Skills, employment and automation

This chapter examines recent trends in labour markets across the EBRD regions, looking at the types of job that have been destroyed and created, future trends in job creation, the skills mix in the various economies and the economic impact of emigration by skilled workers. On the basis of the available evidence, it also discusses policies that can help to mitigate the disruptive effects of technological change and emigration by skilled workers.

The chapter starts by looking at skills from a supply and demand perspective. Like advanced economies, most countries in the EBRD regions have been experiencing deindustrialisation (a decline in industrial employment) and job polarisation (a decline in medium-skilled occupations as a percentage of total jobs). Educational attainment has improved, but mismatches between the supply of skills and the demands of employers persist.

The second section examines the ways in which automation alters demand for labour and skills. On average, workers in the EBRD regions face a significantly higher risk of job automation than their counterparts in advanced economies. For example, the mean probability of a job being automated in the foreseeable future is 57 per cent in the Slovak Republic and 55 per cent in Turkey, compared with around 40 per cent in the United States of America and the United Kingdom.2

The third section looks at the impact that the emigration of skilled workers has had on innovation and the productivity of firms in the EBRD regions. Following the accession to the EU of various countries in central and south-eastern Europe in 2004 and 2007, emigration from that region has increased considerably. Emigrants tend to be young and highly skilled, with Eurostat data indicating that more than 20 per cent have university degrees and another 40 per cent are educated to secondary level. The analysis in this section shows that although skilled emigration has been associated with a decline in firms’ total factor productivity (TFP) – the efficiency with which capital, labour and material inputs are combined to produce final products – emigration has, in the medium term, boosted “knowledge remittances” to those migrants’ countries of origin and contributed to higher rates of innovation.

The EBRD regions have been undergoing deindustrialisation. Indeed, the combined employment share of agriculture, mining and utilities declined by 4 percentage points between 2006 and 2016, while that of manufacturing declined by more than 2 percentage points over the same period (see Chart 2.1). These trends are projected to continue (see Box 2.2 for a discussion of expected changes in sectoral employment and working conditions). Male workers leaving the agriculture and manufacturing industries have typically found employment in construction, while women have mainly switched to non-market services such as public administration or education.

As a result, gender segregation in employment has increased further, with men accounting for 92 per cent of all employees in the construction industry in 2016 (up from 90 per cent in 2006) and women accounting for 59 per cent of all employees in non-market services (up from 57 per cent in 2006). See Box 2.1 for a discussion of gender imbalances in unpaid care work.

The speed of such structural shifts varies across economies. Indeed, additional analysis suggests that the economies of Central Asia and the southern and eastern Mediterranean (SEMED) region have tended to see the strongest declines in the primary sector, while manufacturing has tended to decline most strongly in eastern Europe and the Caucasus (EEC) and Russia. Meanwhile, Turkey has seen a 6 percentage point increase in the employment share of business services.

In line with those sectoral shifts from agriculture and manufacturing to services, the employment shares of service-sector professionals and sales workers have risen by between 2 and 4 percentage points (see Chart 2.2). At the same time, the employment shares of elementary occupations (such as agricultural labourers), craft workers and technicians have fallen by between 0.8 and 2 percentage points. Occupational shifts have been more pronounced among women than among men.

In line with these global trends, the total employment share of medium-skilled occupations (such as clerks, craft workers, and plant and machine operators) has been declining in the EBRD regions (see Chart 2.3),5 while those of high-skilled occupations and low-skilled occupations have both been rising. The extent of such job polarisation has, on average, been greater in higher-income economies (see Annex 2.2). In Bulgaria, Croatia, Cyprus, Greece and Latvia, levels of job polarisation have been similar to those observed in advanced economies and higher than the levels seen in most emerging markets.

However, economies with large primary sectors have seen continued declines in the employment share of low-skilled occupations (with the largest declines being observed in agriculture) and increases in the shares of both medium-skilled and high-skilled jobs. This includes economies such as Azerbaijan, Moldova and Mongolia, where agriculture, mining and utilities still account for more than 35 per cent of total employment.

The number of industrial robots is on the rise around the world (see Chart 2.7), both in absolute terms and as a percentage of the number of workers employed. Indeed, the total global stock of industrial robots is projected to increase by 14 per cent a year, reaching 3 million by 2020.8 The number of industrial robots in use in the EBRD regions stood at 41,000 in 2016, up from 1,500 in 1993 (on the basis of available data for 22 EBRD countries). The vast majority of robots are deployed in manufacturing (particularly in the automotive sectors), although they have also begun to be used more widely in the production of plastic, chemicals and metals.

The extent to which robots are used in manufacturing varies greatly from country to country (see Chart 2.8). The levels seen in the Slovak Republic and Slovenia (where there are more than 93 robots for every 10,000 manufacturing workers) are comparable to those observed in advanced economies and higher than those seen in Brazil, China, India and South Africa. Hungary, Poland, Turkey and Romania have also seen rapid increases in the ratio of robots to manufacturing workers (see Chart 2.9). In contrast, Moldova, Morocco and Serbia have fewer than 2 robots for every 10,000 workers.

Differences across countries and sectors in terms of the use of robots are strongly linked to differences in FDI. Indeed, regression analysis (available on request) suggests that a 1 per cent increase in FDI in a given sector of a given country is associated with a 12 per cent increase in the use of industrial robots.

In order to assess the extent to which the decline in employment can be attributed to the increase in robotisation, regression analysis can be used to explain the causal relationship between the employment rate and the level of robotisation (see Annex 2.4 for a description of the methodology used). That analysis points to a substantial displacement effect: every additional robot per 1,000 workers reduces the employment rate by 0.7 percentage point (see column 4 of Table 2.1).

This suggests that robotisation can explain 13 per cent of the total decline that was seen in the employment rate between 2010 and 2016 in the 11 EBRD economies included in this analysis. Although it is not fully comparable owing to differences in the outcomes used and the time period covered, this estimated impact appears to be somewhat larger than the figures obtained for the United States of America and advanced European economies.9 Increases in the ICT-intensity of production also have a statistically significant negative impact on the employment rate, but the effect of a 1 standard deviation increase in ICT-intensity is half the size of that of a 1 standard deviation increase in the use of robots.

The effect of exposure to robots is strongest for people with low levels of education (see Chart 2.10). This may, however, change as technology evolves. There are no statistically significant differences between men and women or between younger and older workers in terms of the impact of robotisation.

Eurostat data indicate that around 40 per cent of migrants leaving countries that joined the EU in 2004 had medium-skilled or high-skilled occupations in their countries of origin. In the case of Bulgaria and Romania, the corresponding figure was 35 per cent.12 Emigration can raise wages in countries of origin, increase incentives for young people to invest in education, and foster cross-border flows of trade, capital and ideas.13 However, concerns are often raised about “brain drain” and its adverse impact on economic development, and it is noticeable that many of the migrants leaving those new EU member states are highly skilled.14

Firms often complain that they cannot find skilled labour to replace employees who have emigrated. Indeed, the percentage of firms reporting that a lack of skills was a major constraint on their business rose from 12 per cent in 2005 to almost 30 per cent in 2008-09 in the new EU member states, according to the representative Business Environment and Enterprise Performance Survey (BEEPS) conducted by the EBRD and the World Bank. Consequently, some scholars have argued that emigration has the potential to hinder the economic development of those countries, given the record-high labour shortages that they have experienced.

Emigration can affect firms’ TFP via three main channels. First of all, firms that are unable to fill a vacancy promptly may be forced to leave capital equipment idle for extended periods of time. Second, a lack of skilled employees may reduce knowledge sharing, thereby slowing firms’ accumulation of knowledge and hampering improvements in TFP. And third, when experienced employees leave a firm, specialist knowledge about production processes and relationships with suppliers and customers will be lost. Firms may also need to dedicate more resources to training newly hired employees. At the same time, higher rates of labour turnover may actually discourage firms from investing in training, further hampering productivity. On the other hand, an increase in labour market competition may incentivise firms to improve the quality of management and achieve higher levels of TFP.15

The emigration opportunities that were available to workers in those new EU member states post-accession varied by country of origin and sector, with the EU-15 and the four EFTA countries able to impose restrictions on such migrants for up to seven years. Those differences are captured by a labour mobility index ranging from 0 to 1, where higher values correspond to greater opportunities for emigration from a given country in a given year for workers in a specific sector (with 0 denoting a complete absence of freedom of movement for workers and 1 indicating full freedom of movement across the EU; see Annex 2.5 for details).

The regression analysis described in Annex 2.5 finds that increases in emigration opportunities had a negative impact on the TFP of firms in new member states (see Chart 2.14). Indeed, that analysis suggests that, on average, a maximum increase in emigration opportunities for workers in a firm’s country and sector (which corresponds, on average, to an increase of 0.25 in the labour mobility index) resulted in that firm’s TFP one year later being around 7.5 percentage points lower than it would have been in the absence of emigration.16

The cumulative negative impact that emigration has on productivity increases for a number of years following the year with the largest increase in the labour mobility index (which is termed “year zero” in Chart 2.15). It is worth noting that, prior to year zero, firms subsequently exposed to increased emigration were no different from other firms in terms of their productivity trends.

Firms also adjust to increased emigration opportunities for current and potential future employees by increasing personnel costs (which include wages and other employee-related costs, such as hiring and training expenses). An annual increase of 0.25 in the labour mobility index will lead to a 5 per cent increase in personnel costs per employee (resulting in an average rise of €400 per worker per year; see Chart 2.14). The impact on the capital-to-labour ratio is also positive, but not statistically significant. These results are robust to the use of alternative measures of productivity, provided that the 1 per cent of firms that are most and least productive are excluded, firms established after 2002 are excluded and the sample is restricted to the period before the 2008-09 global financial crisis.

The impact on the TFP of firms with at least one foreign owner and firms with at least one patent is smaller and statistically insignificant (see Chart 2.14). This may be because these firms are able to spend significantly more on personnel. In addition, innovative firms (those holding patents) increase their capital-to-labour ratios to a significantly greater extent. Innovative firms may also be in a better position to offer their employees a stimulating working environment, as well as financial incentives conditional on innovations paying off. In the longer term, innovative firms may also benefit from inflows of knowledge facilitated by emigrants working abroad in advanced economies, as discussed in the next subsection.

In order to see which of those two opposing effects of emigration dominates, the analysis below looks at links between emigration and the frequency with which patents filed in migrants’ countries of origin cite patents filed by inventors in migrants’ destination countries, with such cross-border patent citations representing a useful measure of knowledge flows between countries. In addition, that analysis also looks at rates of convergence in terms of the numbers of patents that are filed by inventors in countries of origin and destination countries.

It is noticeable, for example, that the number of cross-border citations in patent applications increased significantly following the introduction of free movement of labour between new EU member states and the EU-15 and the four EFTA countries (see Chart 2.16).

Furthermore, Chart 2.17 shows that the frequency with which patents filed in new EU member states cite patents filed in EU-15 and EFTA countries increases three to four years after the introduction of free movement of labour. At the same time, only a small part of that increase is driven by direct communication between migrant inventors and their former colleagues. If we exclude citations where an inventor cites a patent filed by someone with whom he or she has already filed one or more patents, the estimated impact declines slightly, but remains statistically significant, pointing to the importance of broader spillover effects within individual industries.

Importantly, migration from new EU member states to advanced European economies is also associated with a subsequent convergence in the numbers of patents filed by the two groups of countries. On average, firms in advanced European economies file more than three times the number of patents filed by firms operating in the same industries in new member states. Patents filed in advanced economies are also cited more in other patents and can therefore be regarded as being of higher quality. Migration has, on average, contributed to a narrowing of this gap, both in terms of the total number of patents filed and in terms of quality-adjusted measures of patenting (see Chart 2.17). This analysis suggests, therefore, that migration from new EU member states to advanced European economies has, on balance, resulted in a strengthening, rather than a weakening, of innovation in migrants’ countries of origin.

Unpaid care work by women is often regarded as a cost-free alternative to the provision of care services by the state or the private sector. Care work is broadly defined as activities involved in meeting the physical, psychological and emotional needs of children, the elderly and the disabled.21 Unpaid care and domestic work comprises basic household activities (such as cooking, fetching water and cleaning), plus unremunerated direct care for family and community members.

In the EBRD regions, women spend, on average, 3 hours more per day on unpaid work than men. In SEMED economies, women spend more than 5 hours per day on unpaid work (almost 5.5 hours in the case of Tunisia). Consequently, women’s total working hours (including both paid and unpaid work) far exceed men’s. Effectively, women have less time to dedicate to paid work.

When women do enter the labour market, it is rare for their care-related responsibilities to be sufficiently redistributed. As a means of reconciling unpaid care work and paid employment, women may accept part-time, informal, home-based work, which may be less well-paid. Women also tend to reduce their hours of paid work with each additional child, and mothers are overrepresented in part-time jobs. Data from the 2010 Labour Force Survey, which was conducted in 27 EU countries, indicate that 36 per cent of women with children work part-time, compared with 21 per cent of women without children.22

Part-time work can help women to combine paid work with care-related responsibilities, but this comes at a cost. Mothers who work part-time are much less likely to have managerial or professional occupations and more likely to have elementary occupations or be employed in the service industry. The fact that women are overrepresented in certain sectors and occupations contributes to gender wage gaps that undervalue women’s labour and inflate the numbers of working poor.23 Globally, the gender pay gap is estimated at around 20 per cent, and it is higher for workers with childcare responsibilities.24

Better public care services will allow more women to enter the labour force or increase the number of hours they work. This, in turn, will have a positive effect on economic growth through increased consumption in two-earner households, as well as through job creation in the care sector. Two studies in Australia have found that if the price of childcare is reduced by 50 per cent, labour supplied by young mothers rises by between 6.5 and 10 per cent.25 Similarly, an increase in the subsidised provision of childcare, including public pre-primary school facilities, has been shown to have a positive and statistically significant impact on mothers’ labour supply in Canada and Argentina.26 In the same vein, a programme that increased free public childcare facilities in low-income neighbourhoods of Rio de Janeiro resulted in use of care increasing from 51 to 94 per cent, maternal employment rising from 36 to 46 per cent and employment rates almost doubling among mothers who did not work before giving birth.27 Moreover, ILO analysis indicates that countries which invest more in care enjoy labour force participation rates of 70 per cent or more among women – significantly higher than the global average of 55 per cent.28

If women withdrew from unpaid care work, there would be a significant impact on both the labour market and GDP through the hiring of new care workers and through the new jobs performed by women (who had previously performed such tasks free of charge within their households). One way of estimating that impact is to calculate the number of hours that men and women in each age group spend on unpaid work each year and multiply that figure by an hourly wage rate that is assumed to be equal for men and women.

When estimated in this way, the size of the care economy ranges from 32 per cent of GDP in Bulgaria to 47 per cent of GDP in the Kyrgyz Republic. The care economy tends to be smaller where the labour force participation rate of women is higher, as women who stay at home tend to devote more time to care work. Thus, estimates of GDP would be substantially higher if the care economy was fully accounted for.

Recognising and valuing unpaid care work is central to addressing gender gaps in the labour market and closing the gender wage gap.

This box analyses expected changes in employment relationships and working conditions across the EBRD regions over the period 2017-25, looking at aggregate sectors on the basis of the methodology developed in ILO (2018b).

Most economies in the EBRD regions are expected to experience substantial declines in the percentage of people working in the agricultural sector, with that sector’s employment share falling by between 0.7 and 4 percentage points. That shift out of agriculture is expected to be somewhat less pronounced in central Europe and the Baltic states (CEB), as well as in Russia. In high-income comparator countries (the Czech Republic, France, Germany, Italy, Sweden and the United Kingdom), that shift is expected to be more limited (totalling 0.2 percentage points on average), reflecting the fact that advanced economies already have small agricultural sectors.

Unlike in the past, the structural shift away from agriculture is not expected to be accompanied by rising employment in manufacturing. Instead, employment in market services (such as wholesale and retail trade, transport and financial activities) is projected to increase further. Within market services, real estate and business and administration are projected to generate the most employment in the near future, followed by transport, storage and telecommunications, as well as restaurants and accommodation. The employment share of non-market services (such as public administration, education and healthcare) is also expected to rise, particularly in EU member states.

At the level of the EBRD regions, a significant decline is expected in the incidence of vulnerable employment (defined as self-employment and contributing to domestic work), which will, among other things, reduce the probability of requiring social protection (see Chart 2.2.1). As such working conditions are prevalent in agriculture, the decline in that sector’s employment share across the EBRD regions will translate into an overall reduction in the incidence of vulnerable employment. The total percentage of employment across these economies that is regarded as vulnerable is expected to fall by around 1 percentage point to stand at 19 per cent.

Structural shifts are expected to have a similar – albeit smaller – impact on the prevalence of part-time employment, temporary employment and time-related underemployment (where workers put in fewer hours than a specified threshold and would be willing to work additional hours). At the same time, the incidence of part-time employment and time-related underemployment could rise slightly in the CEB region, owing to the projected decline in the employment share of manufacturing – a sector typically characterised by full-time jobs with relatively stable hours – in those countries.

This problem is most acute among young people. Indeed, according to ILO data, more than a quarter of all people between the ages of 15 and 24 in the SEMED region are not in employment, education or training. Moreover, people educated to tertiary level account for 31 per cent of youth unemployment in Egypt and the West Bank and Gaza, and as much as 37 per cent in Tunisia, based on the latest available data from the ILO. Young women in SEMED economies tend to study for longer and gain higher qualifications than their male counterparts, but their employment outcomes tend to be much worse, especially when it comes to competing for private-sector jobs. Unsurprisingly, many young women end up leaving the labour market prematurely and forgoing further training.

The SEMED economies are among the most youthful in the EBRD regions, but they are also some of the economies that are struggling most to equip their populations with the right skills (as discussed in Chapter 1).

There are three different factors perpetuating skills mismatches in the SEMED region. First of all, shortcomings in primary and secondary education are ultimately resulting in demand for skills remaining unmet. Though levels of enrolment in compulsory education are high, approaching 100 per cent in some economies, teaching continues to rely excessively on passive engagement and rote learning. Outdated classroom resources are also obstacles to effective learning. Some pupils get by with the aid of costly after-school tuition, while others disengage altogether but remain technically enrolled on paper.

As a result, a considerable number of secondary school leavers fail to gain even a rudimentary understanding of key subject areas. Secondary school pupils in Jordan, Lebanon and Tunisia were among the worst performers in the OECD’s 2015 PISA tests, which assessed the skills of 15-year-olds in 73 different economies. Across those three economies, between 60 and 75 per cent of pupils were classified as low achievers in mathematics; 46 to 72 per cent were regarded as low achievers in reading; and between 50 and 66 per cent of pupils were classed as low achievers in science. Furthermore, employers in Egypt, Morocco and Tunisia are among the least satisfied in the world with the quality of their countries’ basic education, according to the World Economic Forum’s Global Competitiveness Index 2017-2018.

Second, vocational education and training does not respond to employers’ needs. In some cases, employers lack the initiative required to properly communicate their needs. And in other cases, providers of training may be too small to establish mutually beneficial partnerships with the business community. Whatever the cause, such dissonance is hindering improvements in the relevance and quality of training.

With EBRD support, sector skills councils (SSCs), such as the SSC for hospitality and tourism in Jordan, are now bringing together business leaders, education providers and policymakers. SSCs seek to improve vocational training at sector level by helping to establish feedback mechanisms between employers and training providers and ensuring that business needs are adequately reflected in sector-level qualifications frameworks.

More broadly, new opportunities for dual learning – whereby technical and vocational programmes combine traditional classroom-based learning with accredited experience in a business setting – represent promising developments. Dual learning enables businesses and education providers to work together to ensure that learning outcomes meet the industry’s expectations. Such schemes enable learners to gain practical experience and develop business contacts.

Third, institutions such as national qualifications frameworks and labour market information systems are often underdeveloped or simply missing. Such institutions help to benchmark employers’ needs against workers’ skills and training outcomes. They also facilitate the forecasting of skill-related needs in the light of technological change. In the absence of such frameworks, potential labour market entrants lack clarity about demand for skills, which makes it harder to make informed decisions about education and training. The establishment of national qualifications frameworks and labour market information systems underpinned by coordinated private-sector involvement could go a long way towards addressing the issue of skills mismatches.

In 2018, the EBRD’s Legal Transition team conducted a study analysing the regulatory frameworks governing lending-based and investment-based crowdfunding platforms in Austria, the Dubai International Financial Centre (DIFC), France, Germany, the United Kingdom and the United States of America.

That group of jurisdictions offers a considerable degree of variety in terms of geographical location, market maturity and the approach to regulating crowdfunding. The UK and the USA are world leaders in this area, with effective regulatory regimes underpinning highly developed markets, while Austria, France and Germany (as civil law jurisdictions) and the DIFC are, in the EBRD’s experience, regarded as key points of reference for the EBRD regions and provide a different perspective on the various issues.

That EBRD study looked at (i) the types of authorisation that are required in order to operate crowdfunding platforms, (ii) capital and liquidity requirements, (iii) know-your-customer (KYC) and anti-money-laundering checks, (iv) restrictions on the size of loans and investable amounts, (v) consumer protection measures (including disclosures to investors), (vi) warnings regarding risks, and (vii) due diligence and pre-funding checks. The study culminated in a report published on 24 October 2018,29 which made recommendations regarding best practices for the regulation of lending-based and investment-based crowdfunding platforms.

Several economies in the EU and in the EBRD regions have adopted (or are in the process of adopting) bespoke crowdfunding legislation, tailoring their regulatory frameworks to the needs of local markets. There is no consensus, however, as to what constitutes best practice in terms of the regulation of crowdfunding.

The EBRD’s study helps to fill this gap, providing guidance on best practice in this area, and is applicable not only in the EBRD regions, but also in other countries around the world.

Besides providing an alternative source of financing, crowdfunding also offers a number of other benefits (see Table A.2.1.1 for an overview). It can, for example, help to attract other sources of funding, such as “business angels” or venture capital, with a recent study reporting that 45 per cent of angel investment in the UK takes place alongside crowdfunding investment.30 Indeed, the presence of crowdfunding investment can help to validate the concept and finances of a novel project. It can also provide access to a large number of interested individuals, giving the entrepreneur valuable insights and information. And if a crowdfunding campaign is successful, it can act as a useful marketing tool. There is also evidence that crowdfunding may help women entrepreneurs to obtain investment funding from other women, in situations where women tend to find it difficult to secure financing for their entrepreneurial endeavours.31

The main conclusions of the EBRD’s analysis were as follows:

• Where platforms’ activities are aligned with other regulated activities, it may be possible to regulate crowdfunding by adapting an existing framework. However, a truly bespoke regime may be more appropriate.
• Imposing minimum capital requirements on platforms can help to ensure that operational and compliance costs continue to be covered in the event of financial distress. Capital requirements should be based on the nature and scale of the activities undertaken by the relevant platform and should be commensurate with the attendant risk.
• Platforms should be required to establish and maintain risk management systems and controls that can identify, track, report, manage and mitigate risks to their business (including operational risk, risks relating to cybersecurity and the protection of personal data, and the risk that the platform could be used to commit financial crimes).
• Platforms need to ensure that their senior management and employees are “fit and proper persons” to perform their roles. Platforms need to be able to assess this themselves.
• Platforms should have primary responsibility for identifying, reporting, managing and mitigating any conflicts of interest.
• The financial services regulator should, where appropriate, have the power to prevent platforms from investing in the projects they feature.
• Platforms should be subject to specific disclosure requirements, in order to ensure that investors and investees understand how platforms operate and earn revenue.
• Disclosures to investors and warnings regarding risks need to be tailored to the relevant product offered by the platform.
• There may be good reasons to differentiate between retail investors and institutional investors when it comes to providing information. Retail investors may benefit from receiving risk warnings and disclosures that are more explicit than those provided to institutional investors.
• A regime which differentiates between different types of investor is preferable to one that requires detailed suitability checks for all investors. Financial services regulators are best placed to decide on appropriate categories of investor.
• Platforms should be required to enter into agreements with their clients governing all key aspects of the client-platform relationship.
• Platforms should be permitted – but not necessarily obliged – to offer automated tools supporting the diversification of investors’ portfolios.
• It is appropriate for lending-based platforms to provide information to investors on their post-investment arrangements and arrangement rights, whether that involves a trustee-type arrangement or a different type of enforcement mechanism.
• Platforms should be required to carry out KYC checks on their clients. The extent of those checks may vary on the basis of a risk assessment performed by the platform. Financial services regulators are best placed to provide platforms with guidance in this regard, which should be in keeping with the KYC requirements of the relevant jurisdiction. Such guidance should also be commensurate with the risk posed by clients.

The recommendations set out above cover just some of the best practices identified by the study, which the EBRD believes would be beneficial in the vast majority of jurisdictions. For a full list of those best practices, as well as details of the optional additional measures identified, see EBRD (2018).

Indeed, the EBRD’s Legal Transition team is already assisting the Capital Markets Board of Turkey and the Astana International Financial Centre in Kazakhstan with the establishment of new crowdfunding regulations, in line with the best practices identified by the study. With various other countries expressing an interest in fostering the development of crowdfunding (and fintech more broadly), similar work is expected to be carried out in other jurisdictions across the EBRD regions in due course.

Those legislation-based dummy variables are, in turn, multiplied by a measure of labour shortages in a given industry in a destination country. This measure captures the percentage of firms that are constrained by the unavailability of labour according to a European Commission business survey.

The resulting variable captures not only legislative barriers to migration, but also implicit industry-specific demand for migrants in destination countries. The weighted average of these variables can be used to measure exposure to emigration for each industry in a given country of origin. The weights applied to potential destination countries are based on the distance between the largest cities in the two countries in question. For instance, firms in Estonia are more likely to be exposed to emigration when Finland’s labour market opens up, rather than when restrictions are lifted in Malta.

In this instance, variation in the rules governing labour market access was driven mainly by the political landscapes in the various destination countries and is unlikely to have been related to the performance of firms in the countries of origin of potential migrants.36 It can therefore be used to establish the causal impact that emigration had on firms’ performance.

The resulting labour mobility index (FLM) is standardised such that it ranges from 0 to 1 and can be used as follows in reduced-form regressions capturing the intention-to-treat effect:

Dependent variables are measures of firms’ performance. Control variables include the age of the firm , as well as industry-specific variables  such as total investment and average mark-up (the ratio of revenues to costs) which reflect common trends in terms of competition or demand for labour that may affect firms’ performance. are industry-specific controls measured at the aggregate EU level. These include total sales and labour shortages. Country-level controls  include an EU membership dummy, GDP and FDI inflows. Specifications include firm and time fixed effects. Standard errors are clustered at the country-industry level (two-digit NACE37 level).

Firm-level TFP is calculated using a TFP index, as in Gorodnichenko and Schnitzer (2013), or using a semi-parametric approach in line with Levinsohn and Petrin (2003). The latter method accounts for the fact that unobserved productivity shocks may affect firms’ inputs and performance at the same time. The return on assets (the ratio of earnings before interest and tax to total sales) is used as an alternative measure of productivity.

The specifications include multiple lags aimed at capturing events several years after the introduction of free movement of labour. In this case, the labour mobility index is a dummy variable indicating whether a specific industry in a destination country was open to migrants from a given country of origin in a given year. Specifications also control for (i) the total number of patent applications in a given industry in the country of origin, (ii) the lagged number of patent applications in a given industry in the destination country, (iii) bilateral industry-specific and country-wide FDI and trade flows (using the same lags as for the FLM index), (iv) the total number of patents in a given industry at the EU level, (v) a dummy variable indicating pairs of countries that are both EU members and (vi) time and origin-destination-industry fixed effects. Standard errors are clustered at the origin-destination-industry level (two-digit NACE level).