CAN FOUNDATIONAL ECONOMY SAVE REGIONS IN CRISIS?

We perform an explorative analysis of employment patterns in the foundational economy producing mundane everyday necessities and providing welfare services across Swedish regional labour markets between 2007 and 2016. We focus specifically on hierarchical patterns in spatial distribution of foundational activities and their association – direct and through integration with other economic activities – with regional employment dynamics in times of crisis, recovery, and growth. Our findings suggest the foundational economy plays an important role as employment provider to a substantial number of Swedish workers, particularly in non-metropolitan regions. Besides, it appears to be associated with improved ability of regions to retain employment in the most acute phases of economic crisis, but only if it is well integrated into regional industrial profiles. However, its overall contribution to regional resilience in the long term appears to be rather limited.


Introduction
Traditionally, research on drivers of economic development and focus of policymakers have tended to emphasise the importance of export-and technology-intensive industries (Moretti, 2012). Thus, the continuous focus on making regional economies more competitive (European Commission, 2018) often translates into support for a narrow set of high-tech industries that invest significantly in research and development (Bentham et al., 2013b, Fothergill et al., 2019, Hansen and Winther, 2014. However, technology-intensive industries generally constitute a limited share of the economy (Hansen and Winther, 2011). To exemplify, in Sweden, high-tech manufacturing activities account for approximately 5 per cent of national employment, ranging between 0.4 and 10 per cent at the municipality level 4 . Thus, suitability and limitations of this (competitive) approach to economic development and policymaking are increasingly debated, in particular for regions far away from large metropolitan areas (Tomaney and Pike, 2020). Increasingly, attention of researchers is given to the role of the everyday (Reeves, 2018) or foundational (Froud et al., 2018) economic activities that deliver services essential for fulfilling basic human needs (Bentham et al., 2013a, Barbera et al., 2016, De Boeck et al., 2019, but remain outside of immediate attention of policymakers. While much of this literature is a normative critique of contemporary economic/industrial policies, there is an increasing stream of papers exploring foundational economy as an economic sector (De Boeck et al., 2019, Froud et al., 2020 and analysing its impact on regional economic performance (Calafati et al., 2020). We contribute to this literature by exploring hierarchical patterns in spatial distribution of foundational activities across Swedish regions and investigating their association -direct and through integration with other economic activities -with regional economic performance in times of crisis and recovery.

Wha f nda i nal in he ec n m ?
Foundational thinking starts from a concept of multiple economies divided into four zones (see Around the FE of daily necessities is an outer zone of the overlooked economy that produces individually provided cultural necessities (e.g., furniture, takeaway food, and tourism) where consumption is taken for granted but is occasional and can be postponed. It also includes lifestyle support goods and services which can be often low-tech goods or mundane support services. Similar to the foundational economy, this zone often remains below the policy radar.
Its share in the municipality employment in Sweden in 2016 ranged widely between 7 and 47 per cent.
Following the definition of Froud et al. (2018), the foundational economy zone encompasses two parts. Material foundational activities connect households to daily essentials through the system of networks and branches. This part of the economy encompasses utilities (electricity, gas, water, etc.), transportation systems, food production and distribution, as well as private banking services and payment systems. Providential foundational services include a subset of (mainly) public sector activities providing universal welfare services to all citizens -education, healthcare, law enforcement, etc. This sector also includes some private supply chain activities, for example, dispensing chemists, which support healthcare, but not pharmaceutical companies.
As the FE delivers necessities for everyday life, these services are consumed by all citizens regardless of income (Moore and Collins, 2020, Hall and Schafran, 2017, Froud et al., 2018 6 . Foundational services are distributed through branches and networks (Barbera et al., 2016, De Boeck et al., 2019, Calafati et al., 2019, and the spatial distribution of foundational activities is expected to follow that of population. Foundational services are generally provided locally as they are either delivered through physical infrastructures or require direct interaction between provider and customer (Froud et al., 2018). While digitization may to some extent affect this by opening up for providing some providential services such as education from distance (Gulbrandsen and Sheehan, 2020), generally local provision of foundational services makes these activities rather immobile and sheltered from inter-regional and international competition (Bentham et al., 2013a, Moore andCollins, 2020).
Thus, while the FE is defined according to the characteristics of demand and opinions of citizens concerning the services that are actually foundational in character (Froud et al., 2018), the production of foundational services and its importance for economic development is arguably central to understand (see Hall and Schafran, 2017). Therefore, the first contribution of our paper is to analyse the distribution of employment in foundational activities across different types of regions and over time.

Wha im an ab he f nda i nal ec n m ?
Whereas an economic development approach focused on technology-intensive industries has little to offer in peripheral, left behind regions, the foundational economy constitutes an important part of the economy in all regions. Moreover, the role of foundational activities is believed to be more pronounced in less-developed regions (Moore and Collins, 2020). The FE thereby proposes a promising approach to embedded regional development, also in peripheral areas (Earle et al., 2018, Heslop et al., 2019a, Morgan, 2019. Crucially, this might be particularly important during times of crisis where he e e e im ac a e being fel in al ead di ad an aged egi n (Tomaney et al., 2010, p. 773).
In relation to the latter point, the role of the foundational economy for economic development may differ significantly from that of traded industries due to non-cyclical demand for foundational services (Calafati et al., 2019). Further, given that foundational activities are labour-intensive (Engelen et al., 2017, De Boeck et al., 2019 and, indeed, employ a large chunk of labour, the FE may exercise a stabilising function on regional economies in periods when economy is contracting (Engelen et al., 2017). Indeed, there are theoretical expectations in the literature on regional resilience that regions with higher share of 'sheltered economic activities and industries with non-cyclical demand can rely, at least in the beginning of a crisis, on a series of automatic stabilisers that soften the blow on employment (Ezcurra, 2011, Webber et al., 2018, enka et al., 2019. More specifically, it was proposed that basic infrastructure, including water provision systems may facilitate regional resilience (Breen andMarkey, 2019, Robinson et al., 2008). Specialisation in sectors closed to international and interregional competition imply that regional employment destruction in times of crisis may be reduced relative to other regions (Fratesi and Rodríguez-Pose, 2016). Our second contribution is, for that reason, to investigate how local presence of foundational activities relates to regional performance in times of crisis, recovery and growth.
While doing so, we move beyond analysing the influence of absolute levels of foundational and traded industries on regional development over time. The third contribution of our paper is to understand how specific compositions of foundational and traded industries in different regions influence regional development outcomes.
This approach not only allows for understanding how regional portfolios of foundational industries influence regional development, but also how integration between foundational and traded industries present in a region matters, rather than simply assuming that their influences can be isolated from each other. The foundational economy is often juxtaposed with traded industries (e.g., Engelen et al., 2017), however, there are multiple ways that foundational and traded regional industries influence each other. Foundational industries are extensively technology-using, highlighting the importance of products offered by traded industries (Coenen and Morgan, 2020). Further, delivery of high-quality foundational services are central for enhancing the competitiveness of traded industries: from ensuring sufficient supply of energy, water and other resources needed in the production of goods and services, to providing the providential services needed by the workforce (Berry, 2018). Also, similar to relations between high-and low-tech industries, collaborations between actors from foundational and traded industries may be of significant importance for some types of innovation processes (Hansen and Winther, 2011). Jointly, this suggests the importance of also considering integration between foundational and traded industries.

Paper outline
Empirically, we analyse the Swedish economy from 2007 to 2016. First, we present an extensive explorative analysis of employment patterns in foundational industries at various spatial scales over this period. Second, we use skill-relatedness approach (Kuusk andMartynovich, 2020, Neffke andHenning, 2013) to operationalise the degree and character of integration between foundational and other zones of the economy. We then investigate the impact of such integration on regional performance in a regression framework. The specific research questions we pose are:

RQ1:
How are foundational activities distributed across various types of regions?

RQ2:
In what ways is the local presence of foundational activities and their integration with other economic zones related to employment growth in Swedish regions? Is this relationship different in times of economic crisis?
The remainder of the paper is organised as follows. Section 2 provides the discussion of the data and definitions employed. Section 3 provides an extensive overview of employment patterns in foundational (and overlooked) zones of the economy over the period between 2007 and 2016. Section 4 presents the summary of regression analysis of relationship between local presence of foundational activities and regional employment growth in different time periods.
Section 5 provides discussion of results through the prism of the literature in the field. Finally, Section 6 concludes.

Data
The data employed in this paper come from the Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA) that is an anonymised linked employer-employee database that aims at complementing traditional labour market statistics and providing a better description of the labour market and people's relationship to the world of work (SCB, 2016 Regions in the paper are defined as local labour markets (LLMs). These are integrated geographical units within which most interactions between workers seeking jobs and employers seeking labour occur. Thus, LLMs are appropriate for linking the supply and demand sides of the labour market and explaining regional labour market performance as a function of endogenous regional factors (including its structural composition). In practice, the boundaries of LLMs are defined by the statistics on commuting between municipalities in the way that maximises the self-containment of commuting flows (SCB, 2010). 290 Swedish municipalities are aggregated into 90 LLMs.
LLMs are further aggregated into four regional families: metropolitan areas, large regional centres, smaller regional centres, and peripheral regions. These are defined by the Swedish Agency for Economic and Regional Growth based on criteria such as the population size and density, regional business dynamics, share of individuals with higher education as well as access to higher education institutions (NUTEK, 2004). We chose this approach instead of a more common rural-urban division because research on Swedish economic development has shown that labour market dynamics varies considerably across hierarchy levels beyond the simple rural-urban dichotomy (Henning et al., 2016, Martynovich and.

Operationalising the foundational economy
We depart from the idea of multiple economies that divides the economy into four zonesfoundational economy, overlooked economy, household economy, and tradeable economy (Bentham et al., 2013a). The description of the former two is provided in the introductory section of the paper. The tradeable economy includes activities that are not classified as foundational or overlooked and comprises predominantly privately-owned export industries that satisfy aspirational private consumption. Household economy is not represented in the official statistics and, therefore, remains outside of our attention.
We employ the classification of economic activities developed by Froud et al. (2018) that allocates four-digit NACE2.0 industries into foundational (material and providential), overlooked and tradeable zones of the economy 7 . Table 1 below summarises the distribution of   industries across zones.   ---------------------------------Table 1 here Almost half of all four-digit industries belong to the tradeable zone of the economy. Overlooked and material foundational activities account for around 23 per cent of industry codes. Finally, providential foundational activities are by far the smallest group with just over 5 per cent of all industry codes.

Employment in the foundational economy in Sweden, 2007-2016
We start by exploring and discussing growth patterns of the foundational economy at the national level over 2007-2016. We pay particular attention to periods of crises respective recovery/growth compared to other parts of the economy. Further, we conduct an extensive regional decomposition of employment patterns in the foundational economy zone comparing three Swedish metropolitan areas with large regional centres, smaller regional centres and peripheral regions.

National employment in the foundational economy
National employment in different zones of the economy is summarised in Table 2. --------------------------------- Table 2 here Looking at the year-to-year variation in employment numbers allows to identify clear lags in the timing of crisis and recovery in different economy zones ( Figure 2).
In the light of the discussion above, it is interesting to observe that the decline in employment is first observed in the FE zone, particularly in its providential services component. It took one extra year for the crisis to hit the overlooked and tradeable zones of the economy. The latter was hit the hardest, as discussed above.
Being the first to fall the providential FE zone is the also first to recover, reaching pre-crisis features of economy stabiliser in the times of crisis, thus making it a potential factor of regional resilience. What is less expected, providential FE activities contributed significantly to the recovery from crisis and subsequent growth. The employment trends in the material FE zone are more in line with expectations: relatively good performance in crisis, but sluggish recovery.
National trends may, however, disguise heterogeneity of regional trajectories with respect to employment in the FE zone. For example, it is hypothesised that its role is more pronounced in the places left behind (Moore and Collins, 2020). In the following section, we perform the regional decomposition of employment trends in the FE zone.

Regional employment in the foundational economy zone
We start by simply mapping the employment shares of the FE zone across 90 Swedish LLMs ( Figure 3). To analyse how this expansion of the FE zone unfolded in space and over time we look at local labour market regions instead of municipalities to account for significant share of commuters (Eliasson et al., 2003). These are further grouped into regional families -metropolitan areas, large regional centres, smaller regional centres, and peripheral regions -to address the question whether foundational economy is distributed hierarchically. These groups of regions (with respect to their population and employment in the FE zone) are presented below (Table 3). Table 3 here Table 3 confirms the claim in the literature that the spatial distribution of FE employment follows that of population distribution as we observe an almost perfect correlation between these two indicators in the beginning and the end of our time period.
Further, the Swedish population increased with more than 800 thousand during the investigated period. Although regions at all levels in the regional system gained population, the increase was strongly biased towards metropolitan regions, in particular Stockholm LLM (see Appendix 4).
Around 75 per cent of the population increase happens in metropolitan areas, while in the remaining regions population growth rate gradually falls at lower levels of the regional hierarchy. This indicates the redistribution of national population from the lower levels to the top of the regional hierarchy. This development is not unique to this specific period but is a continuation, and actually strengthening, of patterns of divergence between metropolitan and more peripheral areas since the 1980s (Lundquist et al., 2017).
Similar hierarchical redistribution patterns can be found regarding the shares of employment in various zones of the economy (Table 4).
--------------------------------- Table 4 here The pattern continues in the recovery and growth period: Stockholm continues (at an accelerating rate) to concentrate employment in all economy zones with the trend reinforced by other metropolitan areas. FE employment is redistributed towards metropolitan areas following the concentration of employment in these regions.
Considering the share of foundational activities in regional employment in different types of regions provides additional details with respect to the hierarchy in their spatial distribution (Table 5). Table 5 here The share of FE in total regional employment increases at lower level of the regional hierarchy.
The differences between the regions are reinforced over time as FE share decreases in two largest metropolitan areas -Stockholm and Gothenburg -while it increases in all other regions.
By 2016, it ranges between 43.5 per cent in Stockholm and 52.5 per cent in the large regional centres.
All in all, the spatial (re)distribution of employment (in absolute numbers) in the FE zone closely follows population patterns in Sweden over the considered time period: regions with higher population have higher employment in foundational activities. This is in line with the necessity of foundational services to be locally provided (Froud et al., 2018). At the same time, the share of foundational activities in local employment tends to increase at lower level of regional hierarchy, particularly when comparing Swedish metropolitan areas to the rest of the economy. This underlines the claims made in the literature that the FE is likely to be of greater economic importance in more peripheral regions (Moore and Collins, 2020).
Focusing on the performance of regions during crisis, recovery and growth reveals a rather interesting sequence (Table 5 and Figure 4). regions at different layers of the regional hierarchy, which fits nicely with redistribution of population discussed above.
When it comes to the foundational zone of the economy, it performed better during the crisis of 2007-2009 than regional economy on the whole in all regions except Stockholm and Malmö.
There are, however, regional differences here: while in larger regions it was providential foundational activities that demonstrated smaller employment decline, in more peripheral areas it was material FE zone that was relatively unaffected. During recovery and growth after 2009, however, material foundational economy kept losing its share in all regions except Stockholm resulting in the overall decline over the considered time period. The providential FE zone, on the other hand, demonstrated faster-than-average recovery and growth in relative terms in all regions and ended up being one of the top economy zones in terms of overall performance over 2007-2016. This contributed to the increase of its share in total regional employment in all regions.
An interesting observation is that employment growth in providential foundational economy accelerated after 2014 in all regions. We have two tentative explanations (which are not mutually exclusive) to that fact. Firstly, in 2014 there was a change in the governing coalition from centre-right to centre-left as a result of parliament election. This might have led to the increased focus on welfare provision which is the major function of the providential foundational economy zone. Secondly, 2015 marked a dramatic increase in the number of immigrants to Sweden which required the proportional increase in demand for welfare services which, in its turn, resulted in the accelerated employment growth in the providential foundational economy zone.
Despite some similarities, there were also dramatic differences between regions at different levels of regional hierarchy. These are briefly summarised below.
In Stockholm, the overall employment falls slightly in the crisis (-0.1 per cent between 2007 and 2009; corresponding to losing 867 jobs) but it is far from the magnitude experienced in other regions. The FE zone is hit by the crisis earlier and stronger compared to the tradeable and overlooked economy zones (see panel (a) in Figure 4). In that respect, neither material nor providential foundational activities act as a stabilising factor for Stockholm LLM; it is rather the other way around. In the recovery and growth period the leading role is played by the overlooked economy zone that somewhat surprisingly shows outstanding growth rates followed by the providential FE zone. Tradeable activities mirrored overall recovery and growth trend of Stockholm economy but never exceeded it (that likely has to do with the strong base effect as In Gothenburg, the overall performance during the crisis was significantly worse but recovery and subsequent relative employment growth were at par with Stockholm LLM. The internal dynamics between the economy zones, however, played out quite differently. The providential FE zone demonstrated better performance during the crisis than any other sector in the economy and also led the recovery and growth after 2009. Tradeable economy zone took some time to recover from the deep decline in employment in 2009, but its fast growth after 2014 ensured that it performed better than the economy as a whole over the 2007-2016 time period. Material foundational activities were strongly affected during the crisis and did not recover until 2016. Malmö was somewhat less hit by the crisis compared to Gothenburg but had a much weaker recovery. The internal dynamics between the economy zones resembled that of Gothenburg with a couple of minor exceptions. The providential FE zone performed quite well both in crisis and recovery, but particularly after 2012 when it leapt forward from other economy zones in terms of employment growth. Between 2012 and 2016, providential economy zone increased its share of employment by more than two percentage points (from 32.5 to 34.7 per cent).

Material foundational activities represented the largest share of Malmö employment in 2007
compared to all other regions but suffered from strong decline during the crisis and never recovered to the pre-crisis level.
When comparing growth trajectories of large regional centres, smaller regional centres and peripheral regions, a shared feature is the dramatic fall in tradeable economy zone during the crisis (7.3-9.2 per cent) that far exceeded that in metropolitan regions (growth in Stockholm, and between 2.8 and 4.2 per cent decline in Gothenburg and Malmö). Demonstrating a relatively fast recovery during a couple of years after 2009, employment hit the ceiling and stagnated/declined thereafter. This had a particularly negative impact in smaller regional centres and peripheral regions as they never returned to pre-crisis total employment levels. In large and smaller regional centres, employment decline in tradeable zone is partly off-set by a steady employment increase in providential foundational activities which are the only economy zone where employment returned to pre-crisis level. Same is true for peripheral regions; however, the recovery in providential foundational employment is smaller in magnitude and much delayed.
All in all, based on the discussion above we can claim that the performance of foundational economy zone during the crisis (for both material and providential FE) and during the recovery and subsequent growth (for providential FE) indicates that it might be a factor of regional resilience helping regions to mitigate employment decline when the crisis hits, and also boosting regional recovery after the crisis. To test this idea further, we perform the regression analysis that relates regional characteristics of foundational economy to the overall employment growth in regions during crisis and recovery.

Measuring the regional presence of the foundational economy
To investigate the impact of FE on regional development we define two measures. The first one is simply the share of foundational activities (separately for material FE and providential FE) in regional employment. This allows us to analyse whether the scale of foundational activities in a region is related to the regional performance.
The second measure makes use of recent advances in research on skill relatedness and associated measures of regional related variety (Fitjar and Timmermans, 2017, Kuusk andMartynovich, 2020, Neffke andHenning, 2013). The starting point is the notion that excessive exchange of labour between two industries signals overlapping skill requirements between them and indicates that these industries are related (Neffke and Henning, 2013). In our case, let be an observed flow of labour between industries i and j (i j) and -an expected flow of labour between them derived from industry sizes, growth, and average wages during the same time period 8,9 . Then the values of ratio of observed to predicted flows , that are significantly larger than 1 indicate that industries i and j are skill-related. We obtain a matrix of skill relatedness indices for each of the 615 614 = 377610 combinations of 615 industries at the four digit level of the NACE2.0/SNI2007 classification.
To aggregate constructed linkage metrics to the regional level it is possible to use the weighting procedure proposed by Fitjar and Timmermans (2017): where 1, if SR 1 and both industries are present in region 0, if one or both of the conditions are not satisfied , -a share of industry i in regional employment; -a number of industries present in region r. In broad terms, this indicator represents the (weighted) average number of related industries per each industry present in the region.
There are three possible situations, both industries i and j belong to the foundational zone of the economy, one of the industries is foundational and another is not foundation, or both industries fall outside of the foundational category. We can decompose ∑ into three categories then: where is relatedness measure for two industries belonging to the foundational economy one, measures relatedness for two industries outside the foundational economy zone, and 8 and calculated at the national level. finally is relatedness for industries across economy zones. We can then decompose the regional measure in the following way: These indicators complement the share of foundational economy in regional employment: if a certain industry i employs a large number of workers (high ) but is not related to any other activities present in the region (∀ : 0) then its contribution to is zero. In that respect, these measures indicate how well (on average) various industries are integrated with each other in the regional industry space. In relation to foundational activities, measures average relatedness of various industries within the regional foundational economy zone while indicates the degree of relatedness across the zones of economy in a region. In the absence of data on input-output linkages this approach proves to be a good way of understanding how well different industries present in a region are connected to each other and the overall degree of integration within regional industry space.

Estimating relationship between foundational economy and employment growth
To investigate how the local characteristics of the foundational economy are associated with regional employment growth we estimate the OLS model of the following form: where represents employment growth in region r, and are the measure of material and providential foundational activities in regional employment in region r, indicates related variety in a region r (total and separately for each of its components), and finally is a matrix containing control variables values. is a standard error term. The model is estimated for the whole time period (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) as well as separately for the period of crisis (2007)(2008)(2009)) and recovery and growth (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). All independent variables are measured at the beginning of each respective period.
Choosing employment growth instead of regional GDP or productivity growth as a dependent variable is motivated by our interest not only in purely economic but broader social impact of the 2007-2009 crisis. Not disregarding the importance of value added or increased productivity, employment is better suited to reflect upon the welfare aspect of foundational economy. Given the potentially discouraging impact of 2007-2009 crisis on the labour market participation rate, employment growth is also preferred to changes in regional unemployment rates (Fratesi and Rodríguez-Pose, 2016).
For the whole time period, we also estimate the random effects panel model of the following form: where -employment growth in region r between t and t+2 10 , and represents unobserved time-specific shocks that are uniform across all regions, such as national or global shocks.
A 3-year period panel model is preferred over a model capturing year-to-year variation in the data for two reasons: First, regional structural preconditions change rather slowly, implying a relatively low year-by-year variation within regions (Firgo and Mayerhofer, 2017). Second, year-to-year models only identify short-run associations between structural factors and regional growth, leaving out long-run effects. Yet, as changes in structural conditions often take time to translate into growth, it makes more sense to employ an interval model rather than a year-toyear model.
There is an agreement in the literature that similar economic structures do not necessarily produce similar economic effects in regions that fundamentally differ from each other ( enka et al., 2019). As regional trajectories are more complex than industrial trajectories (Webber et al., 2018), it is important to differentiate the relationship between industry structures and regional outcomes in different kinds of regions. In our case, to account for the fact that the spatial distribution of foundational activities is related to the spatial patterns of population and 10 The model uses the 'rolling estimation periods where each subsequent period is moved one year forward. For instance, the period (t, t+k) is followed by the period (t+1, t+k+1). This is done to minimise the impact of idiosyncratic fluctuations in the independent variables on the regressions results.
to investigate whether this has an impact on relationship between the local presence of the foundational economy and regional employment growth, we specify the following dummy variable: 1, if population 0 population 00 0, if population 0 population 00 .
By interacting this dummy variable with the measures of regional characteristics of the foundational economy 11 , we can estimate whether the association between the latter and regional employment growth differs between depopulation regions and regions with increasing population.
As control variables, we include some general structural characteristics of local labour markets.
We account for the share of employment in manufacturing to control for the sensitivity of regional labour markets to macroeconomic conditions as 'manufacturing and construction industries have been viewed as being more cyclically sensitive than private service industries (Martin, 2012, p. 13). To capture the regional innovativeness and competitiveness, we define the share of regional employment in high-tech manufacturing 12 and the share of employment in knowledge-intensive services 13 . Human capital effects on regional employment dynamics is captured by the share of regional population with higher education (within the group of workers aged 25+). Finally, urbanisation externalities are captured by population density of regions. 11 The model specified in equation (1) takes the following form: The model specified in equation (2) Table 6 presents the estimation results for the OLS and random effects model for the whole time period. Table 6 here

Regression results
The results of estimating OLS models indicate that the share of material foundational activities does not have any significant relationship with the regional employment growth while the share of providential foundational activities is negatively associated with it. Higher regional related variety tends to correlate with faster increase in employment, driven primarily by related variety within the non-foundational economy zone and across economy zones. Model diagnostics indicate that the former association is somewhat stronger. At the same time, related variety within the foundational economy (FE) zone is not correlated significantly with regional employment growth.
Random effects panel models, in general, confirm the findings summarised above. Two key differences are that (1) related variety within the FE zone is now weakly, but significantly associated with regional employment growth; and (2) model diagnostics prioritise the integration between foundational and non-foundational economy zones over related variety between non-foundational industries. The best model, however, is the one that accounts for the total related variety in the region.
All in all, it appears that having a large foundational sector in a region is not something to be desired as it tends to correlate with slower employment growth (particularly with respect to the providential foundational activities). This negative relationship is mitigated somewhat if foundational activities are well integrated with the regional non-foundational economy zone.
Looking at the models that account for heterogeneity of regional population growth patterns additionally qualifies the results above (Table 7). Table 7 here Most importantly, the negative association of providential activities share with regional employment growth is greatly mitigated in depopulating regions. That is, higher share of providential activities is more damaging in regions with increasing population. This result is observed both in OLS and random effects panel models. This is consistent with the fact that the providential foundational employment is growing across all groups of regions but contributes particularly strong to job creation only in smaller regional centres and peripheral regions which are suffering from depopulation.
When it comes to related variety measure and its sub-components, OLS models do not provide any additional evidence. Panel models, however, suggest that the impact of total related variety as well as integration within and between various zones of the economy is stronger in regions with growing population. This is in line with previous findings in the literature that related variety tends to have stronger relationship with employment growth in more successful regions (Firgo andMayerhofer, 2017, Kuusk andMartynovich, 2020).
Looking at the relationship between characteristics of the local foundational economy and employment growth in times of crisis vs. recovery and growth reveals the differential role of foundational activities (Table 8). Table 8 here In times of crisis, regions with large share of providential foundational activities tend to be disadvantaged, similarly to the observations for the whole time period. Material foundational economy is not associated significantly with employment growth dynamics. What is interesting, however, higher related variety within the foundational economy zone has a positive association with regional employment growth during crisis. There is an additional positive association between employment dynamics and integration across the zones of the economy.
Thus, in times of crisis having a large foundational sector tends to accelerate employment decline in a region, which however can be offset if regional foundational economy zone is coherent within itself and integrated with the rest of the economy.
The dynamics is reversed in the times of recovery and growth. It is now the related variety within the tradeable zone of the economy that seems to support regional employment growth.
That is much in line with most of literature on related variety. Also, the negative association between the share of the foundational zone and employment growth disappears in times of growth. This may be related to the good performance of providential foundational activities as was discussed in Section 3.
Considering regional population growth patterns provides some additional details to the picture (Table 9).
--------------------------------- Table 9 here The population growth patterns, however, seem to have an impact in the recovery and growth models. Here, the related variety within the non-foundational economy continues to have a positive association with employment growth, which is, though, stronger in regions with growing population. Once again, this is consistent with the literature.
What is more interesting, however, the models without regional heterogeneity masked the highly differential relationship between the share of providential foundational activities and employment growth in regions with growing population vs. depopulating regions. In the former, the association is negative. That is, in times of overall recovery and growth in the national economy, having a stronger local presence of providential employment tends to slow down employment growth despite the fact that such activities are expanding rather fast. This points to the fact that in successful regions faster job creation demands strong dynamics in nonfoundational economy zones.
At the same time, the negative association between providential foundational employment share and regional employment growth turns positive in depopulating regions. This is consistent with the findings from the descriptive analysis that in regions at lower levels of regional hierarchy, which are suffering from depopulation, providential foundational economy zone was the only one to return to pre-crisis employment levels during the recovery period meaning that it was the only sector to create additional jobs.
Overall, the findings from regression analyses indicate that the foundational economy does not really generate new jobs: to stimulate employment growth, regions need a strong and large nonfoundational sector. However, in crisis times the foundational sector may act as a stabilisation factor, particularly if it is diverse, but coherent. Importantly, but not surprisingly, the relationship between foundational economy zone and regional growth creation seems to be more positive in depopulating regions, particularly in the recovery and growth period.

5.
Foundational economy: a factor of regional resilience?
In this paper, we set off to explore the employment in foundational zone of the economy in Sweden between 2007 and 2016. Following suggestions in the literature, we structured our discussion along two interrelated dimensions: geographical, in that we investigated hierarchies in spatial distribution of employment in foundational activities across various types of regions, and temporal, as we followed foundational economy employment during and in the aftermath of 2007-2009 crisis. Additionally, we analysed how regional characteristics of the foundational economy -its share, variety, and integration with other economy zones -was related to the overall performance of regional labour markets in crisis and recovery periods. Our findings allow us to discuss the role of foundational economy as stabiliser in turbulent times.
First, despite being largely disregarded in (regional) economic policies in favour of export-and technology-intensive sectors, foundational economy plays an important role as it employs almost a half of workers nationally and more than that in the majority of regional labour markets. In fact, its share in national employment increased over time as 170  of the Swedish regional hierarchy. Their share in regional employment tends to increase from metropolitan areas to peripheral regions. In that respect, the foundational economy s role as provider of jobs is more pronounced in less developed regions (Heslop et al., 2019b, Moore andCollins, 2020).
Third, in all regions but Stockholm and Malmö, the foundational economy zone suffered from less employment decline than the rest of the economy. This corresponds to the non-cyclical demand for foundational goods and services as the foundational economy supplies everyday necessities, consumption of which cannot be postponed. Besides, this is in line with the sheltered nature of foundational activities -that is, limited exposure to international and interregional competition (Bentham et al., 2013a, Moore andCollins, 2020). What is surprising, however, in most regions, employment in the foundational economy zone increased above regional average also after the trough of the crisis in 2009. This is primarily attributed to the providential component of the foundational economy zone, that is, provision of crucial welfare services to the population. The latter grew much faster in metropolitan areas, possibly reflecting the redistribution of population from peripheral to more dynamic regions.
Such employment dynamics in the foundational economy zone, coupled with predictions in the literature about sheltered economic activities and industries with non-cyclical demand as regional stabilisers (Ezcurra, 2011, Webber et al., 2018, enka et al., 2019, may tempt us to conclude that regions with higher share of such activities in their employment portfolios will be less affected by the crisis; that is, foundational economy may be a positive factor of regional First, the relationship between foundational activities share and regional employment growth differs between regions with growing population and depopulating regions, particularly in the time of recovery and post-crisis growth. In depopulating regions, which are likely to be found at lower levels of regional hierarchy, the association turns positive after 2009. In such regions, foundational activities (specifically, providential foundational activities) were the only ones to return to pre-crisis level of employment. Thus, they were the only drivers of job creation.
Whether this is the outcome of regional/national welfare policies or the result of relocation of jobs and/or production facilities to other regions remains outside the scope of this paper and deserves additional research. Whatever the reason, post-crisis recovery and growth in depopulating regions tend to benefit regions with higher share of employment in the foundational economy zone. More generally, we agree with enka et al. (2019) that similar industrial structures (in our case, in terms of share of foundational employment) may produce different regional outcomes with respect to resilience to crises.
Second, while the size of the foundational economy zone tends to have a negative association with employment growth, the related variety between foundational industries as well as integration between foundational and non-foundational economy zones appear to contribute to slower employment decline in the trough of the crisis. We can relate this to the results of Christopherson et al. (2010) who showed that in the USA regions least affected by the economic crisis included those with high diversity in, among others, educational and health institutions (that are parts of the providential foundational economy). A positive role of integration between foundational and non-foundational economic zones in the crisis times (as well as over the whole period between 2007-2016) underlines the importance of considering the regional economies as a whole rather than treating tradeable/competitive and foundational sectors as independent entities 14 . Where 'competitive emphasises the need for change, 'foundational focuses on continuity and stability, and both are needed to make regions more resilient to economic shocks (Boschma, 2015).
The latter points allow us to contribute to the burgeoning literature relating regional resilience skill-related, that is require similar skills, as this enhances regional labour matching (Neffke and Henning, 2013). Besides, related variety enhances the recombination potential of a region 14 Indeed, proponents of 'competitive paradigm of regional development are readily reading off policy lessons from success stories of dynamic regions: if only less dynamic cities and regions develop clusters, build regional innovation systems, pursue 'smart specialization , etc. then their economies will be revitalised (Martin 2015). Supporters of 'foundational view propose that we need to displace the idea of a 'competitive region and focus instead on grounded cities and regions driven by their foundational economy (Engelen et al. 2017). and provides local (related) resources on which new growth paths can build and develop, thus improving long-term regional resilience (Boschma, 2015).
In our case, overall related variety in a region appeared to correlate with faster employment growth over the whole time period between 2007 and 2016, which confirms the results in the literature. There are, however, critical differences between the acute phase of the crisis and the post-crisis recovery. In the former, it is related variety within the foundational economy zone as well as between foundational and non-foundational economy zones that tends to boost the ability of region to retain jobs. This relationship does not seem to differ between regions with growing population and depopulating regions. During the post-crisis recovery, however, variety in non-foundational/tradeable sectors tends to stimulate regional job creation more, particularly in regions with growing population. Thus, it is neither variety as a whole nor total related variety in region that matters for its resilience, but rather related variety within certain groups of industries and at certain times.
So where does this leave us with regards to foundational economy as a factor of regional resilience? On the one hand, being sheltered from international and interregional competition and facing non-cyclical demand, foundational economy zone manages to retain more jobs (in relative terms) than other economy zones. Besides, related variety within foundational economy tends to associate with less regional employment decline during the trough of the crisis. Thus, if we define resilience as the capacity of a region to respond better to short-term shocks, then we have evidence pointing to the foundational economy as contributing positively to regional resilience.
Contemporary definitions of resilience, however, tend to underline the capacity of regions to sustain long-term development, that is their ability to adapt and reconfigure their industrial, technological and institutional structures in an economic system that is restless and evolving (Boschma, 2015). While our analysis does not allow us to say much about the relationship between the foundational economy and the ability of regions to transform, the long-term focus of this definition renders the capacity of foundational economy to contribute to regional resilience as very limited. Indeed, once the economy passes the trough of the crisis, the foundational economy keeps its positive contribution only in the depopulating regions, while related variety in non-foundational sectors tends to promote job creation overall.

Concluding remarks
Overall, we find that the foundational economy plays an important role in providing employment to a substantial number of Swedish workers. Besides, it seems to be associated with better ability of regions to retain employment in the most acute phases of economic crises.
Yet, its overall contribution to regional resilience in the long-terms appears to be rather limited.
When discussing the implications of our results in relation to the foundational economy literature and its suggestion that foundational sectors should take the centre stage in economic development policy, we want to underline two important qualifications to our study. First, while our analysis highlights that the foundational economy is positively associated with regional employment dynamics in some periods (especially during economic crises) and some regions (especially depopulating regions), our results do not suggest that the foundational economy is in general driving economic growth. However, it is important to remember that a central motivation for the work of proponents of the foundational economy is the skewed attention towards the tradeable economy zone in policymaking (Froud et al., 2018). Thus, one might hypothesise that the relative importance of the foundational economy zone for employment growth would be greater, relative to the tradeable economy zone, if policy support for these parts of the economy was balanced. While one could expect that the emphasis in policymaking on the tradeable economy zone would be less predominant in Scandinavian welfare states, previous research suggests that high-tech, traded industries are also prioritised in economic development policies in these countries (Hansen and Winther, 2014).
Second, the argument of foundationalists for placing foundational sectors in the centre of economic development policy is not only related to job creation effects. Additional arguments relate to improving working conditions for employees in the foundational economy zone and enhancing quality and accessibility of foundational services that matter to all (Froud et al., 2018). Thus, it would be beneficial for future research to complement the focus on employment growth with other dependent variables measuring working conditions as well as quality and affordability in foundational service provision. Still, we argue that job creation is indeed a fundamental element allowing citizens to live decent lives, and the current paper thereby provides a first detailed, longitudinal analysis of the role of the foundational economy across different types of regions.
Focusing on the relationship between the foundational economy and regional employment dynamics, we had only a limited opportunity to discuss the former as the driver of regional structural change and economic growth. This, however, is an interesting avenue for further research that would allow for a broader understanding of economic impact of foundational activities. Such research direction would require a different approach, focusing on induced effects on growth rates in value added and productivity not only inside, but also (and, perhaps, more importantly) outside the foundational economy zone. There are many questions to be addressed. When and under which circumstances does regional presence of foundational activities is an enabler for building competitive advantage of cities and regions? Can it be a constraint? Can regions that are 'left behind benefit not only from the stabilising function of the foundational economy zone, but also use it to escape their inferior growth paths.
Alternatively, is the foundational economy zone s role in this context rather to help regions to "shrink with dignity"? Answering these questions would help to better understand the hypothesised role of the foundational economy as a driver of regional resilience, beyond our approach. This would require combining systematic quantitative studies on the role of foundational economy across different types of regions with in-depth qualitative inquiries that would allow exploring various configurations of the mechanisms of interaction between foundational and tradeable economy zones.
In terms of policy, we claim that rather than saying that the foundational economy should be the starting point for economic development policy (as its proponents suggest (Froud et al., 2018)), we would suggest a more context-sensitive approach that would prioritise foundational sectors for job creation and stabilisation of regional labour markets in some places and in certain time periods. More generally, our analysis suggests that we need to move away from the foundational/traded dichotomy. Echoing the work within evolutionary economic geography, our analysis points to the importance for policy of considering interconnections between these parts of the economy, as well as coherency of regional foundational and tradeable industrial profiles. Prioritisations of policymakers should thereby also depend on the sets of foundational and tradeable industries that are already present in specific regions.
Funding information: Mikhail Martynovich s contribution to this paper was supported by a grant from Jan Wallanders and Tom Hedelius foundation (Grant Number: W17-0016). The database used in the paper was supported with funding of Länsförsäkringar Alliance Research Foundation through the project Regional Growth against All Odds (ReGrow).           AIC -708.1282 -700.1278 -708.5081 -705.4677 n/a n/a n/a n/a BIC -685.6299 -677.6295 -686.0098 -682.9694 n/a n/a n/a n/a Within R 2 n/a n/a n/a n/a 0.5393 0.5403 0.5391 0.5389 Between R 2 n/a n/a n/a n/a 0.    -716.6187 -712.8021 -716.4274 -716.1557 n/a n/a n/a n/a BIC -686.6210 -682.8044 -686.4297 -686.1580 n/a n/a n/a n/a Within R 2 n/a n/a n/a n/a 0.5414 0.5392 0.5416 0.5414

AIC
Between R 2 n/a n/a n/a n/a 0.     Table 9. Regional heterogeneity models in crisis and recovery Dependent variable -Employment growth