Chapter 4. Encouraging mobility and entrepreneurship in Latvia’s regions

Active labour market policy in Latvia faces particular challenges in some regions outside the metropolitan area of Riga. This chapter documents differences between Latvia’s regional labour markets before focusing on two aspects: the regional mobility of unemployed persons and the role of entrepreneurship in reducing regional unemployment. In this context, the chapter assesses ALMP programmes that foster mobility between regions and start-ups from unemployment.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.

Regional labour market differences

Several indicators confirm that the divide between regions in Latvia is one of the strongest of all OECD countries (Figure 4.1): only three OECD countries exhibit higher interregional inequality than Latvia, according to a Gini index from 2013. Similarly, only three OECD countries generate a greater share of their GDP in urban areas than Latvia does. Albeit less extreme, the share of the population living in urban areas is also relatively high in Latvia. Latvia’s values on all three indicators exceed those for Estonia and Lithuania. Against this background, analyses of the Latvian labour market should take the regional dimension into account.

Figure 4.1. Indicators for the urban-rural divide in OECD countries, 2013/2014

The divide between regions is primarily driven by a strong concentration of population and economic activity in Riga, the country’s capital, as well as the immediately surrounding area called Pieriga. Riga’s 640 000 inhabitants make it the largest city in the Baltic states and account for more than one-third of Latvia’s entire population. This metropolis is situated near the geographical centre of the country, roughly where three more rural regions meet: Kurzeme in the west, Vidzeme in the north-east and Zemgale in the south. By contrast, the region of Latgale is situated more remotely in the east of the country, along Latvia’s borders with Belarus and the Russian Federation. Daugavpils in Latgale is Latvia’s second largest city, counting 83 000 inhabitants. Following settlements at the time of the Soviet Union, Latvia has a strong minority of ethnic Russians, living primarily in Latgale and Riga (Box 4.1).

In OECD regional statistics, Riga and Pieriga are classified as predominantly urban regions. Vidzeme and Zemgale are classified as predominantly rural, while Kurzeme and Latgale are considered intermediate. For example, this grouping largely aligns with regional population density in 2018, measured as population per square kilometre: with 12 persons per square kilometre, Vidzeme has the lowest population density among Latvian regions, followed by Kurzeme (18), Latgale (18) and Zemgale (22). However, population density is considerably higher in Pieriga (36) and Riga stands out as a dense urban area (2 100 persons per square kilometre), according to Latvia’s Central Statistical Bureau.

Different labour market conditions in regions imply different needs for ALMP

Latvia’s regions also differ markedly in terms of labour market conditions. Figure 1.19 in Chapter 1 shows that unemployment rates, youth unemployment rates and shares of long-term unemployed all varied substantially across regions in 2016. The greatest difference in regional unemployment rates – reaching 11 percentage points – occurred between Pieriga with an unemployment rate of 6.6% and Latgale with a rate of 17.9%. According to Figure 4.2, this range appears large in comparison with ranges observed between the largest subnational units in other OECD countries: greater ranges are only observed in Turkey (23 percentage points), Italy (19), Spain (16) and Greece (13). However, the largest subnational units in Latvia comprise significantly less population than in most other OECD countries, so that these units may not be directly comparable. In OECD statistics on regions, all of Latvia is considered comparable to large regions (TL2 level) in most OECD countries; the same applies to Estonia and Luxembourg (see Table A.1 in OECD (2018[2]) for details). Since smaller regions allow for greater variation in unemployment rates, the range between Latvia’s regions (T3 level) will seem relatively smaller when compared to regions of this size in other OECD countries.

Figure 4.2. Regional differences in unemployment rates, 2017 or latest available year

Data from Latvia’s Central Statistics Bureau on registered vacancies highlight that labour market conditions were much more favourable in Riga in 2016 than in other regions. The vacancy rate (vacancies as a share of total employment) at the end of the fourth quarter of 2016 reached 2%, well above the national average (1.6%) and the second highest vacancy rate, observed in Pieriga (1.5%). The other regions exhibited a vacancy rate around 1%, except for 0.7% in Latgale. Riga stands out most in terms of labour market tightness, i.e. the ratio between registered vacancies (measured as yearly average of levels at the end of the quarter) and registered unemployed persons (annual figure for ages 15-64). In 2016, labour market tightness approached 0.4 in Riga, so that the number of vacancies corresponded to almost 40% of the number of unemployed persons. This level of tightness was more than twice as high as in Pieriga (0.17) and the value for all of Latvia (0.15). By contrast, labour market tightness in Vidzeme, Kurzeme and Zemgale was around 0.06, while it was particularly low in Latgale (0.03). These values for labour market tightness underline that labour market conditions differ more strongly between Riga and Latgale than unemployment rates would suggest: not only is the unemployment rate substantially lower in Riga than in Latgale, but also the vacancy rate is higher.

As Riga is a dense urban area, this suggests that the degree of urbanisation may be a key driver of differences between Latvia’s regional labour markets. Figure 4.3 shows how a number of relevant groups for labour market policy are distributed over cities, towns and rural areas. Compared to the distribution of employed persons, disproportionately many unemployed persons, GMI benefit recipients and discouraged workers live in rural areas: while 39% of all employed persons lived in rural areas in 2016, this applied to 45% of registered unemployed persons, 51% of GMI benefit recipients and 56% of discouraged workers. Towns and suburbs accounted for 20% of employment but 25% of long-term unemployment and almost one-third of GMI benefit recipients. Cities and large urban areas, by contrast, accounted for 45% of employment but only 32% of registered unemployed and 17% of GMI benefit recipients.

Latvia’s State Employment Agency (SEA) delivers labour market policy through 26 branches and 18 smaller offices across all regions (as of November 2018). This regional presence raises the possibility to adapt labour market policy to the regional or local situation. While the SEA centrally defines overarching objectives for its services, such as increasing the coverage of a particular group, regional figures on unemployment and the size of specific groups serves as indicators against which the success of particular branches can be measured. Based on reviews of several public employment services, the European Commission (2013[3]) identified principles for combining centralised decision-making with adaptation to the local context. Firstly, some degree of autonomy for local offices can raise their level of engagement and make services more targeted to local challenges. Secondly, centralised decisions ensure that targets are ambitious and limit the time spent on the target-setting process. While local offices therefore do not necessarily need to influence this progress, setting different local targets can account for different local situations. Thirdly, specific local challenges may be addressed through a small number of additional targets for certain local offices.

Figure 4.3. Relevant groups for labour market policy by degree of urbanisation in Latvia

Box 4.1. Unemployment, ethnicity and language

A considerable share of Latvia’s population is considered to have a non-Latvian ethnicity. According to figures for 2018 from Latvia’s Central Statistical Bureau, ethnic Russians make up one-quarter of the total population and Belarussians, Poles and Ukrainians account for 2-3% each. The ethnic minorities concentrate in certain regions: based on the same figures, ethnic Russians account for 37% of the population in both Riga and Latgale, while their share is below 20% in all other regions. Belarussians and Poles are also most frequent in Latgale, where they account for 5% and 7%, respectively.

According to Latvia’s 2011 Census, the number of Russian speakers in Latvia is larger than the number of ethnic Russians, and Russian speakers similarly concentrate in Riga and Latgale (see Box 1.1 and Figure 1.8 in OECD, (2016[4])). With both Latvian and Russian being widely spoken languages in Latvia, lack of proficiency in either of these languages has been identified as an important barrier to employment (see Hazans (2010[5]), for example). This barrier mostly affects Russian speakers because Latvian is the majority language. However, Lindemann (2014[6]) emphasises for the similar case of Estonia that job prospects of Russian speakers are less affected in regions where Russian is also widely spoken. Results obtained by Toomet (2011[7]) suggest that ethnic Russian men in Latvia and Estonia hardly earn a wage premium from proficiency in Latvian and Estonian, respectively.

While recent data on unemployment by language are not available, Figure 4.4 shows that unemployment rates and the share of long-term unemployed in 2016 were higher for non-Latvian ethnicities than for ethnic Latvians in all regions. At the same time, differences remained limited, with some notable exceptions in Riga and Latgale – the two regions in whose high numbers of Russian speakers make the language barrier especially likely to arise. First, the share of long-term unemployed in Riga is considerably higher for non-Latvian ethnicities than for ethnic Latvians (31% compared with 21%). Second, the youth unemployment rate of non-Latvian ethnicities in Latgale is considerably higher than that of ethnic Latvians (39% compared with 27%). Although it is likely that knowledge of the Latvian language is better among youth, Zvaigzne, Saulāja and Čerpinska (2015[8]) argue that the language barrier also contributes to youth unemployment in Latgale, alongside mismatch between the quality of local jobs and the job quality sought by youth.

Figure 4.4. Regional unemployment of Latvians and non-Latvians, 2014-2016

The language barrier is not the only plausible explanation for differences in unemployment rates between ethnicities. To some extent, higher unemployment rates for non-Latvian ethnicities might reflect the legacy of the economic crisis 2008/2009: the crisis reduced employment of non-Latvians more strongly than employment of Latvians (Hazans (2010[5]), Masso and Krillo (2011[9])). However, Figure 4.5 suggests that, by 2014-2016, these effects did not drive the observed differences in unemployment rates anymore. According to Figure 1.8 in Chapter 1, the following sectors were heavily affected by the economic crisis: agriculture, construction, manufacturing, public administration or defence as well as trade, food and accommodation. In most Latvian regions including Latgale, the shares of unemployed persons who had previously worked in these sectors did not differ much between Latvians and non-Latvians in 2014-2016. The difference was significant only in Riga, where this share was 9 percentage points higher for non-Latvians, as well as in Pieriga, where it was 16 percentage points lower for non-Latvians. When applied to the earlier period 2012-2014, the same analysis confirms notably for Latgale that a substantially higher share of unemployed non-Latvians (66%) than of unemployed Latvians (53%) had been working in heavily affected sectors.

Figure 4.5. Unemployed Latvians and non-Latvians by previous sector of activity, 2014-2016

Ethnic Russians in both Riga and Latgale probably also had comparatively strong trade links with the nearby Russian Federation. While Latvia maintained its fixed exchange rate to the euro during the crisis, the Russian currency lost value, so that goods and services from Latvia would become less competitive in the Russian Federation. This asymmetric loss of competitiveness might thus have reduced employment more among ethnic Russians. For Latgale, there is the additional factor that the distance to Riga is larger than from other Latvian regions, which likely limits access to a range of employment opportunities and thereby contributes to unemployment (Rogers, 1997[10]).

Mobility of the unemployed between Latvian regions

The substantial differences between regional labour market conditions in Latvia create incentives for mobility. According to Hazans (2003[11]), net migration flows between Latvian regions are directed to regions with higher wages and higher population density, and the probability of moving to another region is greater for highly educated and younger persons. Based on qualitative evidence from rural areas of Latvia, Bell et al. (2009[12]) conclude that employment opportunities but also the availability of services determine the likelihood that residents stay. The observed long-run trends in migration flows between Latvian regions highlight the attractiveness of urban and suburban areas, especially the surroundings of Riga but also those of regional centres (Figure 4.6). Central Riga is an exception, which will be discussed below.

Figure 4.6. Internal migration trends in Latvia, 2000-2018

In addition to moves between regions, some analyses have considered commuting in Latvia. A study by Hazans (2004[13]) finds commuting patterns in Latvia are mainly directed to Riga, while residents of Riga or other major cities hardly commute. Results further indicate that highly educated persons and those aged 20-34 are especially likely to commute to work, and that regional unemployment mattered less for commuting than the distance to Riga. Paci et al. (2007[14]) confirm the results for the highly educated and for residents of Riga, but also find that high local unemployment rates are associated with a lower probability for residents to commute. However, low probability of commuting from high-unemployment regions might be explained by a greater distance to Riga, a variable that is omitted in this analysis of commuting. Plausible results were obtained for local levels of GDP per capita: Latvians seem to commute from regions with low GDP per capita to regions with high GDP per capita, and rarely in the reverse direction.

Several studies found that regional mobility was low in Central and Eastern European countries during the post-1990 transition years (see for example Boeri and Scarpetta (1996[15]), Burda and Profit (1996[16])). Within this group of countries, however, Latvia exhibited a relatively high regional mobility, according to Hazans (2003[11]). Figure 4.7 offers new estimates of regional mobility rates in a number of European OECD countries. These estimates suggest that regional mobility in Latvia reaches an intermediate level: 1.3% of the population of working age (15-64) lived in a different region than one year before the survey. This level is well above the estimated regional mobility in Poland (0.1%), the Czech Republic (0.4%) and Hungary (0.6%), but well below that in Finland (1.8%), Denmark (1.9%) and Sweden (2%). With few exceptions, the regional mobility of unemployed persons (i.e. recorded as unemployed one year before the survey) seems to exceed average mobility substantially. The estimate for unemployed persons in Latvia (1.4%) again falls into the middle of the range, while the mobility of unemployed persons in Finland, Denmark and Sweden appears at least twice as high (between 3.0% and 3.3%).

Figure 4.7. Estimated regional mobility rates for unemployed persons in selected European OECD countries, 2013-2016

Two caveats arise with the estimate of regional mobility in Latvia shown in Figure 4.7. Firstly, it refers to mobility between regions at the NUTS3 level, a smaller unit than the NUTS2 level available for the other countries. Given Latvia’s comparatively small population, all of Latvia is considered as a single region at NUTS2 level. By consequence, a part of the mobility between Latvia’s regions would be counted as mobility within the same region in other countries. If, however, the comparison in Figure 4.7 could be made at NUTS3 level throughout, larger countries would exhibit high mobility rates partly because they include a large number of regions at NUTS3 level, which allows for many combinations of regions and thereby creates greater scope for regional migration within the same country. Secondly, the estimate for the mobility of unemployed persons in Latvia covers only registered unemployed persons. Since unemployed persons in the other countries in Figure 4.7 mostly exhibit substantially higher mobility than the average, it appears likely that the corresponding estimate for Latvia would be higher than the estimate covering only registered unemployed persons, which remains close to the average.

Young unemployed exhibit a low willingness to commute but a rather high willingness to move within Latvia

The mobility of young people can play an important role for overall mobility between regions (for example, Hunt (2006[17])). For the age group 15-34, survey data are available on the willingness to move or commute for a job (Figure 4.8). In Latvia, 44% of unemployed persons in this age group indicate their willingness to commute for a job, which is one of the lowest values among European countries. Lower values occur only in the Netherlands (36%) and Lithuania (42%), while around 80% are willing to commute in Spain, Italy and Croatia and the EU average approaches two-thirds.

Figure 4.8. Willingness of unemployed persons aged 15-34 to be mobile for a job, European countries, 2016 or latest available year

By contrast, the willingness of young unemployed persons in Latvia to move within the country for a job is higher than in most other countries for which this information is available (Figure 4.8). More than one-quarter (26%) of those surveyed in Latvia are willing to move, compared with an EU average value of 20%. Significantly higher values are only observed in Greece (31%), Ireland (32%), Belgium (34%), Germany (35%) and Switzerland (49%). The rather high willingness of young unemployed in Latvia to move implies that policies can draw on this willingness and effectively facilitate their regional mobility, e.g. by reducing barriers to mobility.

The share of young unemployed persons who consider emigrating abroad is especially high in Latvia, reaching one-half compared with an EU average of 41%. Emigration intentions appear significantly more widespread only in Lithuania (54%), Slovenia (58%) and Romania (63%). Figure 4.8 further indicates that regional migration and emigration to another country are not clear substitutes for young unemployed persons: across European countries, willingness to move nationally is essentially uncorrelated with intentions to emigrate.

Based on administrative data, Figure 4.9 shows moves of registered unemployed persons that can be associated with taking up employment. Over time, these moves are subject to considerable seasonal variation. When moves between all municipalities are considered, the average monthly mobility rate over the entire period February 2012 to January 2017 is 0.20%, corresponding to an average of 143 moves per month. The highest mobility rate in this period occurred in March 2016 when the mobility rate reached 0.37%, or almost 300 moves. Another series in Figure 4.9 includes only those moves that take place between municipalities that are not adjacent to each other. In this context, a number of municipalities in close proximity are considered adjacent although they do not share a border (see Annex Table 4.A.1). With an average monthly mobility rate over the entire period of 0.13% (on average 92 moves per month), moves between non-adjacent municipalities are considerably less frequent, but appear to follow roughly the same variation over time as moves between all municipalities.

Figure 4.9. Rates of employment-related mobility among registered unemployed, 2012-2017

Figure 4.10 depicts flows between regions that result from moves of registered unemployed persons associated with taking up employment. Over the period January 2012 to January 2017, 2 500 registered unemployed persons moved to Riga to take up employment. Most of them had previously resided in Pieriga (930), while around 400 came from each of the other regions. With a total outflow of close to 1 500 registered unemployed persons, Riga attracted a net inflow of more than 1 000 persons. Pieriga saw a high inflow of persons from Riga (800) but less than 200 from any other region. As total inflows to Pieriga roughly balanced total outflows, the net flow was close to zero in this case. Vidzeme, Zemgale, Kurzeme and Latgale all exhibited net outflows of about 200-300 persons. Their main inflows originated from Riga (150-200 persons) and Pieriga (90-170 persons). The inflows from Latgale were almost always the lowest inflows received from any region. In addition to the lowest total outflow, Latgale also exhibited the lowest total inflow.

Figure 4.10. Employment-related mobility of registered unemployed between regions, 2012-2017

Figure 4.10 further shows that a region’s total inflows appear linked to its job-finding rate, defined as the number of exists from unemployment to employment divided by total unemployment. The highest job-finding rates over the period January 2012 to January 2017 are observed in Riga and Pieriga (42% and 41%, respectively). The same two regions also received the largest total inflows. By contrast, Latgale exhibited a job-finding rate of 34% over this period and received the lowest total inflow. The case of Kurzeme, however, deviates from this pattern: despite a higher job-finding rate than in Zemgale (40% compared with 39%), Kurzeme received a significantly lower total inflow.

Mobility requirements for unemployed persons can better reflect their situation

The SEA expects registered unemployed to take up suitable job offers and can otherwise de-register the unemployed person, which stops unemployment benefits and access to other SEA services. The distance between the workplace and the residence of the unemployed person is one of the criteria for suitability. A job will be deemed suitable if the unemployed person can travel there on public transport within one hour (when unemployed for up to three months) or one and a half hours (when unemployed for more than three months). In addition, the residence of the unemployed persons should not be more than two kilometres from the public transport connection, and likewise for the workplace. In addition, the costs of travelling to work should not exceed one-fifth of the gross wage offered. However, it is unclear to what extent these requirements are enforced: as shown in Chapter 2, hardly any sanctions were applied in 2016 because a suitable job offer was turned down.

In practice, the individual ability to take up a distant job offer likely depends on the family situation of the unemployed person and also on access to a car rather than public transport alone. According to the Mobility Survey of Latvia’s Population in 2017, 46% of all journeys on working days that are shorter than 300 km are undertaken by car. One-third are made walking and another 4% by bicycle. Transport modes that make up public transport – buses, trams and trains – only account for 16% of chosen transport modes on working days. Many unemployed persons might be able to commute to a distant workplace within a reasonable time by using their car instead of public transport, which is necessarily limited in rural areas. Therefore, access to a car could be made an important criterion in the decision whether or not a distant job offer is suitable.

With respect to family situation, stricter mobility requirements could apply to unemployed persons who are single or whose partner does not hold a local job. Especially in the case of young persons, someone without family commitments at the place of residence can be expected to move anywhere in Latvia in order to take up employment, given that financial support for such moves is available from the SEA (see below). This may be an effective policy lever to avoid some cases of long-term unemployment, while the resulting distances would mostly remain limited. In this context, caseworkers play the key role for identifying an unemployed person’s individual barriers to mobility and for highlighting the potential benefits of mobility.

Hofmann (2015[18]) investigates the effect of tighter mobility requirements for unemployed persons without family commitments in Germany: from January 2003, these unemployed persons were required to move in order to take up a distant job unless there is reason to expect that a suitable job opportunity arises in the current region of residence. Hofmann (2015[18]) estimates a substantial positive effect of this change on the employment of women without children from high-unemployment regions (an increase of close to 5 percentage points). The additional employment was generated both in other regions and in the current region of residence, likely reflecting that stricter mobility requirements raised the pressure on them to find employment more generally. This indicates that stricter mobility requirements might not push all unemployed persons concerned out of their current region and may also promote taking up local employment.

High housing costs act as a brake on mobility

House prices underwent a severe correction between 2008 and 2010 (Figure 4.11). While sale prices of (new and existing) residential property had grown by 36% in 2007, they fell by as much in 2009. Since 2011, however, these sale prices have again tended to rise significantly, and growth rates accelerated to 8%-9% in 2016/2017. An index for real house prices has likewise exhibited a rising trend since 2012. Rent levels have risen more strongly than real house prices over the years 2010-2015 but stagnated after 2015. Overall, housing costs in Latvia have recently tended to rise significantly but have remained far below levels reached during the boom. While the link between home ownership and mobility is pointed out in OECD (2017[19]), this section examines to what extent unemployed persons have been affected by rent increases in each region, and discusses the likely consequences for their mobility.

Figure 4.11. Indicators for housing costs in Latvia, 2007-2017

As indicated in Figure 4.6, the centre of Riga attracts relatively little internal migration, in contrasts to areas surrounding the centre and the city or Riga. This likely reflects the strong rise of housing costs especially in the centre of Riga. The monthly median rent paid by an employed person residing in Riga has risen from EUR 80 in 2012/2013 to EUR 140 in 2015/2016 (Figure 4.12). Strong increases in median rent levels also occurred in Pieriga (EUR 30 to EUR 60) and Zemgale (EUR 17 to EUR 50). New hires from unemployment seem to avoid the high rents in Riga, paying about as much in 2012/2013 (EUR 80) as in 2015/2016 (EUR 75). The newly hired former unemployed persons typically pay lower monthly rents than employed persons in total, with the notable exception of Pieriga. New hires from unemployment in Pieriga appear to be bear much of the overall increase in rent levels there, paying a median monthly rent of EUR 100 in 2015/2016 from EUR 40 in 2012/2013. This aligns with the finding in Figure 4.6 that Pieriga has received high internal migration flows. In Kurzeme and Latgale, by contrast, monthly rents have remained low overall.

Several reasons could explain why median rents for new hires were somewhat higher in Pieriga than in Riga in 2015/2016. As a result of the housing boom, Pieriga offers a large part of relatively new accommodation, while renovation of existing dwellings was more important in Riga. Due to its proximity to Riga and extensive transport infrastructure, residents of Pieriga still have access to Riga and its labour market. At the same time, population density is much lower in Pieriga than in Riga, so that living in Pieriga might often be perceived more attractive in terms of quality of life.

Figure 4.12. Monthly rents paid by employed persons in Latvia’s regions, 2012/13 and 2015/16

A further issue is that of housing quality as measured by size and modern amenities, for example. OECD (2017[19]) emphasised that, in 2014, 36% of low-income households (the lowest quintile in the household income distribution) and 34% of middle-income households (the third quintile) lived in accommodation considered to be overcrowded. These values are among the highest among OECD countries, and similar or higher values were only observed in Poland and Hungary. According to a 2018 study by the Latvian Ministry of Economics, total monthly rents of EUR 450 to EUR 530 are charged to new tenants for apartments of 50 square metres that conform with modern standards (thus around EUR 10 per square metre, including utilities and charges). For four-fifths of households, the study concludes, more than 30% of disposable household income would have to be spent on housing in order to afford such an apartment.

The high monthly rents especially in Riga and Pieriga but also in Zemgale might discourage some regional mobility in Latvia. A proportion of job vacancies offer wages that would attract jobseekers from other regions were it not for the difficulties to pay local rents out of this wage. By consequence, these jobs vacancies are more likely to be filled by those who already have affordable accommodation near the workplace, and some vacancies will remain unfilled (see Chapter 1). OECD (2017[19]) argues that the lack of affordable housing, alongside weak public transport, limits the extent to which Riga’s economic strength can emanate into surrounding regions and benefit Latvia more widely.

The link between housing costs and regional mobility has been documented for a number of OECD countries. For the case of Italy, (Cannari, Nucci and Sestito (2010[20]) find that increasing housing costs in norther Italy have offset the impact of high incomes and good employment prospects on South-North migration within Italy. In the United States, internal migrants appear less likely to move to areas with high housing costs, after accounting for other factors (Plantinga et al., 2013[21]). A lack of housing supply may be linked to limited regional migration towards urban areas in the Netherlands (Vermeulen and van Ommeren, 2009[22]). Housing cost differentials were also found to affect regional mobility in Finland (Hämäläinen and Böckerman, 2004[23]) and the United Kingdom (Rabe and Taylor, 2012[24]).

Evaluation of the programme for regional mobility

In order to promote the regional mobility of registered unemployed persons within Latvia, the SEA operates an ALMP programme that offers support with taking up job offers (including subsidised employment) or with attending training measures at rather distant locations. Until 2018, jobs and training measures have been considered distant if they are located at least 20 kilometres from the residence of unemployed person. This threshold was reduced to 15 kilometres in 2018, which is just above the average distance travelled on workdays in order to commute to work (13.6 kilometres), according to the Mobility Survey of Latvia’s Population.

Unemployed persons who are offered a job or training at a distant location are eligible for support under the programme for regional mobility, provided they have been unemployed for at least two months. Up to EUR 150 (EUR 100 until the end of 2018) per month may be reimbursed under the programme for cost of transport or cost of accommodation at the new workplace or the training site. The reimbursement is limited to the first four months of the new job but can cover the entire duration of a training measure. Moves within the region of Riga are currently not eligible, and a second participation in this programme is normally only possible 36 months after the first one.

The programme was introduced in 2013, and the total number of participants approached 9 200 by the end of 2017. For 3 000 of these participants, the programme was provided under the Youth Guarantee. Details on participants can be derived from the BURVIS Database of the SEA. The available sample from this database covers the period from January 2012 to October 2017, in which close to 9 000 participants are observed. Specifically for this Review, observations from the BURVIS Database were linked with the corresponding information from Latvia’s State Social Insurance Data, the Social Assistance Database (SOPA) that covers data collected by municipalities, and the Population Register.

Participants are typically younger than 35, unmarried and not highly educated

The set of linked administrative data offers a wealth of information on participants in the programme for regional mobility. Women make up half of all participants, and almost 60% are under 35 years old. Younger age groups tend to be larger: participants aged 15-24 make up the largest age group (accounting for 37%) followed by ages 25-34 (22%), ages 35-44 (21%) and ages 45-54 (18%), while less than 3% of participants were aged 55-60. Two-thirds were not married, so that relatively young persons, often without family commitments, are predominant among participants in this programme. Close to 6% of participants have disabilities, and 15% are long-term unemployed (more than one year) at the start of the programme.

One-fifth of participants have a higher or a professional higher level of education and another fifth has only a basic level of education (Figure 4.13, Panel A). One-quarter of participants come from Latgale, followed by Pieriga (Figure 4.13, Panel B). Only few participants (4%) come from Riga, which partly reflects the exclusion of moves within the Riga region from the programme but likely also reflects that residents of Riga relatively rarely move elsewhere for employment. Moves from other regions to Riga were excluded until March 2018. The available information on reimbursements (covering 6 000 participants) suggests that almost all reimbursements (98%) were for transport costs and only 2% for cost of accommodation.

Figure 4.13. Participants in the programme for regional mobility, 2012-2017

Eligibility rules can be used to identify the impact of the programme

This section briefly presents the methods used in the impact evaluation of the programme for regional mobility, while the next section will present the results of these analyses. As mobility related to training is included in the evaluation of training (see Chapter 3), the impact evaluation here focuses on mobility to take up employment. Then the programme has a positive impact if it raises mobility to take up employment. However, in contrast to most impact evaluations, it is not possible here to define programme participants as treatment group and compare them to some control group that did not participate. After all, a concrete offer of a distant job (or participation in distant training) is a precondition for participation in the programme for regional mobility. Therefore, mobility to take up employment will necessarily be higher among participants than among non-participants.

Due to this inherent self-selection of participants, a broader perspective is needed that abstracts from the group of participants. The approach proposed here draws on the programme’s eligibility rules: it defines eligible persons as treatment group and ineligible persons as control group. Then the mobility to take up employment is compared over time, before and after the introduction of the programme or the relevant change in eligibility rules: did this mobility increase in the treatment group after this point in time? To ensure that any increase does not simply reflect an overall trend over time towards greater mobility, the mobility of the treatment group is always compared to the mobility of the control group, which should be unaffected because it is not eligible for the programme.

The programme’s impact is therefore identified as the change over time in mobility to take up employment that is observed in the treatment group but not in the control group. As general mobility trends should apply about equally to both groups, they should cancel out in such a comparison. This approach, known as difference-in-difference (DID), is widely considered especially robust (Athey and Imbens, 2006[25]). Box 4.2 presents the details of how this approach is implemented in this Chapter.

Box 4.2. Effects from Latvia’s programme for regional mobility: estimation methods

The estimation is set up using a linear probability model. At the level of individuals indexed by i, the model equation to be estimated relates the outcome variable, the individual’s mobility behaviour $M_{i}$ , to various explanatory variables:

$M_{i} = β_{0} + β_{1} T_{i} + β_{2} A_{i} + β_{3} [T_{i} * A_{i}] + X_{i}^{'} β_{4} + γ_{i} + δ_{t} + ε_{i}$

where $T_{i}$ indicates whether or not individual i is in the treatment group, $A_{i}$ indicates whether the observation comes from the period after the programme introduction or the rule change, and $[T_{i} * A_{i}]$ is the interaction of these two variables. Next, $X_{i}^{'}$ is a vector of observed characteristics of individual i, $γ_{i}$ is a fixed effect of the home region and $δ_{t}$ is a year fixed effect. These fixed effects capture some macroeconomic and institutional influences associated with regions or with developments over time. Further indicators for calendar months are included to account for seasonal variation. Finally, $ε_{i}$ allows for random disturbances in the empirical relation.

The coefficients $β_{0} \dots β_{4}$ have to be estimated, where $β_{0}$ accounts for a constant part of $M_{i}$ . In this model, $β_{1} T_{i}$ captures the difference between treatment and control group, while ${β_{2} A}_{i}$ captures the difference in mobility before and after the programme introduction or rule change. Then $β_{3}$ is the parameter of interest because it will be significantly different from zero if individuals in the treatment group become more mobile after the programme introduction or the rule change, in a way that is not captured by $β_{1} T_{i}$ nor by ${β_{2} A}_{i}$ and not explained by the other explanatory variables either.

The outcome variable $M_{i}$ is an indicator that equals one if individual i is observed to be mobile to take up employment, and equals zero otherwise. Therefore, the model does not estimate the impact on the number of moves – a measure that changes with the number of unemployed persons, for example – but on the individual probability of being mobile to take up employment. The estimated coefficients $β_{0} \dots β_{4}$ can be interpreted as the increase in the probability (between zero and 100%) due to a one-unit increase in the respective explanatory variable.

The programme has succeeded in raising mobility

The estimation uses the linked administrative data arranged in terms of individual histories between January 2012 and January 2017, so that one observation corresponds to one month in the history of an individual. The outcome variable $M_{i}$ is therefore the probability that mobility to take up employment occurs in a particular month. This probability is very low, and the estimated effects will accordingly be very low in absolute terms (but may be substantial when compared to the average probability). However, given large numbers of observations, effects can still be identified reliably. Data on months after January 2017 are not used because most of the mobility in this period does not seem to be recorded in the available data, for unknown reasons. Months in which the individual’s labour force status is recorded as retired are excluded from the analysis. Since none of the participants in the programme for regional mobility is older than 60, months in which individuals are older than 60 are also excluded.

The analyses define mobility to take up employment as follows. Mobility is measured at the level of municipalities: an individual is considered mobile in a given month if the municipality where the individual resides is different from the previous month. Such a move is counted as mobility to take up employment if the individual was not employed at any time up to six months before the move and up to two months after the move but becomes employed up to three months before the move and up to three months after move. Therefore, the definition allows the month of the move to differ from the month in which employment is taken up. In practice, those taking up employment might first live in temporary accommodation and move only several months later, or they might move some time ahead of the beginning of employment. Since moves are identified through changes in registered addresses, delays in registration also need to be allowed for.

The analyses attempt to distinguish between moves over a short distance and moves over a longer distance. In addition to a first measure that includes moves between any two municipalities to take up employment, a second measure disregards moves between municipalities that share a border. The remaining moves between municipalities that are not adjacent to each other are more likely to reach a distance of 20 kilometres or more, as required by the eligibility rules of the programme for regional mobility. Some municipalities that do not share a borer are, however, sufficiently close to each other to count as adjacent, and Annex Table 4.A.1 in Annex 4.A lists these cases.

The linear probability model is implemented in three versions, reflecting three ways to exploit the introduction of the programme or rule change for the impact evaluation (Table 4.1). Key results for these three model versions are presented in Table 4.2, notably the estimate for the parameter of interest ( $β_{3}$ ). Model 1 compares mobility to take up employment before and after the introduction of the programme in March 2013. The treatment group comprises all registered unemployment, as only persons registered with the SEA are eligible for participation in the programme. The control group are persons not in employment but also not registered unemployed. The broad definition of the control group, rather than considering only non-registered unemployed, avoids the problem that changes in the participation margin between unemployment and inactivity affect the estimation results.

Table 4.1. Econometric approaches in the evaluation of the programme for regional mobility
	Considering effects on:	Considering effects from:	Treatment group	Control group
Model 1	Mobility to take up employment	Introduction of the programme for regional mobility in March 2013	Registered unemployed persons	Other persons not in employment
Model 2	Mobility to take up employment	Introduction of the programme for regional mobility under the Youth Guarantee in August 2014	Registered unemployed persons aged 15-29 (eligible under the Youth Guarantee)	Registered unemployed persons aged 30 or above (not eligible under the Youth Guarantee)
Model 3	Mobility to take up employment in the public sector	Inclusion of public sector employers in the programme for regional mobility from March 2016	Registered unemployed persons	Other persons not in employment
Note: Observations on unemployed persons older than 60 years are excluded from all analyses.
Source: OECD secretariat.

The results from the estimation of model 1 suggest that the introduction of the programme for regional mobility has had a positive impact on mobility to take up employment (Table 4.2, Part A). It appears to have raised the probability that such a move occurs in a particular months by 0.032 percentage points, and the probability that such a move occurs between non-adjacent municipalities by 0.023 percentage points. To place the very low magnitude of these effects into perspective, Part A of Table 4.2 also reports the (unconditional) average probabilities of such moves occurring in a particular month before the programme was introduced. This probability reached 0.031% for all moves and 0.019 percentage points for moves between non-adjacent municipalities. Against these average probabilities, the effects from the introduction of the programme appear considerable, but are by definition limited to the treatment group. After accounting for other factors, those in the treatment group of model 1 (registered unemployed persons) exhibit a probability of 0.062%: registered unemployed are especially likely to be mobile to take up employment. The effect of the programme introduction may be interpreted as raising their mobility by one-half.

Model 2 compares mobility to take up employment before and after the introduction of the programme for regional mobility under the Youth Guarantee in August 2014. Programmes operated in Latvia under the Youth Guarantee are only available to registered unemployed aged 15-29. While persons in this age group had access to the programme for regional mobility introduced in the preceding year, the introduction of this programme under the Youth Guarantee might produce additional effects in the age group 15-29. For example, the objective of the Youth Guarantee to offer young unemployed persons employment, education or training within four months introduces a certain time pressure. In model 2, registered unemployed aged 15-29 are therefore the treatment group, and older registered unemployed serve as control group.

While the average probabilities are essentially the same in model 2 as in model 1, the estimated impact is considerably lower, but nevertheless significant. Introducing the programme for regional mobility also under the Youth Guarantee appears to have raised the mobility to take up employment by 0.008 percentage points in the group of registered unemployed aged 15-29, and by 0.006 percentage points when only moves between non-adjacent municipalities are considered (Table 4.2, Part A). The weaker effects may be weaker because persons in this age group are already more mobile than older age groups, which might make it more difficult to raise it further. In addition, the programme for regional mobility had already existed for more than one year outside of the Youth Guarantee.

Table 4.2. Estimation results for the impact of the programme for regional mobility on mobility to take up employment, 2012-2017
Probabilities in percentage points
Dependent variable	Probability before: model 1	Probability before: model 2	Probability before: model 3	Effect in model 1	Effect in model 2	Effect in model 3

A. Full sample of newly unemployed persons
Moves between any two municipalities	0.031	0.031	0.003	0.032	0.008	0.002
Moves between non-adjacent municipalities	0.019	0.020	0.002	0.023	0.006	0.001
B. Subsample of social assistance recipients
Moves between any two municipalities	0.042	0.036	0.003	0.011	0.008
Moves between non-adjacent municipalities	0.025	0.022	0.002	0.009	0.006	0.002
C. Subsample of persons with disabilities
Moves between any two municipalities	0.019	0.024	0.002	0.012	0.012	-0.003
Moves between non-adjacent municipalities	0.012	0.015	0.002	0.011
D. Subsample of Latgale residents
Moves between any two municipalities	0.028	0.030	0.004	0.026
Moves between non-adjacent municipalities	0.018	0.018	0.003	0.020
Note: Only results that are statistically significant at the 5% significance level are reported. In model 3, only mobility to take up employment with public-sector employers is considered. Analysis B includes all persons who received social assistance at some point in the observed history between January 2012 and January 2017. Analysis C includes all persons recorded as having disabilities at some point in this observed history, and analysis D includes all those who resided in Latgale at the beginning of their observed history.
Source: OECD analysis using linked administrative data from BURVIS (SEA), the State Social Insurance Agency (SSIA), the Social Assistance Database (SOPA) and the Population Register (OCMA).
StatLink https://doi.org/10.1787/888933961885

Finally, model 3 compares mobility to take up employment before and after a change in eligibility rules in March 2016, when jobs offered by public-sector employers become eligible for support from the programme for regional mobility. In this model, outcome variable $M_{i}$ only captures mobility to take up employment with public-sector employers. Since such moves only make up a fraction of mobility to take up employment, effects estimated from model 3 may be especially small and more difficult to identify than in the case of model 1 or 2. As in model 1, the treatment group consists of registered unemployed persons and the control group consists of other persons not in employment.

The estimated effects for model 3 are still positive but small compared with the effects estimated for models 1 and 2 (Table 4.2, Part A). However, the effects for model 3 appear substantial when compared with the average probabilities for mobility to take up employment in the public sector in a particular month, which was only 0.003% before March 2016, and 0.002% for such moves between non-adjacent municipalities. In the treatment group (registered unemployed), the probabilities were 0.08% and 0.06%, respectively, so that the effect of the rule change appears to have raised the mobility to take up employment by about one-third.

The estimation of all models includes a range of individual characteristics, most of which are consistently found to matter for mobility to take up employment. Age is negatively correlated with the mobility to take up employment. Being female is slightly positively correlated, while being a Latvian citizen has a somewhat stronger positive correlation (which might proxy unobserved factors such as language proficiency in Latvian). Receiving unemployment benefits or social assistance correlates positively (except for receiving social assistance in model 2). Disabilities correlate negatively with mobility to take up employment. Several variables account for family situation: total household size, the number of children aged under 7, the number of children aged 7-18, and whether a child was born in the preceding year. While household size and recent childbirth correlate negatively, the number of children under 18 correlates positively. This last finding might be linked to age: families with younger parents (who are more likely to have young children) may be more mobile than families with older parents. Finally, having any other region than Riga as home region is positively correlated. Overall, the findings appear plausible and the fact that they are statistically significant supports the empirical validity of the three models.

Education and unemployment duration are further potentially relevant explanatory variables. However, information on educational attainment is only collected for registered unemployed, and unemployment duration is only observed for registered unemployed. Therefore, these two explanatory variables can only be included in model 2 which uses registered unemployed in both the treatment and the control group. When unemployment duration in months and indicators for education levels are included in the estimation of model 2, the effect remains significant at a slightly lower magnitude than before (0.006 percentage points and 0.005 percentage points for moves between non-adjacent municipalities). Compared with a basic education level, every higher level of educational attainment correlates positively with mobility to take up employment. Unemployment duration in months also correlates positively in this case, which might result from the strong focus of activities under the Youth Guarantee on the first four months of registered unemployment – within this time, mobility might be more in the later months. For registered unemployed persons overall, graphical evidence clearly suggests a negative duration dependence of mobility (Figure 4.14).

Especially high mobility rates during the first three months of unemployment raise the question whether a shift to shorter unemployment durations over time drives the estimated effects of the programme when unemployment duration is not included. To investigate this, model 1 is estimated again including only unemployed persons in the first three months of unemployment in the treatment group (without a change to the control group). The estimated effects of the programme are still positive and significant, albeit lower than for all unemployment durations (0.021 percentage points and 0.015 percentage points when only moves between non-adjacent municipalities are considered).

Figure 4.14. Monthly rates of mobility to take up employment by unemployment duration, 2012-2017

A positive impact is found for recipients of social assistance but not necessarily for persons with disabilities

The estimation of the three models can be repeated for specific groups only, in order to examine how estimated effects differ. The groups considered here are recipients of social assistance, persons with disabilities, and residents of Latgale (Table 4.2, Parts B-D). Raising the regional mobility might be especially important for improving the labour market prospects of these three groups because they may often face very limited suitable job opportunities close to their current residence (which applies to recipients of social assistance insofar as it signals sustained problems with integrating the local labour market). In each case, the sample is restricted to certain individuals (while maintaining their entire observed employment history). The estimation for recipients of social assistance includes all observed individuals who received social assistance at some point between 2012 and 2017. The estimation for persons with disabilities includes the individuals who were at some point recorded as having disabilities, and the estimation for Latgale includes individuals who reside in Latgale at the beginning of their observed employment history.

Results for the estimated effect of the programme for regional mobility on recipients of social assistance suggest that their mobility to take up employment in a particular month has increased by about 0.010 percentage points, according to model 1, and by around 0.007 percentage points according to model 2 (Table 4.2, Part B). When model 3 is estimated for this group, only the effect on moves between non-adjacent municipalities is statistically significant but reaches the same magnitude as in Part A of Table 4.2. For persons with disabilities, statistically significant results based on model 1 and model 2 indicate that their mobility increased by 0.012 percentage points due to the programme for regional mobility (Table 4.2, Part C). By contrast, one estimation based on model 3 indicates a negative effect on their mobility to take up employment in the public sector. Together, these results could mean that the programme for regional mobility primarily facilitates the mobility of persons with disabilities to take up suitable employment in the private sector, which allows them to rely less on jobs in the public sector. For Latgale residents, only estimates from model 1 are statistically significant (Table 4.2, Part D). These estimates are close to those obtained in Part A of Table 4.2, which suggests that the introduction of the regional mobility programme in 2013 raised the mobility of registered unemployed persons in Latgale as strongly as on average across Latvia.

The last finding provides a first indication that the estimation results are robust although regional factors are not fully accounted for – such as differences in geographical distance, transport links, and preferred language. To some extent, these differences have been captured by including fixed effects for regions in the estimation. In addition, one can allow for clustered standard errors at the level of regions. This modified estimation indicates that at four sets of results are robust to shortcomings from missing regional variables: the positive effects found for all unemployed persons under model 1 as well as under model 2, the positive results for residents of Latgale under model 1, and the negative result for persons with disabilities found under model 3.

The literature offers a few impact evaluations of comparable programmes. Westerlund (1998[26]) assesses the “starting assistance grant” in Sweden, which was initially introduced mainly to increase mobility. The empirical results suggest that an increase in the level of this grant has not increased regional mobility. Caliendo, Künn and Mahlstedt (2017[27]) evaluate the effect of “relocation assistance” in Germany that subsidises moving costs of unemployed persons. After accounting for self-selection into the programme, they find that participants earned substantially higher wages and enjoyed greater employment stability than a comparison group. However, they do not offer conclusive evidence on the programme’s effect on regional mobility. In Caliendo, Künn and Mahlstedt (2017[28]), they find that mobility support leads unemployed persons to search for jobs across a larger area, which raises their prospects of finding employment.

Le Gallo, L'Horty and Petit (2017[29]) examine the effect of subsidised driving lessons for youth in France on their mobility and labour market outcomes. According to the results, the affected youth are initially less mobile during the training phase before a positive effect on their mobility and job search efforts eventually materialises. Finally, it is worth noting that many other ALMP measures might, as a by-product, reduce regional mobility through lock-in effects: participants in these programmes might stay longer in a region with poor employment prospects than they would have in the absence of ALMP measures in these regions. Hämäläinen (2002[30]) uses data from Finland to show that this effect arises in times when unemployment in potential destination regions is low, so that moving there would have been promising.

Support for mobility can be scaled up in a targeted way

Given that the programme for regional mobility has positive effects on the mobility of unemployed persons to take up a job, it is worthwhile exploring how it can be scaled up. At the same time, the programme should be cost-effective and not support mobility that would also have taken place in the absence of the programme. The BURVIS Database of the SEA includes information on the minimum salaries associated with jobs taken up by programme participants. Where this information is available in the period 2012-2017, 22% of the jobs taken up by programme participants paid at least a monthly salary of EUR 280-299. About one-third paid at least EUR 300-349, 27% at least EUR 350-399, and 6% at least EUR 400-499. Only 7% of jobs paid at least EUR 500-599, and only 5% at least EUR 600. Thus far, the programme therefore appears well targeted.

Under the current rules of the programme, financial support is limited to EUR 400, a substantial amount for many unemployed persons in Latvia. Where couples or entire families would have to move, however, this limit appears too low to cover the up-front costs that such moves often entail. In this context, it may be possible to offer unemployed persons higher amounts as repayable loans from a service provider. The SEA could act as intermediate and substantially reduce the risk of such loans, notably through its knowledge of the expected salary of the new job. As it is unlikely that such loans are taken out by those who can afford to move without further support, unnecessary costs would be avoided.

By another rule of the programme, unemployed persons become eligible for support only after an unemployment duration of two months. As shown in Figure 4.14 above, job-related mobility is especially likely to take place in the first three months of unemployment. Unemployed persons whose mobility depends on financial support then either decline job offers during the first two months or postpone the new job until two months have passed, which needlessly prolongs unemployment and benefit payments from the SEA. Therefore, it would be preferable to abolish this rule and find other ways to limit mobility support to those who would otherwise not move. For example, eligibility for the programme could instead be linked to certain profiling outcomes, so that unemployed persons with very good job prospects are not eligible.

The potential of entrepreneurship for unemployed persons in Latvia’s regions

Policy makers can also address unemployment across Latvia’s regions by fostering entrepreneurship and start-ups, in order to generate sustained employment growth within the regions themselves. Unemployed persons can benefit in two ways, by finding jobs in newly created firms or by becoming entrepreneurs themselves. Based on evidence from Germany, Fackler et al. (2018[31]) find that start-ups are more likely to hire disadvantaged jobseekers. If these results generalise, promoting entrepreneurship in regions may be an effective way to reduce regional unemployment at the same time.

Numbers of self-employed persons have been on the rise in most of Latvia’s regions over recent years (Figure 4.15). Especially in Pieriga, they have risen steadily between 2012 and 2016, rising by almost 40% over this period. Strong growth was also observed in Riga and Vidzeme, where numbers increased by 17% in both cases. Growth rates in Kurzeme and Latgale approached 10%, while the number of self-employed persons in Zemgale was fluctuating. The increasing tendency still arises when these figures are related to the total active population.

The overall impression from the figures on self-employment is corroborated by figures on firms. The net creation rate of firms – the difference between newly created firms and firms being closed – is comparatively high Latvia and reaches about 5% (Figure 4.16). This does not only hold for urban areas (6%) but also for intermediate and predominantly rural areas. Among OECD countries for which this information is available, higher net creation rates are only observed in Estonia and Hungary. Especially high shares of Latvia’s population declare in surveys that they expect to create a business within three years: one-quarter of the population aged 15-64 and 40% of those aged 18-30, close to twice the EU average in both cases (OECD/European Union, 2017[32]). To an unknown extent, however, the creation of firms in Latvia also reflects a preferential tax regime for so-called micro enterprises. This can include the redefinition of activities as a micro enterprise in order to qualify for lower tax rates. In the context of widespread concerns about such abuse, micro enterprises will be limited to a few sectors from 2019.

Figure 4.15. Self-employed persons in Latvia’s regions, 2012-2016

Figure 4.16. Entrepreneurship in selected OECD countries by type of region, 2015

Since the digital economy offers many of the opportunities for entrepreneurship with a high growth potential, entrepreneurship in Latvia’s regions can benefit strongly from access to digital infrastructure. Broadband coverage promises to be a key piece of infrastructure that allows rural regions to catch up with cities. According to the OECD Broadband Portal, the high-performing fibre connections make up an especially large share (63%) of broadband connections in Latvia. Among OECD countries, higher shares only occur in Japan (75%) and Korea (74%). As Latvian authorities increasingly offer e-services, the possibility to register new companies online was introduced in 2018 (OECD, 2018[33]). Using firm-level evidence, Revoltella et al. (2019[34]) find that also regional infrastructure and institutions are an important factor for innovation in firms.

An untapped potential of entrepreneurs exists among the unemployed in Latvia

Survey data from 2015 suggests that 3% of all unemployed persons in Latvia would like to become self-employed (Figure 4.17). This value was above the average for EU countries, and substantially higher values were only observed in Austria (4.2%) and Luxembourg (7.7%). However, only 2% of unemployed persons in Latvia moved into self-employment in 2016. The discrepancy between the share that seeks self-employment and the share that moves into self-employment was larger in Latvia than other European OECD countries for which this information was available (except in Luxembourg and Greece, albeit with a reversed order of shares). This highlights an untapped potential of entrepreneurs among unemployed persons in Latvia: if it could be mobilised, the share of unemployed who move into self-employment would increase by one-half.

Figure 4.17. Transition from unemployment to self-employment, European OECD countries, 2015/2016

For all EU countries combined, some further evidence is available by gender and age (OECD, 2017[35]), while sample sizes do not allow a breakdown of this information by country. The share seeking self-employment was larger among unemployed men in EU countries (2.6%) than among unemployed women (1.7%). It was also higher for unemployed aged 50-64 (2.4%) than for unemployed aged 15-24 (1.1%). The last finding provides a first hint that self-employment could be especially welcome as a tool to reduce unemployment among older workers.

Some existing evidence also suggests that entrepreneurship courses might be a policy lever for mobilising the potential for entrepreneurship. In the age group 18-30, Latvia’s nascent entrepreneurship rate – the share of the population involved in creating a new business – is one of the highest among OECD countries (Figure 4.18). In 2012-2016, it exceeded 12% and was only behind the rates in Estonia (13%) and Chile (14%). This aligns with a comparatively high share of this age group who believes to have the necessary know-how: in Latvia, almost half of those in the age group 18-30 indicated in a survey that they have the knowledge and skills to start a business (see Figure 3.13 in OECD (2017[35])). This share was only higher in Chile, Turkey and Poland.

Figure 4.18. Entrepreneurship and needs for entrepreneurship skills of youth and older workers, 2012-2016

By contrast, Latvia’s nascent entrepreneurship rate in the age group 50-64 falls into the lower half, relative to other OECD countries (Figure 4.18). At 3% in 2012-2016, it was far lower than in the age group 18-30, and the gap of 9 percentage points was the highest among OECD countries except for Estonia (10 percentage points). A part of the gap could be explained by less familiarity with entrepreneurship among older workers: in the age group 50-64, only 40% believed to have the knowledge and skills to start a business, which again falls into the lower half when compared to other OECD countries (Figure 4.18). This indicates that courses in entrepreneurship could help mobilise potential entrepreneurs especially among older workers. According to a recent survey in Poland, participants in entrepreneurship courses were most interested in starting a first entrepreneurial project within these courses, visiting companies and internships, alongside writing business plans and developing ongoing projects (OECD, 2017[36]).

Assessment of start-up subsidies for the unemployed

Against this background, the programme for entrepreneurship and self-employment operated by the SEA (and also offered under the Youth Guarantee) can play an important role for reducing unemployment in Latvia, across regions. The programme assists selected participants with the formulation of business plans and, at a first stage, provides feedback on these business plans in up to 20 consultation sessions for each participant. When submitted to a commission of business professionals, about one-quarter of business plans are approved. To implement approved business plans at the second stage, grants of up to EUR 3 000 can be awarded, as well as monthly stipends at the level of the minimum wage for up to six months.

However, the size of the programme is small compared to many other ALMP programmes in Latvia: according to annual figures from the SEA, about 1 450 participants have reached the second stage during the period 2012-2017, alongside 4 300 participants who only reached the first stage. The annual number of participants at the second stage has fluctuated over this period between 170 and 320. But it has been estimated that between 2 000 and 3 000 unemployed persons could be interested in participating in this programme each year (OECD/European Union, 2017[32]).

The small size of the programme might partly result from high eligibility requirements. To qualify, a registered unemployed person needs to have some experience with business administration, a formal educational qualification at vocational level or higher either in the field of the proposed activity or in business administration, or any formal qualification at vocational level or higher combined with some further training in business administration. The business professionals who offer consultation sessions and select business plans for grants are chosen by the SEA through public procurement (OECD/European Union, 2015[37]).

Completion of the programme is associated with increased employment rates

Figure 4.19 presents employment and inactivity rates of programme participants, distinguishing between those who reached stage 1 and those who reached stage 2. Over the 24 months after the end of participation in the programme, employment rates of participants in stage 2 are initially below employment rates of participants in stage 1. However, employment rates of participants in stage 2 rise faster, overtake employment rates of participants in stage 1 after three months and then remain at levels that are sometimes substantially higher than the employment rates of participants in stage 1. After one year, employment rates of participants in stage 2 have reached 65%-70%, compared with around 60% for participants in stage 1.

Substantial shares of participants (mostly 20%-30%) also move to inactivity after the programme, rather than employment or staying unemployed. The evolution of participants’ inactivity rates broadly mirrors the evolution of employment rates: participants in stage 2 initially have a higher inactivity rate than participants in stage 1, but this eventually reverses after about 18 months. Because the selection of participants into stage 2 should primarily reflect the quality of their business plan rather than their own employability, Figure 4.19 suggests that participation in stage 2 increases the probability of subsequent employment, likely through higher self-employment. However, the evidence in Figure 4.19 is only indicative and cannot be given a causal interpretation because it notably does not account for differences in characteristics of participants.

Figure 4.19. Employment outcomes of participants in the programme for entrepreneurship and self-employment, 2012-2017

Long-term unemployed and persons with disabilities have good chances of being granted start-up subsidies

Given the apparently positive effect of participation in stage 2, it is important which participants are selected for the second stage. At both stages, three-quarters of participants are women, according to the BURVIS Database of the SEA. Participants at the first and at the second stage exhibit a similar distribution over regions, with Pieriga and Latgale being most frequent. Second-stage participants are not systematically younger than first-stage participants – more than 40% of second-stage participants are aged 35-44, compared to 33% of first-stage participants. Distributions by education are very similar, with majorities in both groups possessing a higher education level. While 31% of participants at the first stage are long-term unemployed, this share rises to 44% at the second stage. The programme for entrepreneurship and self-employment therefore appears to reach long-term unemployed persons especially well.

A better understanding of how the characteristics of participants are linked to reaching the second stage of the programme can help optimise its design, identify hurdles and target potential participants accordingly. Table 4.3 presents the results of a regression analysis that links reaching the second stage to a range of individual characteristics. Compared with long-term unemployed (the reference category), persons with an unemployment duration of up to 6 months or 6-12 months are only about half as likely to reach the second stage. Persons aged 15-24 are 1.3 times as likely to reach it as persons aged 35-44 (the reference category), but persons in other age groups are about half as likely to reach it as those aged 35-44. One of the strongest correlations is observed for persons with disabilities, who are more than twice as likely to reach the second stage as persons without disabilities. This suggests that the programme for entrepreneurship and self-employment is especially inclusive for persons with disabilities.

Registered unemployed persons who possess a secondary level of education are less than half as likely as those with other education levels to reach the second stage of the programme (Table 4.3). Compared with residents of Riga (the reference category), residents of Pieriga, Vidzeme, Latgale and Zemgale are all more than twice as likely to reach the second stage. This indicates that the programme especially mobilises potential entrepreneurs outside of Riga and Kurzeme. Finally, it is worth noting which individual characteristics appear uncorrelated with reaching the second stage (and are therefore not reported in Table 4.3). In particular, having children or not possessing Latvian citizenship seem irrelevant for reaching the second stage.

Table 4.3. Observable determinants of reaching the second stage of the programme for entrepreneurship and self-employment, 2012-2017
Logit analysis using observations on participants at first or second stage
	Compared to:	Odds ratio (rounded)
Age between 15 and 24	Age between 35 and 44	1.3
Age between 25 and 34	Age between 35 and 44	0.6
Age between 45 and 54	Age between 35 and 44	0.5
Age between 55 and 60	Age between 35 and 44	0.4
Female	Male	1.2
Married	Not married	0.7
With disabilities	Without disabilities	2.2
Receiving unemployment benefit	Not receiving it	1.4
Receiving social assistance	Not receiving it	0.1
Secondary education	Other education levels	0.4
Unemployed up to 6 months	Unemployed for 12+ months	0.4
Unemployed for 6-12 months	Unemployed for 12+ months	0.6
Resident in Pieriga	Resident in Riga	2.1
Resident in Vidzeme	Resident in Riga	2.5
Resident in Zemgale	Resident in Riga	2.1
Resident in Latgale	Resident in Riga	2.1
Note: Only results that are statistically significant at the 5% significance level are reported. Other education levels include professional secondary, vocational, professional higher and higher education.
Source: OECD analysis using linked administrative data from BURVIS (SEA), the State Social Insurance Agency (SSIA), the Social Assistance Database (SOPA) and the Population Register (OCMA).
StatLink https://doi.org/10.1787/888933962018

In OECD/European Union (2015[37]), Latvia’s programme for entrepreneurship and self-employment is assessed in some detail. Out of 377 persons who received an initial grant for their start-up or self-employment in the period 2008-2014, 71% were still in this business two years later. This survival rate is found to be comparable to survival rates observed in similar programmes in other European countries. The SEA evaluated about 6% of the businesses implemented by the programme cohorts from 2012 to 2014 as very successful, just over half performed in line with expectations, 17% were underperforming and 9% were considered a failure. The costs of the programme approached EUR 300 000 in 2014, which translates into per capita costs of about EUR 1 500 for every participant at the second stage.

The assessment by OECD/European Union (2015[37]) concludes that the programme for entrepreneurship and self-employment functions well overall, mainly due to two factors. Firstly, providing help with designing business plans in combination with financial support and secondly, the selection criteria for participants that led to a participant pool with a high incidence of relevant business experience. At the same time, the small size of the programme is highlighted. The programme procedure in two stages is considered to make the use of the funds for the programme more effective because financial support at the second stage is based on promising results at the first stage. In order to further improve the programme, OECD/European Union (2015[37]) suggests targeting older unemployed persons who can draw on long years of work experience.

Start-up subsidies create rare opportunities in regions with poor job prospects

A small literature has examined similar programmes promoting start-ups or self-employment for unemployed persons in other OECD countries. Criteria used in this context include whether participants are still in (some form of) employment after the programme has ended, their income, and for how long start-ups survive. The existing evaluations have typically found positive effects, as measured by these criteria after accounting for other factors (see for example Wolff and Nivorozhkin (2012[38]), Caliendo et al. (2016[39]). As such programmes might be suspected to generate exits from unemployment into short-lived self-employment, it is important that further evaluations also find positive long-term effects, especially for disadvantaged groups among the unemployed (Caliendo and Künn (2011[40]), Wolff et al. (2016[41])). These positive effects manifest among unemployed persons although the experience of losing a job appears to decrease a person’s willingness to take risks (Hetschko and Preuss, 2015[42]).

Caliendo and Künn (2014[43]) investigate how the effectiveness of ALMP programmes for self-employment or start-ups differs across regions. They found that these programmes can be especially effective where job prospects are relatively poor. This was not driven primarily by the outcomes of the programme participants – which appeared slightly better in regions with favourable economic conditions – but by the especially weak outcomes of non-participants in regions with poor job prospects. These results confirm similar findings that start-up subsidies can reliably create employment options where few others are available. Given studies that emphasise the role of entrepreneurship for lagging regions (e.g. Stephens et al. (2013[44])), programmes that promote entrepreneurship can contribute to both the short-term objective of reducing unemployment and the long-term objective of regional development.

In conclusion, the labour market situation can differ strongly between Latvia’s regions and presents particular challenges for ALMP in regions that are relatively far from Riga. While the regional mobility of unemployed persons in Latvia does not seem low in international comparison, rapidly rising costs of housing in several regions likely discourage some mobility and young unemployed persons exhibit a comparatively low willingness to commute. Against this background, the programme for regional mobility appears to have raised the mobility of registered unemployed persons to take up employment. This notably includes the mobility of social assistance recipients while the evidence is mixed for persons with disabilities and residents of Latgale. Self-employment has risen in Latvia over recent years and significantly more firms are created than closed, also outside urban areas. Latvia’s infrastructure offers possibilities for start-ups in the digital economy across regions. Unemployed persons are themselves a potential reservoir of entrepreneurs, but significantly fewer move into self-employment than indicate a wish to do so. In this context, the well-designed programme for entrepreneurship and self-employment could help mobilise the potential among unemployed persons at a larger scale. Its procedures also offer good chances of obtaining a start-up grant to participants with disabilities, the long-term unemployed and residents of regions outside Riga.

References

[25] Athey, S. and G. Imbens (2006), “Identification and inference in nonlinear difference-in-differences models”, Econometrica, Vol. 74/2, pp. 431-497, https://doi.org/10.1111/j.1468-0262.2006.00668.x.

[12] Bell, S. et al. (2009), “Rural Society, Social Inclusion and Landscape Change in Central and Eastern Europe: A Case Study of Latvia”, https://doi.org/10.1111/j.1467-9523.2009.00480.x.

[15] Boeri, T. and S. Scarpetta (1996), “Regional mismatch and the transition to a market economy”, Labour Economics, Vol. 3, pp. 233-254, https://doi.org/10.1016/S0927-5371(96)00017-6.

[16] Burda, M. and S. Profit (1996), “Matching across space: Evidence on mobility in the Czech Republic”, Labour Economics, Vol. 3/3, pp. 255-278, https://doi.org/10.1016/S0927-5371(96)00013-9.

[43] Caliendo, M. and S. Künn (2014), “Regional Effect Heterogeneity of Start-up Subsidies for the Unemployed”, Regional Studies, Vol. 48/6, pp. 1108-1134, https://doi.org/10.1080/00343404.2013.851784.

[40] Caliendo, M. and S. Künn (2011), “Start-up subsidies for the unemployed: Long-term evidence and effect heterogeneity”, Journal of Public Economics, Vol. 95/3-4, pp. 311-331, https://doi.org/10.1016/J.JPUBECO.2010.11.003.

[28] Caliendo, M., S. Künn and R. Mahlstedt (2017), “Mobility Assistance Programmes for Unemployed Workers, Job Search Behaviour and Labour Market Outcomes”, IZA Discussion Paper, No. 11169, http://ftp.iza.org/dp11169.pdf.

[27] Caliendo, M., S. Künn and R. Mahlstedt (2017), “The return to labor market mobility: An evaluation of relocation assistance for the unemployed”, Journal of Public Economics, Vol. 148, pp. 136-151, https://doi.org/10.1016/J.JPUBECO.2017.02.008.

[39] Caliendo, M., S. Künn and M. Weißenberger (2016), “Personality traits and the evaluation of start-up subsidies”, European Economic Review, Vol. 86, pp. 87-108, https://doi.org/10.1016/J.EUROECOREV.2015.11.008.

[20] Cannari, L., F. Nucci and P. Sestito (2010), “Geographic labour mobility and the cost of housing: evidence from Italy”, https://doi.org/10.1080/000368400425116.

[3] European Commission (2013), Executive Summary: Peer Review on Performance management in Public Employment Services, https://ec.europa.eu/social/main.jsp?catId=89&langId=en&newsId=1827.

[31] Fackler, D. et al. (2018), “Do Start-ups Provide Employment Opportunities for Disadvantaged Workers?”, ILR Review, pp. 1-26, https://doi.org/10.1177/0019793918814476.

[30] Hämäläinen, K. (2002), “Unemployment, selective employment measures and inter-regional mobility of labour”, Papers in Regional Science, Vol. 81/4, pp. 423-441, https://doi.org/10.1111/j.1435-5597.2002.tb01242.x.

[23] Hämäläinen, K. and P. Böckerman (2004), “Regional Labor Market Dynamics, Housing, and Migration”, Journal of Regional Science, Vol. 44/3, pp. 543-568, https://doi.org/10.1111/j.0022-4146.2004.00348.x.

[5] Hazans, M. (2010), “Ethnic Minorities in the Latvian Labour Market, 1997–2009: Outcomes, Integration Drivers and Barriers”, in Muižnieks, N. (ed.), How Integrated is Latvian Society? An Audit of Achievements, Failures and Challenges, University of Latvia Press, Riga, https://www.lu.lv/fileadmin/user_upload/lu_portal/zinas/Integracija_anglu.pdf.

[13] Hazans, M. (2004), “Does Commuting Reduce Wage Disparities?”, Growth and Change, Vol. 35/3, pp. 360-390, https://doi.org/10.1111/j.1468-2257.2004.00253.x.

[11] Hazans, M. (2003), “Determinants of Inter-Regional Migration in the Baltic Countries”, ZEI Working Paper, No. B17-2003, https://doi.org/10.2139/ssrn.742184.

[42] Hetschko, C. and M. Preuss (2015), “’Income in Jeopardy: How Losing Employment Affects the Willingness to Take Risks’”, SOEP paper, No. 813, https://doi.org/10.2139/ssrn.2725934.

[18] Hofmann, B. (2015), “Do Geographical Mobility Requirements for the Unemployed Affect their Exit Rate to Work? Evidence from a Policy Change”, ILR Review, Vol. 68/5, pp. 1195-1219, https://doi.org/10.1177/0019793915592376.

[17] Hunt, J. (2006), “Staunching Emigration from East Germany: Age and the Determinants of Migration”, Journal of the European Economic Association, Vol. 4/5, pp. 1014-1037, https://doi.org/10.1162/JEEA.2006.4.5.1014.

[29] Le Gallo, J., Y. L’Horty and P. Petit (2017), “Does enhanced mobility of young people improve employment and housing outcomes? Evidence from a large and controlled experiment in France”, Journal of Urban Economics, Vol. 97, pp. 1-14, https://doi.org/10.1016/J.JUE.2016.10.003.

[6] Lindemann, K. (2014), “The Effects of Ethnicity, Language Skills, and Spatial Segregation on Labour Market Entry Success in Estonia”, European Sociological Review, Vol. 30/1, pp. 35-48, https://doi.org/10.1093/esr/jct020.

[9] Masso, J. and K. Krillo (2011), “Labour markets in the Baltic states during the crisis 2008-2009: the effect on different labour market groups”, Faculty of Economics and Business Administration Working Paper, No. 79, University of Tartu, https://doi.org/10.2139/ssrn.1734378.

[2] OECD (2018), OECD Regions and Cities at a Glance 2018, OECD Publishing, Paris, https://dx.doi.org/10.1787/reg_cit_glance-2018-en.

[33] OECD (2018), “Responding to business needs: Fostering a doing-business enabling framework in Latvia”, in Access to Justice for Business and Inclusive Growth in Latvia, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264303416-5-en.

[36] OECD (2017), “Building entrepreneurial capacity in Poland through teaching and learning”, in Supporting Entrepreneurship and Innovation in Higher Education in Poland, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264270923-8-en.

[19] OECD (2017), OECD Economic Surveys: Latvia 2017, OECD Publishing, Paris, https://dx.doi.org/10.1787/eco_surveys-lva-2017-en.

[35] OECD (2017), “Self-employment and entrepreneurship by the unemployed”, in The Missing Entrepreneurs 2017: Policies for Inclusive Entrepreneurship, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264283602-10-en.

[4] OECD (2016), “A volatile economy heightens Latvia’s social challenges”, in OECD Reviews of Labour Market and Social Policies: Latvia 2016, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264250505-5-en.

[1] OECD (2016), OECD Regions at a Glance 2016, OECD Publishing, Paris, https://dx.doi.org/10.1787/reg_glance-2016-en.

[32] OECD/European Union (2017), Inclusive Entrepreneurship Policies, Country Assessment Notes. Latvia, 2017, OECD Publishing, Paris, http://www.oecd.org/industry/smes/LATVIA-country-note-2017.pdf.

[37] OECD/European Union (2015), Entrepreneurship support for the unemployed in Latvia, https://www.oecd.org/latvia/Rapid-Policy-Assesment-Latvia.pdf.

[14] Paci, P. et al. (2007), “Internal Labor Mobility in Central Europe and the Baltic Region”, World Bank Working Paper, No. 105, World Bank Group, https://doi.org/10.1596/978-0-8213-7090-2.

[21] Plantinga, A. et al. (2013), “Housing prices and inter-urban migration”, Regional Science and Urban Economics, Vol. 43/2, pp. 296-306, https://doi.org/10.1016/J.REGSCIURBECO.2012.07.009.

[24] Rabe, B. and M. Taylor (2012), “Differences in Opportunities? Wage, Employment and House-Price Effects on Migration”, Oxford Bulletin of Economics and Statistics, Vol. 74/6, pp. 831-855, https://doi.org/10.1111/j.1468-0084.2011.00682.x.

[34] Revoltella, D. et al. (2019), “Framework conditions, innovation and productivity in European regions”, European Investment Bank.

[10] Rogers, C. (1997), “Job Search and Unemployment Duration: Implications for the Spatial Mismatch Hypothesis”, Journal of Urban Economics, Vol. 42/1, pp. 109-132, https://doi.org/10.1006/JUEC.1996.2017.

[44] Stephens, H., M. Partridge and A. Faggian (2013), “Innovation, Entrepreneurship and Economic Growth in Lagging Regions”, Journal of Regional Science, Vol. 53/5, pp. 778-812, https://doi.org/10.1111/jors.12019.

[7] Toomet, O. (2011), “Learn English, not the local language! Ethnic Russians in the Baltic states”, American Economic Review, Vol. 101/3, pp. 526-531, https://doi.org/www.dx.doi.org/10.1257/aer.101.3.526.

[22] Vermeulen, W. and J. van Ommeren (2009), “Does land use planning shape regional economies? A simultaneous analysis of housing supply, internal migration and local employment growth in the Netherlands”, Journal of Housing Economics, Vol. 18/4, pp. 294-310, https://doi.org/10.1016/J.JHE.2009.09.002.

[26] Westerlund, O. (1998), “Internal Migration in Sweden: The Effects of Mobility Grants and Regional Labour Market Conditions”, Labour, Vol. 12/2, pp. 363-388, https://doi.org/10.1111/1467-9914.00072.

[38] Wolff, J. and A. Nivorozhkin (2012), “Start Me Up: The Effectiveness of a Self-employment Programme for Needy Unemployed People in Germany”, Journal of Small Business & Entrepreneurship, Vol. 25/4, pp. 499-518, https://doi.org/10.1080/08276331.2012.10593586.

[41] Wolff, J., A. Nivorozhkin and S. Bernhard (2016), “You can go your own way! The long-term effectiveness of a self-employment programme for welfare recipients in Germany”, International Journal of Social Welfare, https://doi.org/10.1111/ijsw.12176.

[8] Zvaigzne, A., I. Saulāja and A. Čerpinska (2015), “Youth Unemployment and its Main Causes in Latgale Region”, Latgale National Economy Research, Vol. 1/7, pp. 217-227, https://doi.org/10.17770/lner2015vol1.7.1191.

Database references

BURVIS Database (State Employment Agency).

Central Statistical Bureau of Latvia, http://www.csb.gov.lv/en.

Central Statistical Bureau of Latvia, https://migracija.csb.gov.lv/?id=B1e0JusADQ.

European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview.

European Labour Force Survey (Eurostat) ad-hoc module 2016 on young people on the labour market, http://ec.europa.eu/eurostat/statistics-explained/index.php/EU_labour_force_survey_-_ad_hoc_modules.

Gallup World Poll data, http://www.oecd.org/std/43017172.pdf.

Latvian Labour Force Survey (CSB), https://www.csb.gov.lv/en/statistics/statistics-by-theme/social-conditions/unemployment/tables/metadata-employment-and-unemployment.

Linked administrative data from the BURVIS Database (SEA), State Social Insurance Agency (SSIA), the Social Assistance Database (SOPA) and the Population Register (OCMA).

Mobility Survey of Latvia’s Population (CSB), https://www.csb.gov.lv/en/statistics/statistics-by-theme/transport-tourism/transport/search-in-theme/357-results-survey-mobility-survey-latvia.

OECD Broadband Portal, www.oecd.org/sti/broadband/oecdbroadbandportal.htm.

OECD Analytical house prices indicators and Residential Property Price Indices Dataset, https://stats.oecd.org/Index.aspx?DataSetCode=HOUSE_PRICES.

State Employment Agency of Latvia, http://www.nva.gov.lv/.

Annex 4.A. Additional table

Annex Table 4.A.1. List of Latvian municipalities considered as indirectly adjacent
Novads	Code	Indirectly adjacent novads	Code	Novads	Code	Indirectly adjacent novads	Code
Aizputes novads	640600	Alsungas novads	624200	Preiļu novads	760202	Krustpils novads	566900
Alsungas novads	624200	Aizputes novads	640600	Priekules novads	641600	Skrundas novads	621200
Amatas novads	424701	Priekuļu novads	427300	Priekuļu novads	427300	Amatas novads	424701
Babītes novads	804900	Engures novads	905100	Pārgaujas novads	427500	Siguldas novads	801601
Baldones novads	800600	Ikšķiles novads	740600	Pāvilostas novads	641401	Liepāja	170000
Baldones novads	800600	Salaspils novads	801200	Raunas novads	427700	Beverīnas novads	964700
Beverīnas novads	964700	Valkas novads	940200	Riebiņu novads	766300	Vārkavas novads	769101
Beverīnas novads	964700	Raunas novads	427700	Riga	10000	Ādažu novads	804400
Carnikavas novads	805200	Sējas novads	809200	Riga	10000	Ropažu novads	808400
Cesvaines novads	700800	Jaunpiebalgas novads	425700	Ropažu novads	808400	Ādažu novads	804400
Ciblas novads	684901	Zilupes novads	681801	Ropažu novads	808400	Riga	10000
Engures novads	905100	Talsu novads	880200	Rugāju novads	387500	Madonas novads	700200
Engures novads	905100	Babītes novads	804900	Rugāju novads	387500	Rēzeknes novads	780200
Garkalnes novads	806000	Salaspils novads	801200	Rēzekne	210000	Ludzas novads	680200
Garkalnes novads	806000	Sējas novads	809200	Rēzeknes novads	780200	Rugāju novads	387500
Iecavas novads	406400	Jelgavas novads	540200	Rēzeknes novads	780200	Lubānas novads	701400
Iecavas novads	406400	Ķekavas novads	800800	Rūjienas novads	961600	Valkas novads	940200
Ikšķiles novads	740600	Stopiņu novads	809600	Salaspils novads	801200	Olaines novads	801000
Ikšķiles novads	740600	Baldones novads	800600	Salaspils novads	801200	Baldones novads	800600
Ilūkstes novads	440801	Līvānu novads	761201	Salaspils novads	801200	Garkalnes novads	806000
Jaunjelgavas novads	321000	Lielvārdes novads	741401	Siguldas novads	801601	Pārgaujas novads	427500
Jaunpiebalgas novads	425700	Cesvaines novads	700800	Skrundas novads	621200	Priekules novads	641600
Jelgava	90000	Olaines novads	801000	Stopiņu novads	809600	Ķekavas novads	800800
Jelgavas novads	540200	Jūrmala	130000	Stopiņu novads	809600	Ikšķiles novads	740600
Jelgavas novads	540200	Iecavas novads	406400	Sējas novads	809200	Carnikavas novads	805200
Jēkabpils novads	560200	Vārkavas novads	769101	Sējas novads	809200	Garkalnes novads	806000
Jūrmala	130000	Jelgavas novads	540200	Talsu novads	880200	Engures novads	905100
Jūrmala	130000	Mārupes novads	807600	Valkas novads	940200	Beverīnas novads	964700
Krustpils novads	566900	Preiļu novads	760202	Valkas novads	940200	Rūjienas novads	961600
Kārsavas novads	681000	Ludzas novads	680200	Vecumnieku novads	409500	Ķekavas novads	800800
Lielvārdes novads	741401	Jaunjelgavas novads	321000	Vecumnieku novads	409500	Lielvārdes novads	741401
Liepāja	170000	Pāvilostas novads	641401	Vārkavas novads	769101	Jēkabpils novads	560200
Lubānas novads	701400	Rēzeknes novads	780200	Vārkavas novads	769101	Riebiņu novads	766300
Ludzas novads	680200	Kārsavas novads	681000	Zilupes novads	681801	Ciblas novads	684901
Ludzas novads	680200	Rēzekne	210000	Ādažu novads	804400	Riga	10000
Līvānu novads	761201	Ilūkstes novads	440801	Ādažu novads	804400	Ropažu novads	808400
Madonas novads	700200	Rugāju novads	387500	Ķekavas novads	800800	Vecumnieku novads	409500
Mārupes novads	807600	Jūrmala	130000	Ķekavas novads	800800	Iecavas novads	406400
Mārupes novads	807600	Ķekavas novads	800800	Ķekavas novads	800800	Ogres novads	740202
Ogres novads	740202	Ķekavas novads	800800	Ķekavas novads	800800	Stopiņu novads	809600
Olaines novads	801000	Jelgava	90000	Ķekavas novads	800800	Mārupes novads	807600
Olaines novads	801000	Salaspils novads	801200
Source: OECD Secretariat analysis.

End of the section – Back to iLibrary publication page