Service (FSSS) of the Russian Federation. It was impossible to include earlier years in the research due to the different classification of industries before the year 2005.

Moreover, data on some regions were missing (the Republic of Chechnya, the Republic of Crimea and Sevastopol). In addition, the Kaliningrad region was not included in the study because it has no common borders with other regions of Russia. Moreover, during the reporting period, some regions underwent changes of an administrative-territorial nature. This altering of boundaries was taken into consideration, mitigated by an aggregating procedure (see Table 1).

3.2 The Splitting of Regions by Moran Plot

Russian regions are not homogeneous, and employment levels differ considerably among them. We distinguish a group of regions with an employment level above the average and a group of regions with a level of employment below the average. It is also necessary to take account of the weighted average employment rate in neighboring regions (the weights are given by the weighted matrix W matrix): it can also be above or below the average. Thus, we can distinguish four groups of regions. Traditionally used for such a division is the Moran chart, in which the horizontal axis states the standardized values of the employment rate Z, and the vertical axis states spatially weighted standardized values of the employment rate WZ.

In our analysis, a matrix of common borders was formed. It was represented by the following definition of W:

1386 COm/IA-nbHO-flEMOrPAOMHECKMM nOTEH^Afl PErMOHAflbHOrO PA3BMTMJI

W =

0

¡en

W-

len

w:

we

V n1

w

w

len 1n len

w

len

0

(6)

where wlen =

length in km of jont boundaries between regions i and j total length in km of all boundaries of region i

We used information on the length of joint

boundaries taken from the State real estate cadas-

tre.1 The matrix is line-normalized, so that w.. ac' ii

counts for the weights of a region; w.. = 0, if there is no boundary between regions i and j or if i = j. For each region, we have a point on the Moran scat-terplot which can be in one of four quarters. The first quarter represents the High-High group: this means that in the given region the employment rate is high and it is surrounded by regions also with high employment; it is a group of the most prosperous regions that positively influence each other. The second quarter represents the Low-High group: the employment rate in the region is low, but the neighbor region has a high level of employment; this is a group of disadvantaged regions that can receive some benefits from proximity to prosperous regions. The third quarter represents the Low-Low group: both the region and neighbors have low employment rates; this group comprises the most disadvantaged regions. The fourth quarter represents the High-Low group: the employment rate is high for the region and low for its neighbors; this is a group of prosperous regions, but the situation in them may worsen due to proximity to unfavorable regions.

We thus obtained scatterplots for each year from 2005 to 2016.2 Each point was labeled by the number of the corresponding region. To determine the list of regions for each group, we counted how many times the corresponding points were in each quarter. We obtained the results set out in the table below.

It should be stressed that these groups of regions are close to, but not identical with, the group of four unemployment groups of regions in Russia discussed in [25] Danilenko et. al (2017).

In both cases, regions of LL group for employment and HH group for unemployment are geographically split into two parts: South of Russia and South of Siberia. We consequently decided to separate the LL (employment) group into two

1 For the Sakhalin region these boundaries are measured by sea.

2 Available upon request.

groups. Additionally, HL and LH groups contain, respectively, only 5 and 11 regions. We decided to add those regions to larger groups; to determine the destination group, we used the "leader-outsider" approach followed by [1] Oschepkov and Kapelyushnikov (2015, the details are in the next section).

3.3 The "Leader-Outsider" Approach

[1] Oschepkov and Kapelyushnikov (2015) concluded that regions had stable positions in time on all indicators considered. They distinguished between "leader" and "outsider" regions. High employment and purchasing power of nominal wages together with a low unemployment rate are observed in leading regions. By contrast, high unemployment and a low level of economic activity characterize "outsider" regions. The same authors highlight that it is possible to allocate leading and lagging regions to clusters by geographic location. In their analysis, the authors discussed only the first and last ten regions (leaders and outsiders). Moreover, they did not give the exact list of regions for each indicator.

Following the same approach of [1] Oschepkov and Kapelyushnikov (2015), we made a list of regions for three indicators: employment rate, unemployment rate, and the purchasing power of nominal wages using rankings. We took the data on the employment rate from 2005 to 2016 for each region from Table 2. Then, for each year, the regions were arranged in descending order by the employment rate and were ranked (the first region had rank 1, the second one rank 2, and so on). In the final list, the regions were arranged in ascending order by the mean rank for 12 years (the smallest mean rank corresponded to the region with the highest employment rate). The result was the list of regions that follows.

It should be noted that the list of leaders is almost the same for all three indicators. If we now consider Table 1 again, we find that regions from HH and HL groups are at the top of the list in Table 2, while regions from LL and LH groups are at its bottom.

Finally, according to the Moran plot for employment rate and "leader-outsider approach", the Russian regions were divided into 3 final groups. HL and LH groups were attached to the corresponding groups (HL + HH, LH + LL). The LL group was split into two clubs: LL1 group (South of Russia) and LL2 group (South of Siberia) by geographical criteria. Regions 40 and 54 were "outliers". To determine the relevant group, we calculated the mean value of the employment rate in those regions and clubs, and we joined "outliers"

Table 2

List of employment groups

HH (High-High) Regions LL (Low-Low) Regions LH (Low-High) Regions

Vladimir region Belgorod region Bryansk region

Ivanovo region Voronezh region Orel region

Kaluga region Kursk region Ryazan region

Kostroma region Tambov region Republic of Bashkortostan

Smolensk region Republic of Adygea Penza region

Tver region Republic of Kalmykia Ulyanovsk region

Tula region Krasnodar Territory Kurgan region

Yaroslavl region Astrakhan region Tomsk region

Moscow Volgograd region Primorsky Territory

Republic of Karelia Rostov region Amur region

Republic of Komi Republic of Dagestan Jewish autonomous area

Arkhangelsk region Republic of Ingushetia

Nenets Autonomous Okrug Republic of Kabardino-Balkaria HL (High-Low) Regions

Vologda region Republic of Karachaevo-Cherkessia Lipetsk region

Leningrad region Republic of Northen Ossetia — Alania Republic of Mordovia

Murmansk region Stavropol Territory Samara region

Novgorod region Saratov region Chelyabinsk region

Pskov region Republic of Altay Novosibirsk region

Saint-Petersburg Republic of Buryatia

Republic of Marii El Republic of Tyva

Republic of Tatarstan Republic of Khakassia

Republic of Udmurtia Altay Territory

Republic of Chuvashia Zabaykalsky Territory

Perm territory Irkutsk region

Kirov region Kemerovo region

Nizhny Novgorod region

Orenburg region

Sverdlovsk region

Tumen region

Khanty-Mansi Autonomous Area — Yugra

Yamal-Nenets autonomous region

Krasnoyarsk Territory

Omsk region

Republic of Sakha (Yakutia)

Kamchatka territory

Khabarovsk Territory

Magadan region

Sakhalin region

Chukotka Autonomous Okrug

with a group which had approximately the same mean. As a result, regions 40 and 54 were joined to the LL2 group.

The resulting division by employment groups is depicted below.

Regions from the LL1 group are mainly located in the south of Russia. These are mainly agricultural areas characterized by a high level of informal employment.

It should be noted that the authoritative experts on the Russian labor market in their re-

port ([26] Gimpelson et al, 2017) also divide the Russian outsider regions into two groups: a group of southern republics and a group of regions of Southern Siberia (details can be found in [26]). They note that "the regions within each group have very similar structural and natural-geographical characteristics."

In previous research (see [25] Danilenko et al., 2017), we analyzed the difference in spatial effects and the determinants of unemployment rate for three clubs of Russian regions (HH, LL, HL). In the

Group Color

HH Grey

LL1 Dark grey

LL2 Light grey

Fig. Employment groups

Table 3

The list of Russian regions obtained by ranking

Employment (from highest to lowest Unemployment (from lowest to Buying power of nominal wage (from

level) highest level) highest to lowest)

Chukotka Autonomous Okrug Moscow Nenets Autonomous Okrug

Yamal-Nenets Autonomous region Saint-Petersburg Yamal-Nenets Autonomous region

Magadan region Chukotka Autonomous Okrug Tumen region

Saint-Petersburg Samara region Khanty-Mansi Autonomous Area — Yugra

Moscow Tula region Chukotka Autonomous Okrug

Khanty-Mansi Autonomous Area — Yugra Yamal-Nenets Autonomous region Magadan region

Murmansk region Lipetsk region Moscow

Kamchatka territory Belgorod region Saint-Petersburg

Tumen region Republic of Mordovia Sakhalin region

Nenets Autonomous Okrug Yaroslavl region Republic of Komi

Leningrad region Novgorod region Krasnoyarsk Territory

Samara region Kostroma region Murmansk region

Kaluga region Kaluga region Republic of Sakha (Yakutia)

Sakhalin region Leningrad region Kemerovo region

Republic of Mordovia Magadan region Irkutsk region

Yaroslavl region Republic of Tatarstan Tomsk region

Republic of Udmurtia Tver region Kamchatka territory

Kostroma region Nizhny Novgorod region Arkhangelsk region

Novgorod region Ryazan region Republic of Tatarstan

Vologda region Penza region Sverdlovsk region

Continued table 3

Employment (from highest to lowest level) Unemployment (from lowest to highest level) Buying power of nominal wage (from highest to lowest)

Nizhny Novgorod region Chelyabinsk region Zabaykalsky Territory

Sverdlovsk region Tumen region Leningrad region

Krasnoyarsk Territory Ivanovo region Republic of Karelia

Vladimir region Voronezh region Chelyabinsk region

Tver region Vologda region Omsk region

Republic of Tatarstan Kursk region Republic of Bashkortostan

Smolensk region Arkhangelsk region Republic of Khakassia

Kirov region Krasnodar Territory Kaluga region

Lipetsk region Vladimir region Amur region

Khabarovsk Territory Orel region Vologda region

Arkhangelsk region Sverdlovsk region Belgorod region

Chelyabinsk region Khanty-Mansi Autonomous Area — Yugra Khabarovsk Territory

Republic of Sakha (Yakutia) Bryansk region Novosibirsk region

Republic of Komi Ulyanovsk region Novgorod region

Republic of Karelia Orenburg region Republic of Tyva

Tula region Republic of Bashkortostan Astrakhan region

Republic of Chuvashia Krasnoyarsk Territory Yaroslavl region

Ivanovo region Smolensk region Republic of Buryatia

Astrakhan region Stavropol Territory Lipetsk region

Perm territory Khabarovsk Territory Tula region

Primorsky Territory Amur region Orenburg region

Pskov region Saratov region Perm territory

Republic of Marii El Rostov region Republic of Udmurtia

Novosibirsk region Novosibirsk region Primorsky Territory

Belgorod region Tambov region Nizhny Novgorod region

Orenburg region Sakhalin region Saratov region

Kursk region Kamchatka territory Krasnodar Territory

Irkutsk region Pskov region Samara region

Omsk region Kirov region Jewish Autonomous area

Orel region Nenets Autonomous Okrug Tver region

Kemerovo region Republic of Udmurtia Volgograd region

Ulyanovsk region Perm territory Penza region

Volgograd region Volgograd region Smolensk region

Amur region Murmansk region Orel region

Republic of Bashkortostan Kemerovo region Ryazan region

Bryansk region Republic of Karelia Kostroma region

Saratov region Omsk region Rostov region

Tomsk region Republic of Khakassia Kursk region

Republic of Northen Ossetia — Alania Primorsky Territory Republic of Chuvashia

Rostov region Republic of Chuvashia Vladimir region

Penza region Altay Territory Republic of Marii El

Krasnodar Territory Tomsk region Ulyanovsk region

Altay Territory Republic of Marii El Pskov region

Republic of Khakassia Republic of Sakha (Yakutia) Republic of Ingushetia

Ryazan region Jewish Autonomous area Kurgan region

Voronezh region Astrakhan region Voronezh region

Republic of Altay Republic of Komi Bryansk region

Stavropol Territory Republic of Adygea Kirov region

Tambov region Republic of Northen Ossetia — Alania Republic of Mordovia

End of table on the next page

End of table 3

Employment (from highest to lowest level) Unemployment (from lowest to highest level) Buying power of nominal wage (from highest to lowest)

Jewish Autonomous area Irkutsk region Tambov region

Kurgan region Zabaykalsky Territory Republic of Northen Ossetia — Alania

Republic of Kalmykia Kurgan region Stavropol Territory

Zabaykalsky Territory Republic of Buryatia Republic of Adygea

Republic of Karachaevo-Cherkessia Republic of Altay Altay Territory

Republic of Buryatia Republic of Karachaevo-Cherkessia Republic of Altay

Republic of Adygea Republic of Kabardino-Balkaria Republic of Kabardino-Balkaria

Republic of Dagestan Republic of Dagestan Ivanovo region

Republic of Kabardino-Balkaria Republic of Kalmykia Republic of Karachaevo-Cherkessia

Republic of Tyva Republic of Tyva Republic of Kalmykia

Republic of Ingushetia Republic of Ingushetia Republic of Dagestan

Table 4

List of regions with numbers

Number Region Number Region

1 Belgorod region 41 Republic of Marii El

2 Bryansk region 42 Republic of Mordovia

3 Vladimir region 43 Republic of Tatarstan

4 Voronezh region 44 Republic of Udmurtia

5 Ivanovo region 45 Republic of Chuvashia

6 Kaluga region 46 Perm territory

7 Kostroma region 47 Kirov region

8 Kursk region 48 Nizhny Novgorod region

9 Lipetsk region 49 Orenburg region

10 Orel region 50 Penza region

11 Ryazan region 51 Samara region

12 Smolensk region 52 Saratov region

13 Tambov region 53 Ulyanovsk region

14 Tver region 54 Kurgan region

15 Tula region 55 Sverdlovsk region

16 Yaroslavl region 56 Tumen region

17 Moscow 57 Khanty-Mansi Autonomous Area — Yugra

18 Republic of Karelia 58 Yamal-Nenets autonomous region

19 Republic of Komi 59 Chelyabinsk region

20 Arkhangelsk region 60 Republic of Altay

21 Nenets Autonomous Okrug 61 Republic of Buryatia

22 Vologda region 62 Republic of Tyva

23 Leningrad region 63 Republic of Khakassia

24 Murmansk region 64 Altay Territory

25 Novgorod region 65 Zabaykalsky Territory

26 Pskov region 66 Krasnoyarsk Territory

27 Saint-Petersburg 67 Irkutsk region

28 Republic of Adygea 68 Kemerovo region

29 Republic of Kalmykia 69 Novosibirsk region

30 Krasnodar Territory 70 Omsk region

31 Astrakhan region 71 Tomsk region

32 Volgograd region 72 Republic of Sakha (Yakutia)

33 Rostov region 73 Kamchatka territory

34 Republic of Dagestan 74 Primorsky Territory

35 Republic of Ingushetia 75 Khabarovsk Territory

36 Republic of Kabardino-Balkaria 76 Amur region

End of table 4

Number Region Number Region

37 Republic of Karachaevo-Cherkessia 77 Magadan region

38 Republic of Northen Ossetia — Alania 78 Sakhalin region

39 Stavropol Territory 79 Jewish autonomous area

40 Republic of Bashkortostan 80 Chukotka Autonomous Okrug

research reported here, we decided to test similar hypotheses for three employment groups.

H1: spatial effects for the HH, LL1, and LL2 groups differ;

H2: the determinants of employment for the HH, LL1, and LL2 groups differ.

3.4 The Data, the Dependent and the Explanatory Variables

As said, our sample consisted of 80 regions. Due to data availability, the employment rate was calculated on the population aged between 15 to 72.

To explain existing levels of employment rate (variable empl.) and to test two research hypotheses, three groups of variables were chosen: 1) variables concerning the attractiveness of the region; 2) socio-demographic variables; and 3) variables concerning the industrial structure of the employed population. The first group of variables consisted of four indicators: (i) GRP per capita (variable grppercap_1, thousand rubles in prices of basic year), (ii) population density (variable density, people per square km), (iii) the share of urban population (variable urban share, %), (iv) net migration rate (migration_1, number of migrants per 10 000 population, may be positive or negative). The socio-demographic features of the population consisted of variables characterizing the age structure of the population and the stock of human capital. To illustrate the age structure, the shares of people below and above working age were taken as variables (below and above, %). Working age in Russia is above 16 and below retirement age, which was 60 years for men and 55 for women during the studied period. The stock of human capital was measured as the share of the employed population with a higher education (variable high_educ, %), where 'higher education' means that someone has at least a higher professional education according to the FSSS classification. The sectoral structure is one of the most important features in explaining the regional employment rate. Consequently, we used the Hirschman-Herfindahl Index (variable hh_1) to characterize the industrial structure of each region; and it is used as a measure of the region's degree of sectoral specialization.

HHI = S2 + S22 +... + S2n, (7)

where S:2, S22, ..., Sn2 are shares of people employed in sectors of economic activity (agriculture, construction, wholesale and retail trade, public sector (consisting of education and health services), mining, manufacturing, services). The region is more monopolized if the hh value is closer to 1.

To avoid the problem of endogeneity, we used the first time lag of the variables grp, migration, hh.

We now compare the descriptive statistics of variables for different groups (see Table 4).

There are disparities between HH, LL1 and LL2 groups. For example, almost all mean values for variables that characterize the attractiveness of the region are higher in the HH group.

4. Methodology of the Econometric Modeling

To test the two main research hypotheses, we used the following modification of the SAR (Spatial Auto Regression) model:

iv Л

= t

VY'L2 St

+Pl

v YL2

V lL2 St-1

+ Ph

WY;1

(

+ Pl

Xh 0

0

(

Ьн +

t 0 X

0 0

0 0

WY.

i L1 0

ßL1 +

a,.,

V L2 S

iL 2 S 0 0

VXiL2 St

\

ЛШ

V L2 St

(8)

where Y is employment in group 15-72, WY is spatial lag, X is a matrix of explanatory variables, a is a vector of fixed effects, u is a vector of errors (we split them in three parts), c is a vector of time effects (set of dummy variables for 20072016 years).

Let us formulate our hypotheses in a form convenient for empirical verification:

Hypothesis 1. There are no differences of spatial effects in regional groups.

Alternative hypothesis 1. There are differences of spatial effects in regional groups.

+ ct +

Table 5

Descriptive statistics for the variables

Variable Mean Std. Dev. Min Max Observations

All Russia

empl overall 62.528 5.955 16.500 81.200 N = 960

between 5.505 34.475 78.783 n = 80

within 2.345 44.553 78.853 T = 12

wcbempl overall 62.939 3.777 36.439 78.836 N = 960

between 3.386 46.110 76.357 n = 80

within 1.713 53.267 73.657 T = 12

widempl overall 62.138 2.427 41.332 65.947 N = 960

between 1.836 49.799 64.295 n = 80

within 1.600 53.671 70.136 T = 12

grppercap_1 overall 96963.860 117139.900 2133.539 1035858.000 N = 960

between 101129.200 11439.280 589271.300 n = 80

within 60099.540 -387423.000 543550.500 T = 12

urbanshare overall 0.694 0.126 0.260 1.000 N = 960

between 0.126 0.278 1.000 n = 80

within 0.011 0.653 0.738 T = 12

below overall 17.701 3.548 12.100 34.400 N = 960

between 3.450 13.025 31.158 n = 80

within 0.907 15.243 20.943 T = 12

above overall 21.397 4.931 5.500 30.200 N = 960

between 4.638 8.258 28.125 n = 80

within 1.747 16.905 27.805 T = 12

migration_1 overall -9.754 54.749 -499.000 197.000 N = 960

between 43.594 -142.333 107.500 n = 80

within 33.449 -492.338 172.079 T = 12

density overall 73.322 387.276 0.069 3752.572 N = 960

between 388.906 0.070 3482.613 n = 80

within 21.661 -146.362 343.280 T = 12

high_educ overall 26.654 5.678 12.500 50.000 N = 960

between 4.443 18.925 47.758 n = 80

within 3.567 14.671 40.521 T = 12

hh_1 overall 0.293 0.060 0.204 0.635 N = 960

between 0.056 0.215 0.557 n = 80

within 0.023 0.196 0.560 T = 12

HH group

empl overall 65.681 3.855 57.700 81.200 N = 528

between 3.417 61.800 78.783 n = 44

within 1.853 58.898 70.748 T = 12

wcbempl overall 64.648 2.691 58.622 78.836 N = 528

between 2.341 60.473 76.357 n = 44

within 1.369 60.239 67.666 T = 12

widempl overall 62.897 1.521 59.289 65.947 N = 528

between 0.598 61.473 64.295 n = 44

within 1.401 59.961 65.113 T = 12

grppercap_1 overall 124489.100 146080.500 2133.539 1035858.000 N = 528

between 125809.100 34761.040 589271.300 n = 44

within 76432.750 -359897.800 571075.700 T = 12

urbanshare overall 0.755 0.095 0.573 1.000 N = 528

between 0.096 0.588 1.000 n = 44

within 0.011 0.715 0.800 T = 12

below overall 16.929 2.629 12.100 24.800 N = 528

between 2.501 13.025 23.858 n = 44

within 0.887 15.496 19.288 T = 12

Continued table 5

Variable Mean Std. Dev. Min Max Observations

above overall 21.580 5.011 5.500 30.200 N = 528

between 4.704 8.258 28.125 n = 44

within 1.855 17.089 27.989 T = 12

migration_1 overall -7.242 57.181 -223.000 197.000 N = 528

between 48.814 -142.333 107.500 n = 44

within 30.604 -190.325 174.591 T = 12

density overall 107.836 519.274 0.069 3752.572 N = 528

between 523.951 0.070 3482.613 n = 44

within 29.178 -111.848 377.794 T = 12

high_educ overall 26.414 6.174 12.500 50.000 N = 528

between 5.142 19.035 47.758 n = 44

within 3.497 14.431 35.189 T = 12

hh_1 overall 0.291 0.069 0.205 0.635 N = 528

between 0.064 0.215 0.557 n = 44

within 0.028 0.193 0.558 T = 12

LL1 group

empl overall 58.474 6.710 16.500 67.300 N = 264

between 5.966 34.475 62.725 n = 22

within 3.306 40.499 74.799 T = 12

wcbempl overall 60.236 4.420 36.439 67.223 N = 264

between 3.800 46.110 64.996 n = 22

within 2.387 50.565 70.954 T = 12

widempl overall 60.688 3.500 41.332 65.159 N = 264

between 2.887 49.799 63.220 n = 22

within 2.064 52.221 68.686 T = 12

grppercap_1 overall 59339.110 43394.850 3310.776 291518.000 N = 264

between 35287.700 11439.280 161948.200 n = 22

within 26267.330 -33631.360 188908.800 T = 12

urbanshare overall 0.608 0.106 0.384 0.767 N = 264

between 0.108 0.413 0.760 n = 22

within 0.011 0.580 0.644 T = 12

below overall 17.756 4.102 13.600 33.400 N = 264

between 4.127 14.233 30.425 n = 22

within 0.713 15.431 20.731 T = 12

above overall 22.358 4.960 7.600 29.900 N = 264

between 4.839 9.408 27.592 n = 22

within 1.472 19.533 26.474 T = 12

migration_1 overall -2.585 55.750 -499.000 148.000 N = 264

between 37.061 -86.667 75.833 n = 22

within 42.332 -485.169 161.832 T = 12

density overall 45.202 26.455 3.726 143.530 N = 264

between 26.944 3.803 130.743 n = 22

within 2.066 29.739 57.989 T = 12

high_educ overall 28.120 4.833 16.600 46.000 N = 264

between 3.054 23.233 36.417 n = 22

within 3.797 17.461 41.986 T = 12

hh_1 overall 0.308 0.045 0.238 0.456 N = 264

between 0.043 0.253 0.433 n = 22

within 0.016 0.268 0.406 T = 12

LL2 group

empl overall 58.990 3.977 45.800 65.900 N = 168

between 3.625 48.225 62.917 n = 14

within 1.882 52.032 62.932 T = 12

End of table on the next page

1394 социально-демографический потенциал регионального развития

End of table 5

Variable Mean Std. Dev. Min Max Observations

wcbempl overall 61.815 2.500 55.573 67.400 N = 168

between 2.143 57.311 64.955 n = 14

within 1.399 57.884 64.857 T = 12

widempl overall 62.032 1.441 59.104 65.172 N = 168

between 0.559 61.020 63.062 n = 14

within 1.336 59.455 64.142 T = 12

grppercap_1 overall 69580.570 55707.250 5249.425 309864.000 N = 168

between 44904.770 21177.230 170118.100 n = 14

within 34923.890 -35537.610 209326.500 T = 12

urbanshare overall 0.637 0.135 0.260 0.858 N = 168

between 0.139 0.278 0.853 n = 14

within 0.013 0.611 0.663 T = 12

below overall 20.042 4.072 15.400 34.400 N = 168

between 4.026 16.117 31.158 n = 14

within 1.200 17.584 23.284 T = 12

above overall 19.310 3.960 9.200 28.500 N = 168

between 3.648 9.958 24.833 n = 14

within 1.803 16.377 22.977 T = 12

migration_1 overall -28.917 39.109 -159.000 79.000 N = 168

between 30.738 -80.000 28.917 n = 14

within 25.435 -126.167 46.833 T = 12

density overall 9.036 9.080 0.750 29.660 N = 168

between 9.332 1.852 28.982 n = 14

within 1.047 0.628 10.968 T = 12

high_educ overall 25.107 4.697 14.500 36.200 N = 168

between 3.320 18.925 30.842 n = 14

within 3.430 15.490 32.707 T = 12

hh_1 overall 0.277 0.043 0.204 0.371 N = 168

between 0.042 0.217 0.352 n = 14

within 0.015 0.247 0.322 T = 12

Formal main and alternative hypotheses 1:

H0 : Ph = Pu = PL2' H1 : Ph * Pli or Ph * Pl2.

Hypothesis 2. There are no differences in the influence of the factors on employment rates in the regions belonging to different regional clubs.

Alternative hypothesis 2. There are differences in the influence of the factors on employment rates in the regions belonging to different regional clubs.

Formal main and alternative hypotheses 2:

H0 : Ph = PlI = Pl2>

H1 : Ph * PL1 or Ph * PL2.

The split spatial lags in our model were endogenous. To resolve the problem of endogene-ity difference, we adopted the GMM ([27] Arellano and Bond, 1991) method of estimation. However, application of this method to our initial specification (with all explanatory variables, divided into several parts) required a number of instruments much larger than the number of regions.

According to Roodman (2009), this leads to a bias in the parameter estimation. To avoid this problem we had to use the Arellano-Bond approach for the estimation and drastically restrict the number of instruments. Moreover, in order to consider the possible bias in the parameters' estimation at a small time interval but with a large number of observation units, we adopted a GMM modification for models with fixed effects, similarly to [28] Lee and Yu (2010). In addition, a large number of variables may also lead to the problem of multicollin-earity of the data. To increase the efficiency of the estimates, we removed groups of insignificant variables from the model one by one (after a preliminary test of the corresponding statistical hypotheses). The technique that we used was an extension of the conventional backward stepwise method. To test the robustness of the result of the estimation, we re-estimated our model with an inverted distance weighted matrix instead of the matrix of common borders. The results of the estimation are presented in the next section.

5. TheResults of the Estimation

In this section, we present the final results of our main estimations (Table 6). For each model, we also present the results of post-estimation procedures. Estimates of the coefficients obtained by the Arellano-Bond method are consistent under the following conditions ([29] Greene, 2012, p. 400): 1) errors uit must be serially uncorrelated; 2) moment conditions (consisting in the orthogonality of the errors and instruments) must be correct. The Arellano-Bond approach, using the equations in difference, makes it possible to avoid the endogeneity problem with the elimination of individual effects. It is for this reason that the errors in the difference equation must be identified as first-order autocorrelations and not revealed as higher-order autocorrelations (Arellano and Bond test). The second condition is verified by the Sargan test of instruments' validity. These two conditions are verified for all the models estimated.

Spatial effects for the three employment groups were different. For the LL2 group of regions, the spatial effects were insignificant; hence employment in the regions of the LL2 group does not depend on the local labor markets of other regions. For the group of HH regions, only spatial effects for the inverse distance weighting matrix are significant. Consequently, each region of the HH group is affected by the rest of Russia's regions, and the extent of this influence decreases with the increase in geographical distance between regions. Employment in the HH region group varies according to the general situation in the country.

The most interesting spatial effects were revealed for the regions in the LL1 group. For a common boundary weighting matrix, they were negative, and for a weighting matrix of inverted distances, they were positive. Thus, in the LL1 group of regions (in the south of Russia), there is a mechanism of competition for labor resources with neighboring regions. If in one of the southern regions the situation on the labor market improves, then it draws labor resources from neighboring regions. At the same time, if the overall employment situation in Russia improves (or worsens), then similar changes occur in the LL1 group of regions.

In regard to the impact of the explanatory variables, the "group effect" was found for the variables share of urban population, net migration rate, shares of people below and above working age, share of people with higher education.

The influences of GRP per capita and on the level of employment was insignificant.

For LL2 and HH groups of regions, employment was not dependent on the share of the urban population. This can be explained by the presence of

Table 6

The results of estimation

Dependent Variable empl modelwcb modelwid

Weighted matrix Matrix of common borders Matrix of inverted distance

empl

L1. (time lag) 0.562*** 0.553***

Spatial lags

wcbempLl —0.l84***

wcbempL2 0.l05

wcbemplH 0.l25

widempLl 0.l0l*

widempL2 0.234

widemplH 0.39l***

urbanshareLl —5l.l86*** —29.2l6***

belowLl —l.0l7*** —0.785***

belowL2 —0.4l5** —0.l39

belowH —0.373** —0.l54

aboveLl 0.82l*** 0.8l3***

migr_1L1 0.006*** 0.007***

migr_lL2 0.006* 0.004

migr_lH —0.007*** —0.006***

dens —0.004*** —0.003*

high_edL2 0.l29** 0.l47**

hh_l 5.0l7* 4.857*

d2007 0.206 0.239

d2008 —0.l64 —0.332

d2009 —l.l88*** —l.l38***

d20l0 0.l75 —0.062

d20ll 0.583 0.l22

d20l2 l.237*** 0.338

d20l3 0.698 —0.235

d20l4 l.272** 0.095

d20l5 0.992* —0.206

d20l6 l.30l** —0.098

_cons 35.l03*** ll.964*

Number of instruments 62 62

p-v AB test for zero autocorrelation

in first-differenced errors

order l 0.000 0.000

order 2 0.l33 0.l2

p-v Sargan test 0.477 0.254

* p < 0.05; ** p < 0.01; *** p < 0.001.

two opposite tendencies: in a city, it is usually easier to find a job; but for monotowns, of which there are more than 300 in Russia when the city-forming enterprise closes, the situation changes to the op-

1396 COm/IA-nbHO-flEMOrPAOMHECKMM nOTEH^Afl PErMOHAflbHOrO PA3BMTMJI

posite.1 The increase in the share of urban population reduced employment in the Low-Low1 group. This can be explained by the fact that a large proportion of businesses in the southern regions are engaged in agriculture.

With an increase in the population aged under 16, the level of employment was reduced according to both estimated models only in the LL1 group of regions. This can be explained by the fact that in the North Caucasus regions that belong to this group, the share of young people is very high, and for them, it is difficult to find jobs. Conversely, increasing the proportion of the population over working age increases employment in the LL1 group, because in these regions there are relatively low wages; therefore, on reaching retirement age, people often leave the labor market, preferring informal employment and work on personal plots of land.

The influx of migrants stimulates employment in the LL1 group of regions and decreases employment in the group of HH regions. This may be because more educated migrants able to compete for well-paid jobs go mainly to the regions of the HH group.

An increase in the proportion of the population with higher education stimulates employment only in the LL 2 group of regions.

Group effects were not found for the variables density and Herfindahl-Hirschman index.

The negative dependence of employment on population density can be explained 1) by competition for jobs, 2) by the problem of mono-towns mentioned above, 3) by the favorable situation in the labor markets of some sparsely populated northern regions. Perhaps, for this variable, a functional dependence more flexible than the linear one should be used.

A higher Herfindahl-Hirschman index is associated with higher employment. The higher the Herfindahl-Hirschman index, the more the level of specialization in the given region. Thus, in 20052016 in Russia Marshallian effects prevailed.

The results of the estimation also make it possible to draw a conclusion about the negative impact of the 2008-2009 crisis on the level of employment, whereas the crisis of 2014 did not affect it. This is probably because the 2008-2009 crisis was global, while the crisis in 2014 had a local character with effects more delayed and distributed over time.

6. Conclusions

1 On the topic of monotowns and the political economy of industrial restructuring, see [30] Crowley (2016).

In this paper, we have investigated the spatial effects for the regional employment groups in Russia and the differences in the impact factors that explain regional employment. The key results are the following: (1) boundary spatial effects for three groups are different; (2) the differing influence for selected groups of regions was apparent for the share of urban population, net migration rate, the shares of people below and above working age, the share of people with higher education; (3) the influence of GRP is insignificant.

According to our results, these are several (general and specific) policy implications. From a general perspective, (national and regional) economic and labor policies must take the specificity of each group of regions into account; in fact, the same policy measures can generate significantly different consequences in HH, LL 1 and LL 2 groups of regions; moreover, policies should consider that the (positive) impact on a group of regions can indirectly (negatively or positively) affect other groups of regions. As regards the specific policy implications, the following aspects seem most important: (i) policies favoring a higher level of specialization are suggested due to the positive employment effects in all groups of regions; (ii) an increase of the higher educated proportion of the labor force is recommended because it would increase employment in the South of Siberia and in Zabaikalye; (iii) the increase in migration flows favorably affects the employment of only the LL1 group of regions, where the agricultural sector is well developed; therefore, it is desirable to create favorable conditions for migrants in the South of Russia, while in more densely populated regions of Russia, migrants can compete for jobs with indigenous people, worsening both wage and working conditions; (iv) national and regional economic, educational and labor policies could focus especially on improving the dramatic employment quantity and quality in the group of regions LL 1, especially considering the supply-demand mismatches and the difficulties of transition — from education and from unemployment — to employment, also reducing the problem of competing for resources with neighboring regions.

These results are consistent with the findings of the report based on [26] that federal policy in the labor market in Russia should be built with "possible consideration of the heterogeneity of the regions", there are no "simple and quick solutions" to smoothing the differentiation of Russian regions, this problem requires a "strategic and integrated approach."

References

1. Oschepkov, A. & Kapelyushnikov, R. (2015). Regionalnyye rynki truda: 15 let razlichiy [Regional labor markets: 15 years of differences]. Higher School of Economics. WP3 series "Problems of the labor market". (In Russ.)

2. Caroleo, F .E. & Pastore, F. (Eds.) (2010). The labor market impact of the EU enlargement. Berlin: Springer, 342. doi. org/10.1007/978-3-7908-2164-2_2.

3. Mussida, C. & Pastore, F. (Eds.) (2015). Geographical Labor Market Imbalances. AIEL Series in Labor Economics, Berlin and Heidelberg, Springer, 370. Retrieved from: https://econpapers.repec.org/RePEc:ail:labook:08.

4. Dolton, P., Bondibene, C. R. & Stops, M. (2015). Identifying the employment effect of invoking and changing the minimum wage: A spatial analysis of the UK. Labor Economics, 37, 54-76. doi.org/10.1016/j.labeco.2015.09.002. Retrieved from: https:// s100.copyright.com/AppDispatchServlet?publisherName=ELS&contentID=S0927537115001062&orderBeanReset=true.

5. Vega, S. H. & Elhorst, J. P. (2016). A regional unemployment model simultaneously accounting for serial dynamics, spatial dependence and common factors. Regional Science and Urban Economics, 60, 85-95. doi.org/10.1016/j.regsciurbeco.2016.07.002. Retrieved from: https://s100.copyright.com/ AppDispatchServlet?publisherName=ELS&contentID=S0166046216300862&orderBeanReset=true.

6. Manning, A. & Petrongolo, B. (2017). How local are labor markets? Evidence from a spatial job search model. American Economic Review, 107(10), 2877-2907. doi: 10.1257/aer.20131026.

7. Head, K. & Mayer, T. (2006). Regional wage and employment responses to market potential in the EU. Regional Science and Urban Economics, 36(5), 573-594. doi.org/10.1016/j.regsciurbeco.2006.06.002.

8. Ketterer, T. D. & Rodriguez-Pose, A. (2016). Institutions vs.'first-nature'geography: What drives economic growth in Europe's regions? Papers in Regional Science. doi 10.1111/pirs.12237.

9. Huber, P. (2007). Regional labor market developments in transition: A survey of the empirical literature. The European Journal of Comparative Economics, 4(2), 263-298.

10. Bah, E. & Brada, J. (2014). Labor Markets in the Transition Economies: An Overview. The European Journal of Comparative Economics, 11(1), 3-53.

11. Vakulenko, E. S. & Gurvich, E. T. (2016). Gibkost realnoy zarabotannoy platy v Rossii: sravnitelnyy analiz [Real Wage Flexibility in Russia: Comparative Analysis]. Zhurnal novoy ekonomicheskoy assotsiatsii [Journal of the New Economic Association], 3(31), 67-92. (In Russ.)

12. Kapelyushnikov, R., Kuznetsov, A., & Kuznetsova, O. (2012). The role of the informal sector, flexible working time and pay in the Russian labor market model. Post-communist economies, 24(2), 177-190. doi.org/10.1080/14631377.2012.6 75154.

13. Solanko, L. (2008). Unequal fortunes: a note on income convergence across Russian regions. Post-Communist Economies, 20(3), 287-301. https://doi.org/10.1080/14631370802281399.

14. Ledyaeva, S., & Linden, M. (2008). Determinants of Economic Growth: Empirical Evidence from Russian Regions. European Journal of Comparative Economics, 5(1), 87-105.

15. Kholodilin, K. A., Oshchepkov, A. & Siliverstovs, B. (2012). The Russian regional convergence process: Where is it leading? Eastern European Economics, 50(3), 5-26. doi.org/10.2753/EEE0012-8775500301.

16. Akhmedjonov, A., Lau, M. C. K., & izgi, B. B. (2013). New evidence of regional income divergence in post-reform Russia. Applied Economics, 45(18), 2675-2682. doi.org/10.1080/00036846.2012.665600.

17. Lehmann, H. & Silvagni, M. G. (2013). Is There Convergence of Russia's Regions? Exploring the Empirical Evidence: 1995-2010. IZA Discussion Papers, 7603. Retrieved from: http://dx.doi.org/10.2139/ssrn.2321098 (date of access: 02.10.2018).

18. Dolinskaya, I. (2002). Transition and Regional Inequality in Russia: Reorganization or Procrastination? IMF Working Paper, 2, 169.

19. Demidova, O. & Signorelli, M. (2012). Determinants of Youth Unemployment in Russian Regions. Post-Communist Economies, 2, 191-217. doi.org/10.1080/14631377.2012.675155.

20. Demidova, O., Marelli, E. & Signorelli, M. (2013). Spatial Effects on Youth Unemployment Rate: The Case of Eastern and Western Russian Regions. Eastern European Economics, 5, 94-124. doi.org/10.2753/EEE0012-8775510504.

21. Demidova, O., Marelli, E. & Signorelli, M. (2015). Youth Labor Market Performance in the Russian and Italian Region. Economic Systems, 39(1), 43-58. doi.org/10.1016/j.ecosys.2014.06.003.

22. Blinova, T., Markov, V. & Rusanovskiy, V. (2015). Youth unemployment in Russia: Models of interregional differentiation. Regional Formation and Development Studies, 15(1), 7-18. http://dx.doi.org/10.15181/rfds.v15i1.975.

23. Blinova, T., Markov, V., & Rusanovskiy, V. (2016). Empirical study of spatial differentiation of youth unemployment in Russia. Acta Oeconomica, 66(3), 507-526. doi.org/10.1556/032.2016.66.3.7.

24. Rusanovskiy, V. & Markov, V. (2016). Youth unemployment in Russian Regions and assessment of the economic loss. Indian Journal of Science and Technology, 9, 30. Retrieved from: http://www.indjst.org/index.php/indjst/article/view/98754 (date of access: 02.10.2018).

25. Danilenko, T., Demidova, O. & Signorelli, M. (2017). Unemployment Clubs in Russian Regions. Emerging Markets Finance and Trade, 54(6), 1337-1357. https://doi.org/10.1080/1540496X.2017.1281799.

26. Gimpelson, V. E., Zudina, A. A., Kapelyushnikov, R. I., Lukyanova, A. L., Oschepkov, A. Yu., Roshchin, S. Yu., Smirnykh, L. I., Travkin, P. V. & Sharunina, A. B. (2017). The Russian labor market: trends, institutions, structural changes. In: Gimpelson, V. E., Kapelyushnikov, R. I., Roshchin S. Yu. (Eds). Moscow: Center for Strategic Research. Retrieved from: https://lirt.hse.ru/data/2017/03/21/1170068107/Doklad_trud.pdf (date of access: 02.10.2018). (In Russ.)

27. Arellano, M. & Bond, S. (1991). Reviewed Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. The Review of Economic Studies, 58(2), 277-297.

28. Lee, L. F. & Yu, J. (2010). Estimation of spatial autoregressive panel data models with fixed effects. Journal of Econometrics, 154(2), 165-185. doi.org/10.1016/j.jeconom.2009.08.001https://s100.copyright.com/ AppDispatchServlet?publisherName=ELS&contentID=S030440760900178X&orderBeanReset=true.

29. Greene, W. H. (2012). Econometric analysis, 7th ed. Upper Saddle River, NJ: Prentice Hall. 1188.

30. Crowley, S. (2016) Monotowns and the political economy of industrial restructuring in Russia. Post-Soviet Affairs, 32:5, 397-422. doi.org/10.1080/1060586X.2015.1054103.

Authors

Olga Anatolyevna Demidova — PhD in Mathematics, Associate Professor, National Research University Higher School of Economics; Author Scopus ID: 6602926736 (26, Shabolovka St., Moscow, 119049, Russian Federation; e-mail: demidova@ hse.ru).

Pierluigi Daddi — Professor, Department of Economics, University of Perugia; Author Scopus ID: 55832980300 (Via A. Pascoli, 20, 06123, Perugia, Italy; e-mail: pierluigi.daddi@libero.it).

Ekaterina Vladimirovna Medvedeva — Leading Specialist, PJSC "Post Bank" (32/1, Kutuzovsky Ave., Moscow, 121165, Russian Federation; e-mail: medvedevaekvl@yandex.ru).

Marcello Signorelli — Professor, Department of Economics, University of Perugia; Author Scopus ID: 24774145900 (Via A. Pascoli, 20 06123, Perugia, Italy; e-mail: marcello.signorelli@unipg.it).