Enhancing inclusive growth to create new evidence of rural diversity: an analysis in the Highlands and Islands of Scotland

ABSTRACT

This article applies a modified perspective of inclusive growth in order to understand the diversity of communities within a remote and predominantly rural region of northern Europe. A transdisciplinary process co-constructed an enhanced and regionally appropriate framework of inclusive growth, which recognised the importance of evaluating the concept as multidimensional and affected by the uneven contexts of physical geography and connectivity, and social characteristics, which influence development potential and community resilience, and which extend the relevance of inclusive growth to rural places. Operationalisation of this framework was achieved at the local level through quantitative multivariate analysis, which identified underlying dimensions of inclusive growth ‘performance’ representing interactions of broad economic outcomes and variation in geographical and social contexts, and then clusters of locations with similar characteristics. The evidence of multidimensional and geographical inequalities from this analysis was evaluated using a spatial justice perspective, enhancing the applicability of an inclusive growth framing as a means of understanding the wellbeing economy and rural diversity. Conceptually, the regionally-sensitive understanding of inclusive growth performance builds on and develops existing broader framings of inclusive growth. To avoid leaving places behind, policies which seek to achieve inclusive growth should recognise and adapt to nuanced and localised inequalities.

KEYWORDS:

• Inclusive growth
• spatial justice
• wellbeing economy
• regional diversity
• inequality
• geography

1. INTRODUCTION

Institutions in more developed countries have a growing interest in addressing economic inequality. For example, the comparatively very high levels of regional inequality within the UK are clear (Martin et al., 2021; UK2070 Commission, 2020) with a ‘consensus’ that this should be addressed (UK2070 Commission, 2020). It is argued that EU cohesion policies have not successfully addressed inequality or supported spatial justice (Madanipour et al., 2022). Within this context, inclusive growth (IG) has emerged as a ‘new (policy) mantra’ (Lee, 2019, p. 424). This review identifies the value of a regionally-sensitive operationalisation of IG which resonates with policy interest in wellbeing economies and emerging learning about ‘left-behind’ places.

IG is within a suite of alternative economic approaches which have been advanced due to unequal development (Crisp et al., 2023). Previously, IG was applied in developing countries, where key characteristics of the critical place of economic growth, and a ‘non-redistributive’ ethos, emerged (Ali & Zhuang, 2007). Fundamentally, IG assumes that economic growth and increased equality can be delivered tangentially (Evenhuis et al., 2021, p. 13) and IG policies intend for quality economic development (rather than redistribution) to drive lower inequalities (Statham & Gunson, 2019, pp. 10–11). However, the nature of ‘inclusion’ in IG definitions is not consistent (Lee, 2019).

Additionally, central to many definitions of IG is economic growth, which has been criticised as a metric of success. Costanza et al. (2009) reflect that gross domestic product (GDP) is inadequate as a measure of wellbeing, as it does not capture the multiple capitals supporting economic activity, or issues of sustainability and inequality, and (they advise) there is an urgent need to develop more appropriate metrics of development. The breadth and ‘dimensionality’ of prosperity is central to this discussion. Jackson (2009) criticises definitions of prosperity which are tied solely to wealth or life satisfaction, arguing that the ability of people to achieve multiple needs which support their health and wellbeing should be how prosperity is understood. This perspective aligns with the view that sustainable development results from contributions of multiple ‘community capitals’ (Flora et al., 2018). This re-prioritisation of development outcomes towards wellbeing, recognising diverse influences on wellbeing and requiring a more nuanced and holistic measurement of value, forms the ‘wellbeing economy’ (Fioramonti et al., 2022) which is considered a larger departure from mainstream economic development (Crisp et al., 2023).

These debates question the value of IG as a concept and development goal, and arguably, two perspectives on IG have emerged (Evenhuis et al., 2021; Table 1). Aligning closely with the definitions of IG noted above, the first emphasises the centrality of maintaining economic growth and the need for more people to participate in this, while the second is a broader and more radical view which prioritises plural outcomes, de-emphasising and criticising the importance of economic growth (Hughes & Lupton, 2016; Waite et al., 2020). The latter constitutes an extension of the conceptual scope of IG. Similarly, Hill O’Connor et al. (2024) have situated IG policies on a continuum from incremental adaptations to transformative changes. The de-prioritisation of growth increases the relevance of IG to wellbeing (Evenhuis et al., 2021), and broader framings of IG are also necessary in a context where the economic growth model has had negative impacts on several factors affecting wellbeing, leading to alienation (Hay et al., 2020), and which recognises that ‘inequality is multidimensional’ (Patias et al., 2022, p. 160).

Table 1. Established definitions of IG. The ‘narrow’ and ‘broad’ descriptions were also used by Hughes and Lupton (2016) and Waite et al. (2020).

IG definition type	Definition	Adapted from
narrow	Reduced inequality through economic growth	Statham & Gunson, 2019, p. 10
		Hughes & Lupton, 2016, pp. 6–7
		Waite et al., 2020, p. 1824
broad	Reduced inequality through reshaping economic development to contribute to holistic goals	Hughes & Lupton, 2016, p. 7
		Waite et al., 2020, p. 1824

These questions of IG may be becoming visible in policymaking. Waite and Roy (2022) and Hill O’Connor et al. (2024) suggest that in Scotland IG is increasingly being de-emphasised in favour of wellbeing: indeed, the Government have stated that the ‘Vision’ of the Wellbeing Economy ‘builds on our previous inclusive growth approach’ (Scottish Government, 2022, p. 12). However, others suggest that actions supporting IG and wellbeing are mutually supportive (Statham & Gunson, 2022). Practically, recommendations to extend IG to a multidimensional concept raise the issue of developing and extending evidence and data (Hay et al., 2020).

While ‘narrow’ understandings of IG are clearly limited, it is uncertain how the concept should be enhanced and applied to better understand multidimensional development associated with wellbeing. This arises partly as IG’s ‘fuzziness makes it hard to operationalize’ (Lee, 2019, p. 429), with evidence of ‘consistent ambiguity’ in understandings of IG among practitioners and difficulties in implementing the flexible concept and forming measures to represent it (Hill O’Connor et al., 2024, citation: p. 123). However, two further areas of thought are relevant to this uncertainty.

Firstly, the economic inequality which is a focus of ‘narrow’ framings of IG, and a reason why the concept has received cautious support (Lee, 2019), remains salient given the emergence of the geography of ‘left-behind places’ suffering economic decline and disaffection (Rodríguez-Pose, 2018), which identify part of the complex relationship between uneven economic outcomes and wellbeing. Their existence emphasises the critical importance of inclusion, but the need to understand this and geographical inequalities as complex is also evident, as left-behind places are heterogeneous (Martin et al., 2021), are created by diverse processes and challenges (Velthuis et al., 2022), and their meaning has evolved to reflect inequalities which ‘are not solely economic in cause, expression or solution’ (Pike et al., 2023, citation: p. 10). Left-behind places therefore demand development approaches which go ‘beyond narrow conceptions of the economic’ (MacKinnon et al., 2022, p. 51). Furthermore, there is recognition of the need to understand these places as situated within the ‘spatial continuum’ as within- and between-region inequalities can affect place-based disaffection (Larsson et al., 2021), and links have been found between local economic conditions (and perceptions of these) and satisfaction with how community interests are taken on board by national decision-makers (McKay, 2019).

Secondly, the observation that IG has often been considered as an ‘urban’ development approach (Crisp et al., 2023; Lee, 2019) raises an important question about how it should be understood and applied in rural areas. This is supported by spatial variation in interpretations and implementations of IG between (urban) regions in the same country, associated with distinctive economic processes (Waite & Roy, 2022). Separately, Jones et al. (2020) and Mahon et al. (2023) argue that spatial justice thinking can support the formulation of more diverse, regionally-grounded and locally-formed definitions and visions of development priorities, with benefits for rural and disadvantaged places that are considered economically ‘left-behind’. In addition to the potential to supply new evidence to identify and inform plural development goals, the demand for understanding rural IG is increased by the argument that focusing on cities for economic development sits uncomfortably with IG aspirations (Clelland, 2020) and the potential for ‘city region’ policies to generate uneven impacts in rural areas has been recognised due to variation in their economic characteristics and infrastructure, contributing to a ‘need for alternative economic development approaches, sensitive to the geographies of rural localities’ (Beel et al., 2020, citation: p. 729). Identifying the potential of IG to make sense of rural development patterns could support the formulation of customised ‘place-sensitive’ approaches (Iammarino et al., 2019) which are aligned closely with IG.

Given the question of how IG can best maintain its valuable focus on inequality, and evolve to more effectively resonate with wellbeing and rural characteristics, this article applies a modified perspective of IG to understand the diversity of communities in the Highlands and Islands of Scotland. This article argues that for IG to be useful as a practical means of understanding variability in the wellbeing economy, rural diversity, and issues in left-behind places, it must be transformed into a flexible and regionally-sensitive framework where prosperity (rather than economic growth) and inclusion outcomes overlap with underlying spatial and social contexts, and key concepts for evaluation emerge from these relationships, rather than being specified a priori. We aim to advance conceptual development by arguing that understanding IG and strengthening its links with wellbeing requires going beyond applying a multidimensional and broad framing.

2. MATERIALS AND METHODS

This paper is based on outputs from a project which focused on understanding and measuring IG within Scotland’s Highlands and Islands (Figure 1) which included collaboration between researchers and practitioners at Highlands and Islands Enterprise (HIE), a regional development agency.¹ Scotland represents a particularly relevant country for IG research, because of the devolved Government’s prioritisation of this in economic strategies: descriptions of this have appeared within recent IG research (e.g., Waite et al., 2020; Waite & Roy, 2022) and will not be described in depth here. Partnership-based regional deals also aim to deliver inclusive economic growth,² and IG forms part of the strategic ‘vision’ of regional development agencies (HIE, n.d.; South of Scotland Enterprise, 2022). However, Scotland appears representative of the ‘fuzziness’ of IG (Lee, 2019): Statham and Gunson (2019) recommended that the Scottish Government’s IG definition should be clearer, and noted difficulties in assessing progress towards IG at different scales; furthermore, a key development index (the Scottish Index of Multiple Deprivation) does not identify rural disadvantage well (Clelland, 2021).

Figure 1. Map: Scotland and the Highlands and Islands, showing places referred to in the text.

Source: Spatial data shown: Appendix 1.

A map focused on Scotland, showing the Highlands and Islands region in the north and west of the country and urban areas within Scotland. Major cities and places referred to in the article text are indicated and labelled.

The project was informed by a need identified by – for a more nuanced and granular understanding of rural development. This was carried out through a transdisciplinary approach, aligning with the reflection of Waite and Roy (2022) that implementing IG constitutes a wicked problem. The term IG was used throughout the research and the term ‘IG performance’ used within this paper reflects multidimensional and contextual considerations of inclusion and prosperity. Collaboration and analysis took place in 2020, and this paper describes outputs from a revised version of the quantitative analysis (reproducing main steps, but with some differences in details) formulated after the project. Key stages and milestones in collaboration and analysis are shown in Figure 2.

Figure 2. Simplified overview of the research which contributed to this paper. Labels referred to in the text. Part of the framework represents the Inclusive Growth Monitor.

Source: Beatty et al., 2016.

An infographic diagram showing key stages of the research described in this paper, with labels reflecting the seven stages referred to within the ‘Material and methods’ section, from the co-construction of a conceptual framework (stage 1) to the interpretation of the inclusive growth classification (stage 7).

Firstly, a framework of IG which was sensitive to the characteristics of the Highlands and Islands region was co-constructed (Figure 2: step 1), forming a basis for indicator selection and analysis. The regional scale is critical, as the predominantly remote rural region has a very different geographical character from the cities where IG is more typically considered (Lee, 2019). The collaborators provided details of limitations in existing knowledge: the two-fold ‘Fragile Areas’ classification (HIE, 2014) used for the spatial targeting of interventions; the conceptual basis for the study: the definition of IG, provided by the Poverty and Inequality Commission, that should be used (Statham & Gunson, 2019, p. 4); and a starting framework: the IG Monitor published by the Joseph Rowntree Foundation, which includes themes of prosperity and inclusion and underpinning dimensions reflecting parts of the relationship between growth and poverty (Beatty et al., 2016). The collaborators stated the key project goal as producing typologies to identify places with similar characteristics of prosperity and inclusion, in the context of physical geography and population, and so provided directions for contextual development of the framework. An extension of this framework was co-developed to enhance alignment with regional characteristics (Table 2). Adding contextual themes acknowledged that economic development processes interact with and utilise place-based resources, assets, and barriers to development, and affect communities with different levels of resilience and vulnerability; these spatially uneven influences on communities’ potential and capacity for development are essential for evaluating the inclusiveness of development. These themes were influenced by hazards research which emphasises the dual influences of geographical and social characteristics on place-based vulnerability (Cutter, 1996). Factors supporting rural community resilience, including financial and social capital, community spirit, and access to diverse services and infrastructure (Currie et al., 2023) also align with dimensions across the extended framework. Dimensions of social strength, infrastructure and vulnerability, the latter including indicators reflecting legally protected characteristics (Equality Act 2010 (c.15): Section 4³), are included in addition to human capital. Contextual characteristics reflect characteristics pertinent to rural areas, or which are prone to rural variation, and can form both resources and vulnerabilities for development. Sparsely populated regions face economic challenges (Gløersen et al., 2005, pp. 23–27), but rural uplands contain considerable environmental assets (Bradley et al., 2005) and environmental quality and outdoor amenities may attract creative workers to some rural areas (McGranahan & Wojan, 2007). Within the prosperity theme, additional indicators were included representing diversity in economic activity across industries and private sector businesses, corresponding with the view that these support rural economic resilience (Steiner & Atterton, 2014). The extended conceptual breadth described above aligns with nuanced and multidimensional understandings of prosperity which encompass non-economic values (Jackson, 2009; Moore & Woodcraft, 2019). The framework thus reflects a novel scope for measuring multidimensional IG.

Table 2. IG framework.

Theme	Dimension	Potential indicators
Inclusion*	Income*	Benefits*
		In-work tax credits*
		Low earnings*
	Living costs*	Housing affordability*
		Housing costs*
		Fuel poverty*
	Labour market inclusion*	Unemployment*
		Economic inactivity*
		Workless households*
Prosperity*	Output growth*	Output*
		Diversity of business interests
		Private sector businesses*
		Wages/earnings*
	Employment*	Workplace jobs*
		Private/public sector employment balance
		People in employment*
		Employment in low pay sectors*
	Human capital*	Educational attainment*
		Intermediate/higher-level skills*
		Type of workforce, occupations*
Physical geography and connectivity	Physical geography	Urban-rural classification
		Mainland or island
		Population density or sparsity
		Natural hazards
	Access to services	Transport infrastructure
		Access to key services
		Digital connectivity
	Environmental assets	Area of carbon ‘sinks'
		Renewable energy potential / existing schemes
		Area protected / designated in some way
Population and social characteristics	Social strength	Community strength and cohesion
		Commuting
		Second or empty homes
	Social infrastructure	Existence of community centres, libraries and similar facilities, community owned land
	Social vulnerability	Age structure
		Rates of disability, poor health
		Ethnicity
		Gender
		Public transport infrastructure

Note: Some descriptions of potential indicators have been shortened.

Source: Adapted and extended from IG Monitor framework (Beatty et al., 2016). Themes, dimensions and indicators replicating or partly reflecting those in the IG monitor (Beatty et al., 2016, p. 4) are starred (*).

Indicators were compiled and selected in an iterative process which began with a review of data sources, indicators and metadata aligned with potential indicators from the framework (Figure 2: step 2). This focused on online resources and indicators from relevant outputs within Scotland (Appendix 1 in the online supplemental data). Datasets and indicators available or calculable at Data Zone level (small areas drawn to contain 500-1000 people⁴) were flagged. Gap analyses identified levels of data availability for potential indicators. Next, an iterative and evolving indicator selection process (Figure 2: step 3) led to full coverage of the twelve dimensions in the framework. Criteria for indicators noted during the project included conceptual ‘fit’, data recency, spatial granularity, regularity of data collection, and applicability to a wide range of the population. These correspond with qualities typically considered during indicator selection (see Niemeijer & de Groot, 2008).

Additionally, further criteria were noted which influenced the form of some final indicators, which were informed by the required granularity of data, and challenges posed by rural landscapes and regional geography for measuring socio-economic characteristics. The modifiable areal unit problem – the influence of boundaries on properties of aggregated data (see Buzzelli, 2020) is pertinent as Data Zones are less compact in rural and island areas (Flowerdew et al., 2007⁵); interpreting spatial autocorrelation (see Buzzelli, 2020, p. 169) in similar areas is complicated as nearby Data Zones may have poor links. Also, the residence-based geographies mean that important dimensions of IG may not be recorded if they are outside unit boundaries. Corresponding with the influence of geography and connectivity on IG in the region (Table 2), and the potential importance of within-country comparisons (Larsson et al., 2021) and place-based experiences of the economy (McKay, 2019) in understanding being left behind, many final indicators representing economic outputs and characteristics, and accessibility, were calculated from aggregations of data which intersected the estimated area within 30 min’ travel of each Data Zone. This threshold has been used to identify sparsely populated areas based on access to people in Scotland (Hopkins & Piras, 2020). Two indicators representing environmental assets were calculated based on proximity, and indicators of output growth and employment used data reflecting workplace locations. Indicators representing totals and counts were expressed as rates per 1000 people: an appropriate weighting, given the uneven regional population distribution.

More precisely defined indicators were developed after data were downloaded for the majority of selected indicators (Figure 2: step 4). The final dataset (Table 3) covered all dimensions across the framework. Some indicators represent adapted versions of ‘final’ specific indicators, some specific indicators were not calculated or not included in the analysis, and two indicators were calculated at a later stage.

Table 3. Indicators included in the dataset: raw data sources and descriptive statistics (to two decimal places).¹⁰

Indicator name	Code	Data sources	Type	n	mean	min	max	sd	n unique values
Data Zone code	DZ11		C	630	NA	NA	NA	NA	630
Percentage of people who are employment deprived (2017) (%)	INC_EMPDEP	Scottish Government (SG)	N	630	7.7	1	35	4.98	29
Number of children in working families, receiving tax credits (2017/18) (estimated rate per 1,000 children)	INC_CHILDR	HM Revenue and Customs; National Records of Scotland (NRS)	N	630	316.22	56.82	735.29	120.95	404
Percentage of people who are income deprived (2017) (%)	INC_INCDEP	SG	N	630	9.59	1	46	6.37	36
Percentage of people in households that are overcrowded (2011) (%)	INC_OCROWD	NRS (Census); SG	N	630	7.07	0	37.62	4.42	618
Percentage of dwellings which are vacant or second homes (2019) (%)	LIV_VSHOME	NRS	N	630	7.95	0	45.79	6.82	600
Percentage of dwellings which are in council tax bands A–C (2019) (%)	LIV_TAXABC	NRS	N	630	61.47	0	100	24.71	615
Claimant count^11, as a proportion of residents aged 16–64 (2019) (maximum value, %)	LAB_CCOUNT	Office for National Statistics (ONS)	N	629	3.28	0	15.1	2.4	101
Average annual range of claimant count (2016–19) (percentage points)	LAB_ARANGE	ONS	N	609	1.32	0	6.12	0.71	206
Impact of COVID-19 pandemic on claimant count (March–September 2019 to March–September 2020) (percentage point change between maximum claimant count)	LAB_CORONA	ONS	N	625	3.26	−1	9.6	1.81	203
Estimated GVA of accessible area (2017) (£ million)	OUT_GVAEST	ONS; Geographic Information Science and Analysis Team, Scottish Government (hence in this table: GISAT, SG); ONS Geography; NRS (Census); SG; Hopkins and Piras (2020)	N	630	1006.61	6.43	3063.18	938.36	470
Estimated GVA of accessible area (2017) (£ million per 1,000 employees)	OUT_GVAEST1000	ONS; GISAT, SG; ONS Geography; NRS (Census); SG; Hopkins and Piras (2020)	N	630	53.54	37.21	88.21	5.71	470
Accessible workplaces: industrial diversity (2020) (Simpson Index^12)	OUT_DIVERS2020	ONS; GISAT, SG; Hopkins and Piras (2020)	N	630	0.88	0.59	0.91	0.04	250
Accessible workplaces: proportion of micro-businesses (2020) (%)	OUT_MICROB2020	ONS; GISAT, SG; Hopkins and Piras (2020)	N	630	83.22	75.84	93.88	3.35	244
Number of accessible private sector workplaces (2020) (number)	OUT_PWORKN2020	ONS; GISAT, SG; Hopkins and Piras (2020)	N	630	2543.75	120	6260	1847.57	229
Median gross household income estimate (2017) (£/week)	OUT_INCOME	SG	N	630	610.09	319.74	1170.58	111.68	629
Accessible employment within the private sector (2018) (employment)	EMP_PEMPLN2018	ONS; GISAT, SG; Hopkins and Piras (2020)	N	630	13,918.19	78	46,188	13,909.79	468
Percentage of accessible employment within the private sector (2018) (%)	EMP_PEMPLP2018	ONS; GISAT, SG; Hopkins and Piras (2020)	N	630	73.7	12.7	100	7.69	470
Percentage of accessible workplaces within the private sector (2020) (%)	EMP_PWORKP2020	ONS; GISAT, SG; Hopkins and Piras (2020)	N	630	93.41	87.5	96.84	1.57	236
Attainment of school leavers (2015–16 to 2017–18) (score, based on highest level of qualification attained for pupils of publicly funded secondary schools)	HUM_ATTAIN	SG	N	627	5.54	3.65	6.45	0.43	342
Degree of remoteness from urban areas	PHY_REMOTE	SG	C	630	NA	NA	NA	NA	4
Settlement size	PHY_SETTLE	SG	C	630	NA	NA	NA	NA	3
Population sparsity (within or outside the Sparsely Populated Areas)	PHY_SPARSE	Hopkins and Piras (2020)	C	630	NA	NA	NA	NA	2
Number of accessible child care places (2020) (within 30 min of Data Zone centroid, per 1000 children)	ACC_CHILDCPL1000	Care Inspectorate; ONS Geography; NRS; GISAT, SG; Hopkins and Piras (2020)	N	630	175.42	44.28	465.75	46.72	510
Sum of average drive times to a GP surgery and retail centre (2018) (minutes)	ACC_DRIVET	SG	N	630	16.74	1.79	195.97	20.51	630
Sum of average travel times by public transport to a GP surgery and retail centre (2018) (minutes)	ACC_PUBLIC	SG	N	630	38.88	6.79	265.44	33.61	630
Percentage of premises receiving a superfast connection (January 2019) (%)	ACC_BBFAST	Ofcom; NRS	N	628	76.79	0	100	27.4	481
Capacity of renewable schemes, operational or awaiting/under construction (2020) (MW) (within 50 km of Data Zone centroid)	ENV_RENEWA	Department for Business, Energy and Industrial Strategy; GISAT, SG; ONS Geography	N	630	1378.83	0	3222.8	826.02	310
Area within National Scenic Areas (hectares) (within 50 km of Data Zone centroid)	ENV_SCENIC	GISAT, SG; ONS Geography	N	630	58,456.72	0	276,092.82	61,861.08	548
Number of accessible active charities, which operate locally (2020) (within 30 min of Data Zone centroid, per 1000 people)	SOCS_CHARIT1000	Office of the Scottish Charity Regulator; ONS Geography; NRS; GISAT, SG; Hopkins and Piras (2020)	N	630	6.43	2.82	37.5	3.68	567
Number of accessible key community facilities (2020) (within 30 min of Data Zone centroid, per 1000 people)	SOCI_FACILI1000	Ordnance Survey; NRS; GISAT, SG; Hopkins and Piras (2020)	N	630	6.18	2.57	36.28	4.25	577
Old age dependency ratio (2019) (old age population (65 or above) as % of working age population (16–64))	SOCV_OLDAGE	NRS	N	630	40.25	2.1	97.5	14.66	621
Percentage of single births of low birthweight (2017–19) (%)	SOCV_BIRTHW	Public Health Scotland	N	630	1.88	0	25	3.82	49
Comparative Illness Factor (2017) (standardised ratio)	SOCV_ILLFAC	SG	N	630	80.71	20	280	36.01	40
Standardised mortality ratio (2014–15 to 2017–18) (indirectly standardised ratio)	SOCV_MORTAL	SG	N	630	90.94	19.49	719.78	39.67	630
Percentage of residents who belong to ethnic groups other than ‘White: British’ (2011) (%)	SOCV_NONWBR	NRS (Census)	N	630	4.73	0.3	25.99	3.14	620
Gender balance (2019) (50% – female %, absolute value)	SOCV_GENDER	NRS	N	630	2.01	0	18.58	1.82	593

Source: Data analysis, fuller details of data sources: Appendix 1. Some ‘Code’ values and indicators are based on dimension names and indicators (respectively) in IG Monitor (Beatty et al., 2016, p. 4).

The framework was implemented using multivariate analysis, which identified specific ways that uneven regional contexts intersected with evidence of prosperity and inclusion. This identifies components for measuring IG a posteriori and evaluates these in the context of spatial justice. This adds considerable depth to measurement, meeting the demand for research into ‘a broader range of variables to capture the multi-dimensionality of inequalities’ (Patias et al., 2022, p. 162) and learning about the diversity of left-behind places (e.g., Martin et al., 2021).

Data handling and analysis used R (R Core Team, 2022; packages: Appendix 1; RStudio, PBC, 2022, 2023) with some processing in ArcGIS (Esri Inc., 2019). The regional dataset was prepared for analysis (Figure 2: step 5): 32 missing values in five variables were infilled using mean values for area types in the Urban Rural Classification (Scottish Government Geographic Information Science & Analysis Team, Rural and Environment Science and Analytical Services Division, 2022; variables: PHY_REMOTE, PHY_SETTLE). High Pearson coefficients (absolute r > 0.7) between numeric variables were identified and eight variables were removed, reducing potential multicollinearity (see Field et al., 2012, pp. 770–1, who used 0.8). Remaining numeric variables used different measurement scales and so were standardised.

The analysis used exploratory factor analysis and cluster analysis (Figure 2: step 6), techniques regularly used to produce rural classifications (Nelson et al., 2021). Factor analysis aims to produce measures explaining variation in a larger dataset (Watkins, 2018), and was used to identify and measure combinations of variables from the primary and contextual themes representing intrinsic elements of IG performance. 22 numeric variables assessed as representing strengths and weaknesses in inclusion or prosperity were analysed. The Kaiser-Mayer-Olkin Measure of Sampling Accuracy for this dataset was 0.8, signifying good suitability for factor analysis (Kaiser, 1974). Parallel analysis, recommended for identifying the number of underlying factors (Hayton et al., 2004), suggested six factors, while the scree plot of eigenvalues (see Tabachnick & Fidell, 2001, p. 621) indicated four factors as optimum. Analyses specifying 3–10 factors were implemented using oblimin rotation: oblique rotations are more generally preferred (Osborne, 2015) and the factor correlation matrix for the seven-factor solution shows correlations above an absolute value of 0.32, favouring oblique rotation (Tabachnick & Fidell, 2001, pp. 622–3). As factor analysis should seek to improve interpretability (Osborne, 2015), substantial loadings (absolute thresholds: 0.3 and 0.4) within the 5–10 factor solutions were reviewed; the seven-factor solution had the lowest number of variables loading on multiple factors.

Factor scores: estimated values for these concepts (see Tabachnick & Fidell, 2001, pp. 626–7) were calculated using regression, labelled, and incorporated into hierarchical cluster analyses (a means of identifying groups of similar cases within a dataset (Crawley, 2013, p. 819)), specifying Ward’s method and Euclidean distance. Two sets of variables were investigated: (a) the seven factor scores, and (b) the factor scores plus three standardised demographic indicators measuring legally protected characteristics, corresponding with social vulnerability. 26 metrics within the R package ‘NbClust’ (Charrad et al., 2014)⁶ were used to identify the most suitable number of clusters, finding lower variation in numbers of clusters for the second set of variables with two and five cluster solutions having most recommendations. Cluster counts below five were judged to lack sufficient nuance, and classifications of Data Zones into between five and nine categories were generated. The five- and nine-cluster solutions were explored in more detail, using descriptive statistics and their distribution across the Urban Rural Classification. The former, with the highest number of recommendations, is interpreted (Figure 2: step 7).

This analysis is conceptually grounded in the multidimensional IG framing and is regionally-sensitive by including contextual themes; framework co-construction with regional experts; and indicators which are sensitive to regional and rural measurement issues. The analysis is also inductive, and is well suited to the analysis of a vague concept with a limited evidence base (e.g., Lee, 2019; Statham & Gunson, 2019) but where collaboration had identified relevant topics to consider.

3. RESULTS

The factors representing intrinsic dimensions of IG performance were interpreted using high loadings (Table 4), based on the commonly-used 0.3 threshold (Field et al., 2012, p. 767). The factor explaining most variance is labelled ‘disadvantaged’, based on very high loadings for benefit claimant rates (positive), income (negative) and lower-cost and overcrowded housing (positive). Other loadings reflect correlated negative outcomes. Two other factors explain over 9% of variance: ‘big output’ has large loadings for accessible economic output (positive) and the proportion of accessible workplaces which are micro-businesses (negative). ‘Rural services’ has positive loadings for journey times to services and the proportion of second homes, and a negative relationship with superfast broadband access: collectively suggesting limited access to essential services. ‘Quality of life’ explains over 8% of variance: the largest loading signifies the nearby extent of National Scenic Areas, a designation reflecting landscape attractiveness (Countryside Commission for Scotland, n.d.); renewable energy capacity has a negative loading, likely reflecting planning policy for wind farms (Scottish Government, 2014). The remaining factors explain below 6% of variance, reflecting private enterprise, the availability and accessibility of supportive services (childcare places and charities), and the last factor was labelled ‘small diverse businesses’. Mapping (Figure 3) indicates considerable geographical variation: for example, high values for ‘disadvantaged’ are apparent in built-up areas, including parts of Inverness and nearby towns, but relatively high values are found near towns in the west (e.g., Fort William, Oban, Rothesay, Campbeltown) and within some remoter rural areas and outlying islands. High values for ‘big output’ correspond with economic centres and surroundings: particularly Inverness and Elgin, but clusters also appear near west coast towns, in the Western Isles (Stornoway), and on the Shetland and Orkney mainlands. ‘Rural services’, ‘quality of life’ and ‘community support’ appear to have varying west–east gradients.

Figure 3. Maps: intrinsic dimensions of IG performance.

Source: Spatial data shown: Appendix 1.

A series of seven maps of Scotland, each including an inset map of the Inverness and Elgin area, showing the distributions of factor scores representing intrinsic dimensions of inclusive growth performance in the Highlands and Islands – disadvantaged, big output, rural services, quality of life, private sector, community support, and small diverse businesses.

Table 4. Factor analysis: 22 standardised variables, oblimin rotation.

	Disadvantaged	Big output	Rural services	Quality of life	Private sector	Community support	Small diverse businesses
LAB_CCOUNT	0.844
LIV_TAXABC	0.779
INC_OCROWD	0.734
OUT_INCOME	−0.707
INC_CHILDR	0.595
HUM_ATTAIN	−0.535
SOCV_MORTAL	0.384
SOCV_BIRTHW
OUT_GVAEST		0.777		−0.344
OUT_MICROB2020		−0.763
ACC_DRIVET			0.868
ACC_BBFAST			−0.68
LIV_VSHOME			0.524
ENV_SCENIC				0.633
EMP_PWORKP2020				−0.543	0.391
ENV_RENEWA				−0.459
LAB_CORONA				0.419
EMP_PEMPLP2018					0.885
ACC_CHILDCPL1000						0.866
SOCS_CHARIT1000				0.335		0.419
OUT_DIVERS2020		0.343					0.58
OUT_GVAEST1000							−0.521
Variance explained (%)	15.26	9.62	9.04	8.23	5.88	5.64	4.42

Note: Absolute loadings of 0.3 or more shown to three decimal places.

Source: Data analysis. Some ‘Code’ values are based on dimension names and indicators (respectively) in IG Monitor (Beatty et al., 2016, p. 4).

The emergent dimensions of IG are thematically influenced by the regional context of extensive rurality and sharp urban-to-rural transitions, while others reflect established concerns of IG with economic inequality (disadvantaged, big output). Spatial concentrations in dimensional strength near large settlements, and rural variation, are visually apparent (Figure 3). The potential utility of this knowledge in driving understanding of multidimensional development and spatial justice, and the value of extending IG’s scope to reflect regional, rural-influenced contexts affecting development, are apparent.

The five-fold typology of IG performance (Figure 4) is significantly associated with settlement type and accessibility to urban areas (Fisher's Exact Tests: p < 0.001⁷) as defined by the Urban Rural Classification. However, within-cluster distributions of the latter (Table 5) suggest a more complex relationship between these and heterogeneity in IG performance. While cluster 2 is predominantly urban and cluster 5 is only found in very remote rural areas, the same area types do not form a reliable guide for a location's profile of IG performance; therefore, an indication of the diversity of rural areas has begun to emerge. For example, very remote areas form over two-thirds of the small areas in three clusters (1, 3, 5), and rural areas are a majority of small areas in these clusters, and are approximately 33% of those in cluster 4.

Figure 4. Map: five cluster typology.

Source: Spatial data shown: Appendix 1.

A map of Scotland, including an inset map of the Inverness and Elgin area, showing the distributions of the five clusters which form a typology of inclusive growth performance.

Table 5. Descriptive statistics showing means and standard deviations (in brackets) to two decimal places and distribution of Data Zones.

Cluster	1	2	3	4	5	Region
Disadvantaged	0.03 (0.68)	0.8 (1.04)	0.16 (0.95)	−0.32 (1.03)	0 (0.59)	0 (0.95)
Big output	−0.55 (0.52)	0.96 (0.63)	−0.64 (0.56)	0.98 (0.4)	−2.02 (0.14)	0 (0.94)
Rural services	0.48 (1.02)	−0.66 (0.27)	0.04 (0.58)	−0.48 (0.41)	4.03 (1.06)	0 (0.92)
Quality of life	1.13 (0.53)	−0.03 (0.88)	−0.33 (0.68)	−0.61 (0.3)	−0.24 (0.27)	0 (0.9)
Private sector	0.05 (1.14)	0.26 (0.51)	−0.1 (0.92)	0.11 (0.36)	−2.6 (2.1)	0 (0.91)
Community support	0.57 (1.33)	−0.12 (0.21)	−0.25 (0.64)	−0.2 (0.28)	−0.34 (1.16)	0 (0.89)
Small diverse businesses	0.5 (0.6)	0.18 (0.34)	−0.44 (0.78)	0.15 (0.32)	−4.61 (1.57)	0 (0.88)
Old age dependency ratio	0.37 (0.96)	−0.96 (0.69)	0.07 (0.79)	−0.25 (1.09)	0.86 (0.6)	0 (1)
Ethnic group: not ‘White: British'	−0.11 (0.63)	2.92 (1.17)	−0.42 (0.6)	0.03 (0.7)	−0.4 (0.39)	0 (1)
Gender balance	0.04 (0.82)	0.68 (1.33)	−0.22 (0.7)	0.07 (1.24)	−0.33 (0.42)	0 (1)
Number of Data Zones	175	35	202	210	8	630
Urban (population > = 10,000) (%)	5.71	74.29	0.99	50	0	22.7
Small town (population 3000–9999) (%)	20	22.86	43.56	17.14	0	26.51
Rural (population < 3000) (%)	74.29	2.86	55.45	32.86	100	50.79
Accessible (within 30 min drive from urban settlement) (%)	2.29	5.71	4.95	34.76	0	14.13
Remote (30–60 min drive from urban settlement) (%)	9.14	14.29	26.24	14.29	0	16.51
Very Remote (more than 60 min drive from urban settlement) (%)	82.86	5.71	67.82	0.95	100	46.67

Source: Data analysis. Urban Rural Classification definitions adapted from Scottish Government Geographic Information Science & Analysis Team, Rural and Environment Science and Analytical Services Division, 2022, p. 4.

New evidence on spatial justice is provided by evaluation of cluster profiles (Table 5), and the most important dimensions for differentiating between the clusters, measured using mean scores for IG performance and demographics, and multiple statistical comparisons (using the Mann–Whitney U test) of indicators between all cluster pairs. Firstly, cluster 1 covers a large area across western and mountainous areas, including several of the western islands. This cluster is distinctive in terms of high mean scores for quality of life (1.13), community support (0.57) and small diverse businesses (0.5): these are the highest averages for all clusters. It is generally relatively remote from centres of economic activity (Big output: −0.55) and services (Rural services: 0.48) and has evidence of population ageing (Old age dependency ratio: 0.37).

By contrast, cluster 2 is concentrated in urban areas and towns, and is notable for its demography and evidence of socio-economic disadvantage: it features a relatively healthy population structure (Old age dependency ratio: −0.96) with evidence of high diversity (Ethnic group: not ‘White: British': 2.92) although a gender imbalance (Gender balance: 0.68). The cluster has the highest average score for ‘Disadvantaged’ (0.8), and its situation in built-up areas is reflected in the lowest mean score of any cluster for ‘Rural services’ (−0.66) and a high score for ‘Big output’ (0.96).

The types of places in cluster 3 share similarities with those in cluster 1: many small areas are rural and remote, but small towns are more common in cluster 3, and its remoteness is slightly lower. Most small areas in this cluster are in the north, forming a periphery to Inverness and Elgin, alongside more central parts of Shetland and Orkney and scattered Data Zones in the Western Isles and south-west. Across dimensions of IG performance, cluster means are not extreme, but the low average for big output (−0.64) and other negative mean scores could suggest experience of slight disadvantage.

Cluster 4 contains a third of the Data Zones, appearing as a ‘commuter belt’ to population centres near the Moray Firth, also containing parts of these built-up areas. The cluster has the lowest average for quality of life (−0.61) but is close to economic centres as the big output mean is the highest of any (0.98). This cluster is also the least disadvantaged of any (−0.32).

Cluster 5 contains eight small areas in remote parts of Orkney and Shetland. It has a distinct profile, having the lowest average score for big output (−2.02) and the highest (4.03) for rural services. It has very low scores for private sector (−2.6) and small diverse businesses (−4.61) and supportive services in communities are typically less available (community support: −0.34, the lowest average). The highest average old-age dependency ratio (0.86) signifies likely demographic imbalance.

Bivariate comparisons (Figure 5) find that the cluster pairs 1 and 2, and 3 and 4, show the largest differences, as only one of ten indicators do not significantly differ between these. For clusters 2 and 3, and 2 and 5, there are two indicators with no evidence of significant differences in values. The magnitude of differences between clusters 3 and 4 is notable, as these share a considerable virtual border near the Moray Firth. Some Data Zones in cluster 2, associated with smaller built-up areas (Aviemore, Oban, Fort William) are also contiguous to cluster 1. The aspects of IG performance and demographics which contribute most to these differences can be evaluated by calculating, for each variable, the proportion of cluster comparisons where the null hypothesis (of equal values between clusters) was rejected. This ratio was highest (1) for rural services, followed by small diverse businesses and the old age dependency ratio (0.9), big output (0.8) and quality of life (0.7). Other indicators had a ratio of 0.6, except community support (0.4). Therefore, the enhanced IG conceptual framework has been implemented to generate a fine-grained geography of multidimensional development, providing insights into rural diversity and potential issues of spatial justice aligning with IG.

Figure 5. Graph: the number of variables (intrinsic dimensions of IG performance, demographic variables) which are not significantly different between each pair of clusters.

A network graph, with the five clusters represented as points, linked by lines between these points. The number of lines between two points shows the number of variables (out of ten) which are not significantly different between the clusters.

4. DISCUSSION

Although ‘narrow’ definitions of IG are being marginalised, the term resonates due to its concern with inequality (Lee, 2019). As there is increasing evidence that inequalities are driven by diversity in local-level trends and their distribution across space (Patias et al., 2022), and that disadvantaged regions are heterogenous (Martin et al., 2021; Pike et al., 2023; Velthuis et al., 2022), both the methods for understanding IG, and its conceptual framing require transformation. The former is related to the view that current knowledge, aims and practices in regional studies are inadequate for the urgent need to address plural non-income inequalities (Martin, 2021). The latter is supported by the view of prosperity as the ability of people to access holistic, enabling ‘entitlements’ (Jackson, 2009, p. 35) and the contribution of multiple capitals to sustainable development (Flora et al., 2018).

This article argues that ‘IG performance’ is a function of the intersections of emergent intrinsic dimensions of prosperity and inclusion, which are affected by the varying influences of geographical and social contexts and are likely to be regionally distinctive. Evaluating how differences in these development profiles and their geographical distribution reflect spatial justice enhances the perspective that can be provided by a framing of IG on the wellbeing economy. This position supports the need to understand prosperity as multidimensional and grounded in local experiences (Moore & Woodcraft, 2019) and represents a nuanced development from broad definitions of IG and arguments that it should be reframed as multidimensional (Hay et al., 2020). This is validated by the spatial distribution of the clusters, suggesting effects of settlement type and location on IG performance, and place-based and region-specific influences on intrinsic dimensions of IG. Applying the extended framework has also led to greater understanding of rural IG, contributing to reducing uncertainties in interpreting and measuring this, partly due to its more common urban framing (e.g., Lee, 2019) and aligning with understanding of differing regional interpretations of IG (Waite & Roy, 2022). ‘Rural services’ identifies challenges in access to services and housing, which are important influences on community sustainability, and this factor emerged as a major point of difference between clusters. ‘Quality of life’ captures attractive rural landscapes, which form an asset as similar locations have been a preference of creative workers (McGranahan & Wojan, 2007). The benefit of enhancing an existing IG framework using regional expertise was forming a boundary of conceptual interest, from which underlying dimensions of IG performance could emerge.

Factors underpinning IG, and groups of locations with similar characteristics, have transformed the ‘fuzzy’ concept of IG (Lee, 2019) within the region. By measuring inequalities within the region and defining the meaning and scope of IG at this level, the analysis applies and extends the concept of spatial justice and its holistic consideration of multi-scalar, cross-boundary inequality (Madanipour et al., 2022). This is valuable as a spatial justice perspective on rural development means opening up the potential for multiple, holistic, regionally-informed definitions of such development to form (Mahon et al., 2023). Differences between the output clusters in IG performance could help to prioritise development issues and inform place-specific policy interventions (Iammarino et al., 2019) to support spatially just IG. For example, in the Moray Firth area, development strategies should aspire to meet the distinct needs of populations in some built-up areas, who are potentially experiencing disadvantages and limited ability to access local economic opportunities (cluster 2); a more prosperous commuter zone (cluster 4); and a remoter periphery (cluster 3).

Cluster 2 also includes parts of smaller built-up areas which are contiguous with cluster 1. The latter covers many upland, coastal and island areas: average values suggest ageing communities in attractive landscapes, with good access to supportive community services. High numbers of differences between clusters 2 and 1, and 2 and 3 suggests that cluster 2 may reflect types of left-behind places. Cluster 5 also appears left behind for different reasons, showing evidence of remoteness and demographic fragility. Identification of the detailed characteristics and spatial distributions of (potentially) left-behind areas is valuable, given a call to understand multidimensional profiles of left-behind places to ‘offer(s) the prospect of interpreting their predicaments and potentials in new ways’ (Pike et al., 2023, p. 2) and building on existing understanding of the diversity of left-behind, disadvantaged regions (Velthuis et al., 2022). The heterogeneity of peripheral and island areas, represented by the distributions of clusters 1, 3 and 5, and parts of cluster 2, is particularly notable. This work has strengthened and developed the IG concept, finding that an interpretation which is multidimensional, regionally-sensitive, and implemented at a granular level can identify rural diversity; this could form part of the long-term advances in the understanding and measurement of rurality (Nelson et al., 2021) and complement the ‘Inclusive Growth-in-cities agenda’ (Lee, 2019, p. 428). Just as left-behind places require ‘a reframing of development thinking and policy’ grounded in their circumstances and experiences (MacKinnon et al., 2022, p. 51), we argue that places may also be left behind through inadequate identification of challenges and opportunities which are multidimensional, rural, localised, or not aligning with existing geographies.

5. CONCLUSION INCLUDING POLICY RECOMMENDATIONS

Evidence suggests that IG is conceptually enhanced through closer integration with regional contexts of geography and society, and their implications for development potential and community resilience. This transformation is essential due to the multidimensional form of wellbeing and inequalities affecting left-behind places and goes beyond specifying ‘broad’ understandings of IG. The bounded, but data-driven and inductive, approach to operationalising this framework has identified within-region heterogeneity, represented as intrinsic dimensions of IG performance and a multidimensional classification, to which a spatial justice perspective on multidimensional inequalities was applied.

Although the analysis was regional, we suggest that themes and dimensions in the multi-level framework (Table 2) are widely applicable, while more detailed potential indicators could be adapted to different circumstances. Limitations of the analysis include the challenge of quantitative measurement in rural areas posed by less compact population-based units (Flowerdew et al., 2007). The analysis utilised publicly accessible datasets containing rounded and estimated values which are subject to some uncertainty, which has not been quantified; some selected indicators were also not calculated (e.g., due to computing speed limitations), or not used due to re-use permission constraints. Calculating aggregated data based on travel time provides arguably more valid measurements for a geographically diverse region, however these times are estimates due to assumptions and inaccuracies in the transport network model (Hopkins & Piras, 2020, pp. 21–22). The analysis also presents a ‘snapshot’ of development, however, other analyses have examined inequalities through changes to classifications over time (Hughes & Lupton, 2021; Patias et al., 2022) and an analysis using the same indicators could calculate typologies based on multiple years’ data (as in Patias et al., 2022) and assess change in IG performance over time. Finally, the analysis has not investigated within-cluster variability. Although averages are used to interpret typologies (Velthuis et al., 2022), the clusters are not homogenous. Development to support wellbeing should address inequalities at multiple scales, given the effect of surrounding regions on feelings of being left-behind (Larsson et al., 2021).

A key implication of this analysis is that policies and strategies which prioritise IG or similar goals, which are targeted at existing regions or geographies, should be sensitive to localised multidimensional inequalities which do not align with boundaries or established typologies. Operationalising context-informed IG performance can provide important information given the potential unequal impacts of city region-based development approaches in diverse rural regions, and questions over their appropriateness in rural areas (Beel et al., 2020). The diversity in IG performance in peripheral areas, and the role of rural services and infrastructure in generating this, confirms the value of developing the conceptual debate on IG beyond ‘narrow’ and ‘broad’ framings (e.g., Evenhuis et al., 2021) to recognition of it as regionally distinctive interactions of prosperity, inclusion, geography and society.

GEOLOCATION INFORMATION

Highlands and Islands region, UK (57.366032^oN, −4.47425^oW). Co-ordinates source: Appendix 1.

ACKNOWLEDGEMENTS

We acknowledge the contributions of Andy Sarjeant, Catherine MacNeil, Eilidh MacDonald, Rachel Forrest and Heather Smith (HIE) and Ruth Wilson (James Hutton Institute) to ToWards Inclusive Growth; and acknowledge feedback on a previous version of this paper from David Brown and Paula Duffy within the Transatlantic Rural Researchers Network (Newcastle, 2022). An earlier version of this paper was presented at the Regional Studies Association Annual Conference (Hopkins et al., 2023) and the research has been described in a newsletter article (BioSS, 2023) and in lectures; a previous version of the analysis has been published online (Hopkins et al., 2021 and ⁸) as has the framework (⁹).

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The ToWards Inclusive Growth project was funded by the SEFARI Gateway Responsive Opportunity Initiative and the Rural and Environment Science and Analytical Services (RESAS) Division as part of the Scottish Government's Strategic Research Portfolio (2016-22: funding from 2022-27 Programme acknowledged [JHI-E1-1, JHI-E2-2]). Views expressed are those of the authors and are not necessarily those of the Scottish Government or RESAS.

Notes

1 https://www.hie.co.uk/about-us/.

2 https://www.gov.scot/policies/cities-regions/city-region-deals/ and https://www.gov.scot/policies/cities-regions/regional-growth-deals/.

3 https://www.legislation.gov.uk/ukpga/2010/15/section/4.

4 https://spatialdata.gov.scot/geonetwork/srv/eng/catalog.search#/metadata/7d3e8709-98fa-4d71-867c-d5c8293823f2.

5 This analysis considered an older definition of Data Zones.

6 Maximum number of clusters specified: 32.

7 Based on Monte Carlo simulation (2000 replicates).

8 https://sefari.scot/research/understanding-inclusive-growth-in-the-highlands-and-islands.

9 https://researchontheedge.org/what-makes-up-inclusive-growth/.

10 The ‘Code’ (before the underscore) indicates the framework dimension which the indicator represents. Indicators describing ‘accessible’ characteristics calculated as aggregations using 30 min service areas. Variable types: ‘N’: numeric/integer, ‘C’: character/categorical, n: non-missing values.

11 Datasets on claimant counts (used to calculate the three LAB_ variables) were downloaded separately in 2020 and 2021 and gave different indicator values, likely due to data updates. The 2020 data was used in analysis as it reproduced earlier results.

12 Royal Geographical Society (n.d.).