Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia

Journal of Rural Studies xxx (2014) 1e14

Contents lists available at ScienceDirect

Journal of Rural Studies journal homepage: www.elsevier.com/locate/jrurstud

Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia Matthew Tonts a, *, Paul Plummer a, Neil Argent b a b

School of Earth and Environment, The University of Western Australia, Australia Division of Geography and Planning, The University of New England, Australia

a b s t r a c t Keywords: Evolutionary economic geography Path dependence Resilience Western Australia

This paper draws on the emerging field of evolutionary economic geography to offer insights into the transformation of rural economies. In particular, it focuses on the concepts of path dependence and resilience, and the ways in which these help to explain change within four case study local areas in rural Western Australia. The paper draws on recent advances in dynamic econometrics to examine the ways in the major economic shock of the late 1980s and early 1990s restructuring process ‘unlocked’ these local economies from existing development pathways and reshaped their trajectories. The paper finds that while common trends were evident across the four case study areas, the ways in which they responded and recovered from the shock were quite different reflecting the diverse ways in which multiscalar processes play out across rural space economies.
2014 Published by Elsevier Ltd.

1. Introduction One of the ongoing themes within rural studies over the past two decades has been the attempt to understand how localities and regions respond to economic, political and environmental upheaval. Using concepts and methods drawn from perspectives as diverse as Marxian political economy, behavioural studies, neoclassical economics and post-structural theory, rural social scientists have paid close attention to subjects ranging across agricultural adjustment and the farm crisis, local economic restructuring, poverty and disadvantage and environmental degradation. In very simple terms, one of the ongoing themes in this work is to understand both the drivers and implications of change and continuity. Parallel to these interests in rural studies, economic geographers have been addressing a similar meta-narrative about change and continuity, albeit in quite different locational and sectoral contexts. One of the perspectives to emerge out of this engagement with the dynamics of the uneven development is ‘evolutionary economic geography’ (MacKinnon et al., 2009). This emerging paradigm emphasizes the significance of history and geography in understanding the development of space economies by bringing together a number of evolutionary concepts and metaphors, such as ‘selection and adaptation’, ‘path dependence’ and ‘hysteresis’. Essentially,

* Corresponding author. E-mail address: [email protected] (M. Tonts).

the evolutionary economic geography project is about accounting for the transformations in capitalist economies over multiple temporal and spatial scales. While much of the work on evolutionary economic geography has been focused on urban systems and industries such as manufacturing and services, it also has considerable potential to contribute to interpretations of rural economic change (Tonts et al., 2012). Yet, to date, rural geographers have tended not to engage directly in debates about the theoretical and methodological efficacy of evolutionary economic geography for understanding the dynamics operating both within and across rural economies. There has, however, been considerable engagement with one of the concepts that is increasingly becoming prominent within evolutionary economic geography e the notion of resilience (e.g. Wilson, 2010, 2012; McManus et al., 2012). For both rural geographers and those engaged more broadly in articulating an evolutionary perspective on the geography of uneven development, resilience has become a popular conceptual lens through which to consider the ability of places to respond to shocks and upheaval (see Adger, 2000). Yet there are some key differences in the way the idea has been applied. In rural geography, the focus has tended to be on individual, social and/or community resilience, with an explicit focus on rural economies being less common (e.g. Franklin et al., 2011; Graugaard, 2012; Skerratt, 2013). This is not to suggest that rural economies have been ignored (e.g. McManus et al., 2012), but that social relations, cultural attributes, and policy considerations have tended to form the bulk of inquiry. Perhaps not surprisingly,

http://dx.doi.org/10.1016/j.jrurstud.2014.04.001 0743-0167/
2014 Published by Elsevier Ltd.

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

2

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

economic geographers have tended to focus their attention more squarely on local and regional economies and related policy and institutional dynamics (e.g. Rafiqui, 2009; MacKinnon, 2012; Mas and Hassink, 2013). Against this background, this paper examines the potentials and limitations of evolutionary economic geography in understanding rural economic transformations, including the emergence of ‘new rural economies’. More specifically, the paper considers a number of core concepts embedded within evolutionary economic geography, including path dependence, hysteresis and resilience, and the ways in which these help to explain continuity and change in four selected local case study areas from rural Western Australia. The paper explores these concepts by examining three standard measures of economic performance: the unemployment rate; employment growth: and wage rates. An important contribution here is the application of recent advances in dynamic econometrics that have rarely been applied in rural geography as a means of both testing some of the conceptual claims in evolutionary economic geography, and also unpacking some of the longer run dynamics and adjustments that have occurred in rural economies. The paper also offers a more qualitative and discursive assessment of the evolution of rural economies, teasing out a number of conceptual issues relevant to the evolutionary economic geography ‘project’ and broader conceptualisations of rural economic transformation. 2. Towards an evolutionary rural geography The idea that economic systems are ‘evolutionary’ is not new. In explaining the development of capitalist systems Marx (1967), for example, placed considerable emphasis on the out-of-equilibrium and dialectical nature of the mechanisms underpinning the dynamics of economic growth and crisis. The economist Veblen (1898) famously posed the question ‘why is economics not an evolutionary science’, while Joseph Schumpeter (1939) gave careful attention to economic change and transformation in describing the process of ‘creative destruction’. In a more spatial context, geographers too have paid due consideration to the nature, causes and consequences of the development of capitalist space economies (e.g. Harvey, 1982; Smith, 1984; Sheppard and Barnes, 1990; Barnes, 1995). One of the common threads underpinning much of this work was that both historical and geographical contingency matter in explaining the structure and performance of local and regional economies. In the field of economics, however, the significance of history tended to be pushed to the background as a result of the dominance of neoclassical general equilibrium theorizing. Typically, neoclassical models of spatial competition postulate globally stable adjustment processes, prioritizing the spatial configuration of the economic landscape over out-of-equilibrium dynamics. Recent research in geographical economics has embraced the possibility of multiple equilibria and the associated issue of selecting amongst locally (un) stable equilibria (Fujita and Mori, 2005; Redding, 2010). In so far as equilibrium selection depends on the adjustment path taken with respect to equilibrium this raises the possibility of a rapprochement between geographical economics and evolutionary economic geography. Nonetheless, the focus on the equilibrium thinking derived from the logic of ‘constrained optimization’ largely overlooks the question of ‘economic emergence’ and the role of behavioural routines, norms, and social and political contingency in determining the complexities of economic landscapes that evolve in ‘historical’ rather than ‘logical’ time (Foster and Metcalfe, 2012). As part of a critical response to neoclassical theorizing, the emergence of a ‘neo-Schumpeterian’ economics in the 1980s began (re)emphasizing the importance of history in explaining economic development (Nelson and Winter, 1982; Dopfer, 2005). This work has focused in particular on

technological innovation, the behaviour of boundedly rational firms, and the role of institutions in driving economic change, which increasingly has drawn on a number of key Darwinian concepts, including ‘selection’, ‘variety’, and ‘self-replication’, or continuity (e.g Essletzbichler and Rigby, 2007). Moreover, considerable attention has been given to the routines and norms that help to explain the behaviour of individuals, firms and institutions in shaping economic systems. One of the most significant ideas to emerge out of this work is the notion of ‘path dependence’ (Hassink, 2005; MacKinnon, 2012). Path dependence has typically been used to describe the development of particular technologies and the ways in which their development is the product of their industrial histories (David, 2001). In other words, the emergence and adoption of new technology is often followed by a gradual process of refinement that sees the technology become ‘locked in’ to a particular development pathway through increasing returns and other positive feedback mechanisms (Arthur, 1994). While much of the initial work on path dependence focused on very specific technologies, it quickly ‘jumped scale’ to incorporate the analysis of economic sectors, institutions and wider economic systems (MacKinnon et al., 2009). Path dependence has also been of considerable interest to geographers, and is one of the central ideas underpinning evolutionary economic geography. For economic geographers, path dependence is often constructed as a form of place dependence in which local or regional outcomes are, in one way or another, shaped by past events and outcomes (Plummer and Tonts, 2013a). In analytical terms, this means understanding the ways in which applications of specific technologies, previous rounds of investment, sunk costs, dynamic increasing returns, institutional structures and social routines all contribute to cumulative and self reinforcing spatial development (Plummer and Tonts, 2013a,b). One of the concerns of economic geographers has been how these cumulative and self-reinforcing processes eventually lock regions into a particular development trajectory. Once this ‘lock in’ occurs, it often takes a large external shock to destabilize (or delock) the system and create entirely new development paths. In a recent contribution, Tonts et al. (2012) have argued that evolutionary economic geography and, in particular, the notions of path dependence and lock-in have considerable value in interpreting processes of rural development. They argue that rural economies and particularly those based around agriculture often have development histories characterized by large sunk costs, specialized skills and knowledge, land use systems, institutional structures, and deeply embedded social and cultural routines that tend to be cumulative and self reinforcing. However, they also point to the role of economic and political shocks in contributing to economic restructuring and the emergence of new rural industries. Most apparent here was the upheaval in the Australian agricultural sector during the 1980s and 1990s as a severe cost price squeeze, rising debt-equity ratios and widespread deregulation led to an economic and social crisis across many rural regions (Lawrence, 1987, 1996; Smailes, 1997; Pritchard, 2000). It is in this context that the complementary notion of resilience is important. Resilience occupies an increasingly prominent position within evolutionary economic geography as a means of explaining the ways in which local and regional economies are able to cope with disruption or shocks and subsequently regain functional capacity (Hudson, 2010). In essence, the work on resilience is focused on the ability of economies to recover following major upheaval. However, it is important to stress that this recovery is rarely implied to mean returning to the same economic state, but, in many cases, a new set of functional arrangements that underpin economic and social wellbeing (Christopherson et al., 2010). Concepts embedded within the resilience literature, such as ‘adaptive

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

capacity’, ‘self organisation’, ‘complexity’ and ‘learning’ have become increasingly influential in recent theoretical work in evolutionary economic geography (Hudson, 2010; Wilson, 2010, 2012; Franklin et al., 2011). However, it is important to note that, as with the other concepts within evolutionary economic geography, the meaning of economic resilience in a spatial context remains ambiguous and contested (MacKinnon and DriscollDeridson, 2012). For example, questions remain about how to account for multi-scalar relations, the extent to which temporal scale is adequately accounted for, and how to unravel the endogenous and exogenous factors that lead to recovery and transformation (see MacKinnon and Driscoll-Deridson, 2012). Moreover, as Christopherson et al. (2010) point out, regional economic resilience is in reality the restatement of an old question: why are some places able to overcome adversity, while others fail? Rural geographers too have become increasingly interested in resilience. The focus here has been more expansive than within evolutionary economic geography, incorporating aspects of socioecological resilience, social and cultural resilience, economic resilience, and resilience policy (McManus et al., 2012; Wilson, 2010, 2012; Glover, 2012). Indeed, one of the characteristics of this work has been a reluctance to reduce resilience to ‘functional’ categories, such as ‘the economy’ or ‘the social’, with an emphasis on a more integrative assessment within the context of particular places. Yet, in broad terms, the narrative remains similar to economic geography in so far as it attempts to understand how places respond to changing circumstances. In the context of rural Australia, this has been particularly apposite where the combination of globalization, neoliberalism, environmental degradation and major social and demographic shifts has resulted in three decades of upheaval, disturbance and uneven development (Lawrence, 1996 Smailes, 1997; Argent, 2011; Pritchard et al., 2007; Pritchard and Tonts, 2011). 3. Modelling rural economic evolution and resilience: path dependence, stability and equilibrium In this section, we outline a theoretical framework for an ‘evolutionary rural economic geography’ through the specification of a relatively simple econometric model of economic evolution and resilience. While formal econometric modelling is rarely used in rural geography, it is our contention that it offers considerable value on at least two fronts. First, it provides a novel basis for operationalizing some of the conceptual arguments and associated claims regarding both economic evolutionary processes and resilience. Second, it offers a means of moving beyond the types of casual empiricism that are often employed in quantitative rural geography, thereby providing deeper insights into some of the processes at work. 3.1. Evolutionary thinking and the resilience framework The recent emergence of evolutionary economic geography as a perspective from which to view the dynamics of local economies typically prioritizes irreversible, historically and geographically contingent processes over neoclassical perspectives regarding spatial equilibrium configurations (Page, 2006; Setterfield, 1995; Robinson, 1979). As we have outlined above, part of a broader ontological and epistemological commitment to evolutionary thinking has seen resilience emerge as a key idea within the literature on evolutionary economic geography, focussing on the ability of local economies to adjust and/or adapt to external disruptions (Hassink, 2010). In this paper, we confine our conceptualization and interpretation of regional economic resilience to the ability of a local rural economic system to absorb shocks, such that it returns or

3

adjusts to a new stable growth path following a ‘big’ event (shock) or regime shift. Put differently, we frame our conceptualization by asking the counterfactual question: ‘what would it mean for a system to not display resilience when subject to (exogenous) disturbances or ‘random’ shock? Two forms of resilience are often reported in the literature: ‘engineering resilience’ and ‘ecological resilience’ (Holling, 1996; Adger, 2000). According to notions of engineering resilience, globally stable dynamic systems typically have a unique equilibrium, such that out-of-equilibrium adjustment mechanisms will drive the system back towards that equilibrium in response to an exogenous shock. In contrast, ecological resilience refers to systems that are only locally stable and have multiple basins of attraction, or multiple equilibria, which raises the possibility that local economies may move between basins of attraction as a result of exogenous shocks. This conceptualization of ‘ecological resilience’ is consistent with a more general theory of discontinuities that encompasses models of hysteresis and structural stability, and is beyond the scope of this current paper (see Elster, 1976; Casti, 1979; Amable et al., 1995; Adger, 2000; Barkley Rosser, 2000; Gocke, 2002). Framing the question of resilience in terms of the response to exogenous shocks is consistent with recent advances in dynamic econometrics, which emphasises the ways in which persistent nonstationary processes and structural breaks can be employed to understand out-of-equilibrium dynamics (Katzner, 1993; Setterfield, 1988, 2010; Fingleton et al., 2012). In particular, by using econometric models it is possible to take advantage of the distinction between stochastic and deterministic dynamics when thinking about the notion of rural economic resilience. If a stochastic process is non-stationary, then a system can display persistent out-ofequilibrium dynamics, or path dependence, with no tendency to return to equilibrium after exogenous shock. Under our definition, this would be a system that is non-resilient. In other words, once a system is disrupted from an equilibrium growth path there is no tendency to return to the original steady state growth path. Rather, external shocks to such a (non-stationary) dynamic system have persistence effects on the subsequent trajectory of that system. In contrast, systems that can be characterized in terms of deterministic trends tend to return rapidly to their long run steady state following an exogenous shock and would be defined as resilient. However, the resilience of systems that are characterized in terms of deterministic steady states does not preclude either the existence of multiple equilibria or evolutionary transitions between local basins of attraction. On the contrary, exogenous shocks may ‘de-lock’ a system from one steady state growth path to another resulting in structural discontinuities and, hence, permanent shifts in the long run equilibrium of such a system. Substantively, structural breaks would be indicative of a regime shift in a rural economic system. For example, a local economy may switch from a low steady state growth path to a high steady state growth path as a result of a change in the institutional or regulatory environment, with no long run tendency to return to the original growth path after the regime shift. Accordingly, the local economy is not resilient to regime shifts. Thus, in this context it is important not to conflate a lack of resilience with economic ‘weakness’. Indeed, the shift to a new stable developmental trajectory in response to a shock can be viewed as a positive outcome where it results in stronger economic performance. 3.2. Modelling rural economic resilience: path dependence, hysteresis, and structural change In the context of this paper, we consider the nature and degree of rural economic resilience in terms of the dynamics of the

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

4

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

unemployment rate (U), employment growth (E) and annual personal income (W) in each locality. Consistent with the existing literature on economic resilience, these dynamics are modelled as a univariate stochastic data generation process, with the possibility of both path dependence and structural breaks in response to exogenous disturbance or ‘shocks’ (Tong, 1990; Hendry and Nielsen, 2007). To capture the possibility that a local economy evolves in historical rather than logical time, path dependent dynamics and structural breaks are considered in terms of a data generation process (yt ¼ lnYt) evolving over time in the sense that the moments of the probability model representing that process are nonconstant (Hendry, 2009; Plummer and Tonts, 2013a, 2013b). In the context of this paper, the process under consideration yt could represent the dynamics of unemployment, employment growth, or annual personal incomes. Specifically, evolutionary dynamics are postulated such that the current value of the stochastic process, yt, depends on the past history of the data generation process, t¼N,.t1, and a series of random shocks, 3 t, drawn from a normal probability distribution. Here, both the function mapping the past into the present (fj) and the variance ðs2t Þ of the random shocks are, at least potentially, time varying parameters and, hence, historically contingent data generation processes:

yt ¼

Xj¼t1   f j yj þ 3 t j¼N

3t

  ¼ N 0; s2t

(1)

This general model specification represents a formalization of the resilience framework in so far as it permits both transitional, out-of-equilibrium, dynamics with respect to a pre-determined equilibrium (ft¼f ct˛T) and permanent ‘shifts’ in equilibrium configurations (fjsfk, dksj). Pragmatically, in this paper we limit our model specification to the simplest scenario that is capable of sustaining evolutionary dynamics through both path dependent stochastic dynamics and deterministic structural discontinuities, or structural breaks. Specifically, we consider a second order linear stochastic process in which bo captures the deterministic components (constant, trend, and indicators) and b1, b2 are the out of equilibrium adjustment parameters.

yt ¼ b0 þ b1 yt1 þ b2 yt2 þ 3 t

b0 ¼ a1 þ a2 T þ a3 D1t þ a4 D1t T þ a5 D2t þ a6 D2t T

(4)

where D1t ¼ 1 for t equals 1989e1992, zero otherwise. Similarly, D2t ¼ 1 for t equals 1993e2011, zero otherwise. As a corollary, the pre-1989 parameters a1, a2 represent the ‘baseline’ deterministic components against which the resilience of a local economy can be evaluated in response to the regime shift that occurred between 1989 and 1992, a3,a4, and the post-1992 regime shift, a5,a6 (Chatterjee and Hadi, 2006). If the regime shift parameters are statistically significant, then this indicates that there was a shift in the data generation process such that the regime shift resulted in a step change (a3) and/or trend change (a4) in long-run equilibrium. Similarly, if the post-regime deterministic parameters are statistically significant then this indicates that the locality is not resilient in the sense that it has not returned to its long-run steady state growth path after the 1989e1992 regime shift. Put differently, at equilibrium, when the local economic process under consideration is on its long run steady state growth path, then

(2)

where a series of disturbances, or random shocks, to the system, 3 t, are assumed to be normally distributed with mean zero E(3 t) ¼ 0, constant variance Eð3 2t Þ ¼ s2 , and serial independence E(3 i3 j) ¼ 0, for isj. Finally, the deterministic components can be decomposed into constant (a1), trend (T), and indicator (It) components, which are intended to capture shifts due to structural discontinuities in the data generation process:

b0 ¼ a1 þ a2 T þ d1 It

that the regime shift occurred over the period 1989e1992, being the result of a confluence of events over this period, impacting on the regulatory environment. The period between 1989 and 1992 was one in which the deregulation of the agricultural sector hit hard, following the floating of the currency in 1983, deregulation of the banking sector in 1984, the 1987 stock market crash, and the ‘recession we had to have’, brought on by stringent monetary policy that saw farm overdraft rates reach 25% (Smailes, 1996). It also saw the withdrawal of the wool reserve price scheme in 1991 that left many woolgrowers in extreme difficulty (Smailes, 1997). The period 1991/92 also saw a global economic downturn. Based upon the timing and duration of the regime shift, we partition the parameter model represented in equation (3) into three periods: pre-1989 regime shift, the period of the regime shift, 1989e1992, and post-1992. This allows us to consider both the initial resistance of localities to this regime shift and the subsequent post-1992 recovery stage. Formally,

(3)

where It s 0 for t  T0, zero otherwise, and T0 is the timing of a structural break, a1,a2 define the long run steady state equilibrium, and d1 measures change in the equilibrium level and/or growth rate resulting from a structural break. Typically, structural breaks are modelled as discontinuities in either the mean (step shifts) or trend (trend shifts) or both (Perron, 1994; Baddeley et al., 1998). In the absence of prior knowledge about the timing of regime shifts, structural breaks can be determined endogenously by testing for parameter constancy using recursive, rolling or sequential techniques (Banerjee et al, 1993; Charemza and Deadman, 1997). However, in the context of this paper, we are able to identify the timing of potential structural breaks based upon our prior theoretical and local knowledge. Specifically, rather than assuming that the structural discontinuity occurred in a single year, we assume

  a þ a2 T þ a3 D1t þ a4 D1t T þ a5 D2t þ a6 D2t T E y*t ¼ 1 1  b1  b2

(5)

From which it follows that the locality displays economic resilience if the locality is resistant to the shock resulting from the regime shift (a3 ¼ a4 ¼ 0) and, subsequently, returns to its original long run steady state growth path (a5 ¼ a6 ¼ 0). Out of equilibrium, the stability or resilience of the process subject to exogenous disturbance, or shocks, depends on the magnitude of the adjustment parameters, b1 and b2. Assuming that there are no shifts in the deterministic components of the model, then if b2 ¼ 0 there is no hysteresis in the sense that the equilibrium defined in equation (5) does not depend on the adjustment process described by equation (1). Conversely, if b2 s 0 then the process displays hysteresis and, hence, the locality is not resilient to exogenous ‘shocks’. As is well known, other things being equal, if b1 ¼ 1 then the data generation process is non-stationary and, hence, path dependent (Hendry, 1995). That is, it is a unit root process displaying persistence, with the current value of yt depending on the cumulative random shocks to the system (Harvey, 1993). Conversely, if b1  1 then, following an exogenous shock, the system will return to its long run equilibrium growth path. As a corollary, the data generation process displays resilience.

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

3.3. Empirical modelling methodology: “stylized facts”, model specification and model selection In order to gain a deeper understanding of the dynamics of the localities we engage in some simple exploratory data analysis prior to testing for path dependence and rural economic resilience. Essentially, this entails a descriptive summary of the distributional properties of unemployment rates, employment growth, and annual personal incomes across the local economies with the aim of establishing some overall patterns, or “stylized facts”, that characterize the local economies. Whilst this can provide a thumbnail sketch of what has been happening over the past three decades, “stylized facts” are no substitute for careful empirical model specification and hypothesis testing. Indeed, it is often the case that simple empirical regularities can mask more than they reveal: the age-old problem of spurious correlation. Accordingly, we move beyond casual empiricism, bringing the data to bear on our operationalization of the concepts of path dependence and rural economic resilience by fitting the empirical model specification of equation (2) and equation (4) to the unemployment, employment and wage time series for each of the four localities. The model specification defined in equation (2) and the submodel defined in equation (4) allow empirical testing of the nature and degree of rural economic resilience, at least in terms of how these concepts are defined in this paper. This empirical model specification can be easily estimated using conventional ordinary least squares (OLS) techniques, with the parameter estimates being the best linear unbiased estimates conditional on the assumptions about the underlying data generation process being met. For the OLS model these assumptions are readily tested using a battery of conventional misspecification tests for linearity, normality, constant variance (homoscedasticity), and series independence (temporal autocorrelation) (McAleer, 1994). If the model passed the suite of misspecification tests then it is congruent with the data. To evaluate the empirical significance of both path dependence and rural economic resilience for each of the local economies we employ a general-to-specific modelling methodology, which has been found to be an efficient strategy in designing a theory consistent and parsimonious model specification that is congruent with the data (Plummer and Taylor, 2001a, 2001b). Recently, Hendry and his colleagues have automated a general-to-specific strategy using Autometrics and PcGETS (Doornik, 2009; Castle and Hendry, 2011). Essentially, Autometrics starts with a general unrestricted model (GUM) and acts like a large sieve, removing nonsignificant variables, searching multiple paths and testing for congruence at each stage in the model reduction process. It has been shown to outperform alternative automated model selection algorithms, such as stepwise regression, in the sense that it is highly successful in detecting the correct data generation process in Monte Carlo experiments. In this paper we employ Autometrics to derive a theory consistent and parsimonious model specification that is a reduction of the general model defined in equation (2) and equation (3). The ‘final’ reduced model specification tells us which parameters are statistically significant and, hence, whether our local economies display either path dependence or resilience, or both. 4. The case study localities This analysis employs the model specification and model selection strategies outlined above in the context of a comparative case study of rural local areas. Specifically, we draw on four rural local government areas in Western Australia: Beverley, Plantagenet, Toodyay and York (see Fig. 1). These localities were selected for their potential to provide quite different insights into the nature of

5

evolutionary processes and resilience. Two of the localities, Toodyay and York, were historically dominated by mixed crop and livestock production, with wheat and wool the two main commodities. While their populations were declining during the late 1960s, York began to experience a population turnaround in the 1970s, with Toodyay following suit in the 1980s. These towns were some of the early beneficiaries of the counterurbanisation movement that affected selected parts of rural Australia during the 1970s and 1980s (Hugo and Smailes, 1985) largely on the basis of their amenity landscapes (Argent et al., 2011) and relative accessibility, being located less than 100 km from the Perth metropolitan area. In contrast, the other two localities, Plantagenet and Beverley, are located 130 km and 350 km from Perth respectively, and have tended to remain focused on traditional agricultural enterprises, mainly in the wheat, sheep and to a lesser degree cattle and (in the case of Plantagenet) viticulture sectors. Only in the 2000s did these localities begin to experience lifestyle migration, albeit on a much more limited basis than in Toodyay and York. Table 1 provides an overview of the broad contours of population change in the case study localities between 1971 and 2011. This shows the rapid increase in population in Toodyay and York during the 1980s and 1990s as a result of a strong inflow of lifestyle migrants, with many of these people purchasing hobby farms that had been created from the subdivision of larger holdings. In addition, both towns successfully capitalized on their natural amenity and heritage architecture, with tourism and recreation becoming important parts of the economy (and no doubt helping to increase the number of eventual lifestyle migrants). In contrast to York and Toodyay, the localities of Beverley and Plantagenet experienced steady decline during the 1970s and 1980s, with both towns only beginning to arrest the decline in the 1990s. However, even this turnaround was modest with Beverley only recording a 0.2 per cent annual growth between 1991 and 2001, and Plantagenet 0.9 per cent over the same period. The growth over the next decade was also modest, although remains a distinct contrast to the decline of the 1970s and 1980s. In both Plantagenet and Beverley a modest degree of lifestyle migration helped to arrest the decline and, in the case of Plantagenet, the expansion of a viticulture industry that counteracted the contraction linked to the mixed crop and livestock industry. Table 2 provides a snapshot of recent changes in the local employment structure for the period 2001e2011. While ideally a longer time series would be provided, changes to the Australian and New Zealand Standard Industrial Classification over the period 1971e2011 make longitudinal analysis impractical. However, the shorter timeframe is indicative of some of the longer run changes. Beverley and Plantagenet recorded the greatest increases in the total labour force between 2001 and 2011, although recorded sharp declines in agriculture, forestry and fisheries in line with restructuring in the agricultural sector, and in particular processes of farm amalgamation and expansion. Typically it was service sectors that recorded growth in these localities, notably in education and training, and health care and social assistance. Manufacturing employment increased in Plantagenet on the basis of an expanding viticulture industry. As might be expected given their population growth, the labour force also expanded in Toodyay and York. However, as with the other case study localities agriculture, forestry and fisheries declined, while increases were recorded in retail trade, construction and a number of service sectors. The overall picture that emerges is of local economies that have transitioned from a dependence on agriculture and related services to more complex, multifunctional spaces that incorporate not only farming but other more service oriented sectors. In large part, this appears to be the outcome of lifestyle migration driving population and employment growth, and therefore contributing to the

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

6

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

Fig. 1. Case study local areas.

emergence of new rural economies in the case study localities. Yet, this cursory overview of population and employment change is suggestive of a smooth transition, and in reality this belies the considerable upheaval and disturbance that these localities have endured over the past three decades or so. To better understand these dynamics, and in particular the extent to which concepts of path dependence and resilience are important, we turn to a more detailed examination of three key indicators of economic performance: unemployment, employment growth and annual personal income.

Table 1 Populations of case study localities, 1971e2011.

Beverley Plantagenet Toodyay York

1971

1981

1991

2001

2011

1628 4296 1725 2044

1520 4069 1396 2108

1436 3964 2461 2500

1464 4329 3750 3005

1506 4684 4256 3283

Source: Australian Bureau of Statistics, various issues.

5. An overview of unemployment rate dynamics, employment growth and wage inflation Fig. 2 shows the time series of local unemployment rates, number of persons employed and real annual incomes for Beverly, Plantagenet, Toodyay, and York for the period 1981e2011. These data are drawn from two sources. The employment growth (E) and unemployment (U) data are drawn from the quarterly publication Small Area Labour Markets, produced by the current Department of Education Employment and Workplace Relations. These data are available for the period 1984e2011. The real annual personal income (W) data are from the Australian Taxation Office and are for the period 1980e2008. The vertical lines on the figures indicate the regime shift between 1989 and 1992, described earlier. The first observation is that there appears to be a high degree of synchronicity across localities for both unemployment rate dynamics and real wage dynamics, with less synchronicity in employment. There appears to be a marked increase in the unemployment rates and a corresponding fall in real incomes across all localities during the 1989e1992 regime shift. This, of course, is the period characterized

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

by economic recession and the major upheaval in the agricultural sector, and appears to have played out across all of the case study areas. Subsequently, unemployment rates stabilized up until around 2000, after which they fell. This coincides with a period of strong economic growth in Australia, and in particular the ‘resources boom’ from about 2004 onwards. While none of these localities were directly affected by mining, most of Western Australia saw a sharp decrease in the unemployment rate over this period as a result of the more general expansion of the economy. In contrast to falling unemployment, real incomes appear to have risen consistently across all localities throughout the post-regime shift period. Finally, the dynamics of employment show differences between localities: Plantagenet and Beverley appear to have experienced consistently stable employment levels over the past thirty years, which coheres with their broader population trends and is a reflection of their dependence on a ‘mature’ agricultural economy. York and Toodyay, on the other hand, have experienced rising numbers of persons employed, which is consistent with their lifestyle-led and counterurbanisation-led population growth. Moving beyond Fig. 2, and temporarily setting aside the temporal properties of the data generation process, each measure of local economic performance for each locality is summarized in Table 3. First, both average unemployment and volatility, as measured by the standard deviation, was highest in Toodyay, followed by York, Beverley, and Plantagenet. Overall, the rate of change in the unemployment rate was negative across all localities, with York experiencing the most rapid fall in its unemployment rate, followed by Toodyay, Plantagenet and York. For each locality the dispersion of reductions in the unemployment rate over this period is similar, although the variability around the overall decline in unemployment rates across this period is marginally higher in

Table 2 Employment by industry sector, 2001e2011. Beverley

Plantagenet

2001 2011 2001 Agriculture, forestry 243 and fishing Mining 5 Manufacturing 26 Utilities 0 Construction 17 Wholesale trade 18 Retail trade 32 Accommodation and 20 food services Transport, postal and 19 warehousing Information media 8 and teleco’s Financial and 11 insurance services Rental, hiring and real 0 estate services Professional, scientific 13 and tech services Administrative and 3 support services Public administration 32 and safety Education and training 37 Health care and 56 social assistance Arts and recreation 6 services Other services 14 Inadequately described/ 6 Not stated Total 566

2011

Toodyay

York

2001 2011 2001 2011

181

709

521

177

133

285

247

18 32 3 35 8 39 16

8 180 12 95 61 119 79

30 219 8 131 56 184 78

28 129 15 144 67 113 78

106 108 17 186 47 164 93

20 58 16 70 35 117 73

53 47 27 105 53 160 84

30

45

95

60

108

36

63

0

8

6

38

11

14

3

7

13

27

16

24

11

18

3

19

27

32

25

20

19

15

47

46

52

65

51

68

11

31

35

42

49

29

17

42

92

137

120

148

60

78

48 69

134 96

153 174

105 127

138 197

100 99

102 160

0

9

12

7

16

9

12

15 9

54 45

65 44

62 52

87 49

39 53

40 37

581

1856

2048

1464 1771 1195 1393

7

Toodyay and York than it was in Plantagenet and Beverley. This is not an uncommon pattern, with ‘lifestyle’ localities often being linked to welfare migration, while in those places dependent on agriculture, outmigration amongst unemployed is quite typical (Hugo and Bell, 1998; Argent et al., 2007; Argent and Walmsley, 2008). With the exception of the unemployment rate in Beverley, there is no evidence of deviations from a normal distribution for either the unemployment rate or rate of change in the unemployment rate. Second, on average, the employment level was highest in Plantagenet, followed by Toodyay, York, and Beverley. In contrast to the unemployment rate, the rank ordering of localities is different for the volatility of employment, with the highest being in Toodyay, followed by York, Plantagenet, and Beverley. That said, and consistent with the evidence on unemployment rates, those localities with the highest employment tended to experience the greatest variability in job creation. In terms of the rate of job creation as measured by the percentage change in employment levels, there are two distinct groupings with Toodyay and York experiencing higher growth rates than Plantagenet and Beverley. Here there is a distinct difference between the two localities most affected by amenity migration, Toodyay and York, and the two places with a strong dependence on agriculture and with few other industries. Over this period all localities experienced positive job creation (on average), with Toodyay experiencing a higher growth rate than the other localities. There is no evidence that either employment levels or job creation across the SLA deviates significantly from normality. Finally, both the average real wage and wage volatility for the 1981e2008 period was highest in Beverley, followed by Toodyay, York and Plantagenet. That is, the localities with the highest (lowest) average real wage rate experienced (on average) the highest volatility in their real wage rates. Overall, all localities experienced positive wage inflation over this period and we find an identical pattern of wage dynamics between the localities, with Beverley experiencing the highest wage inflation, followed by Toodyay, York, and Plantagenet. The pattern of wage inflation is similar to wage rate dynamics, with Beverley and Toodyay being the most volatile and Plantagenet and York the least volatile (although their ranking was reversed). However, there is evidence of non-normality in the rates of wage inflation for Toodyay and York, which warrant further consideration as part of the formal empirical modelling. Table 4 shows the simple correlations between localities for each measure of local economic performance. For the relationship between unemployment rates, the highest correlations are between York, Toodyay, and Beverley, with the lowest correlations being between Plantagenet and the other localities. Although the simple correlations between the unemployment rates in each SLA are statistically significant at the 5% level these results should be treated with caution and are likely to be spurious in the presence of non-stationary time series. Exploring the rate of change in unemployment, there remain significant correlations between York, Toodyay and Beverley but the correlation between Plantagenet and the other SLAs is no longer significant. Note that using the rates of change in variables effectively removes the non-stationarity in the data and, hence, the likelihood of correlations due to common trends. Overall, the results of the simple correlation analysis indicate that there is synchronicity between the unemployment dynamics of York, Toodyay, and Beverley, but not Plantagenet. Exploring the correlation between employment levels and job creation across localities, with the exception of the correlation between job creation in Beverley and Plantagenet, the pattern of simple correlations is statistically significant at the 5% level for both employment levels and rates of job creation. However, whilst the

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

8

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

Fig. 2. Local economic dynamics, 1980e2011.

correlation structure for employment levels is positive, the pattern of correlation is more complex once we control for possibility of common trends (i.e remove non-stationarity, see above). There are significant positive correlations between the rate of job creation between Beverley, Toodyay and York and between Plantagenet, Toodyay and York, whilst there is a negative correlation between Toodyay and York. Finally, over this period there is evidence of strong positive correlations between the wage rates of all localities, with all simple correlations being statistically significant at the 5% level. Whilst simple correlations for wage inflation are lower across the localities they are nonetheless statistically significant and positive, reflecting

Table 3 Descriptive statistics across localities.

Un Rate (U) Mean St. Dev Normality DU Mean St. Dev Normality Employment (E) Mean St. Dev Normality DE Mean St. Dev Normality Real Wage (W) Mean Standard Dev Normality DW Mean Standard Dev Normality

Beverley

Plantagenet

Toodyay

York

6.23 2.04 6.41*

6.13 1.66 4.37

7.16 2.62 0.48

6.51 2.37 0.57

0.002 0.294 0.744

0.011 0.306 0.179

0.037 0.315 0.097

0.043 0.315 4.632

712.4 83.82 2.89

2315.8 195.59 1.50

1437.6 539.29 4.07

1278 289.22 2.58

0.006 0.072 3.515

0.008 0.063 3.529

0.049 0.091 3.124

0.026 0.070 0.236

40,253 5384.4 0.095

35,644 4221.9 0.564

39,972 4975.4 2.619

38,563 4696.9 0.454

0.009 0.092 4.068

0.003 0.081 5.518

0.009 0.083 7.44*

0.006 0.075 8.82*

a degree of synchronicity across local labour markets. This synchronicity suggests that wage dynamics are in part driven by wider processes from ‘beyond the local’, and might include commodity prices and other wider macroeconomic forces. 6. Testing for rural economic resilience The stylized facts presented in the previous section suggest that the local economies in the case study areas were subject to considerable upheaval and transformation over a near threedecade period. Taking this analysis further, the application of econometric modelling offers a means of going beyond casual empiricism to provide a more nuanced set of insights into the processes of change. In this section we seek to better understand local economic dynamics and, in particular, resilience, by fitting equation (2) and equation (4) to the data for each locality and employing the automated general-to-specific model selection algorithm. Tables 5e7 summarize the final model specifications. Recall that the b1, b2 parameters capture the stochastic adjustments with respect to a long-run steady state, including non-stationary trends and hysteresis. The deterministic parameters a1, a2 determine the location of the steady state equilibrium prior to the regime shift, whereas the a3, a4 parameters determines the reaction of unemployment rates, employment and annual personal incomes to the 1989e1992 regime shift. Finally, the a5, a6 parameters determine the long run responsiveness of localities to the regime shift. If both the initial reaction to the regime shift and the subsequent response are not significant, then a locality is resilient to the regime shift, otherwise the locality is not resilient. Table 5 and Fig. 3 focus on unemployment rates. In this case, the model accounts for a statistically significant proportion of the variability in unemployment rates across all localities. Furthermore, each model is congruent with the data. Based upon the final model specification, although there are similarities in unemployment dynamics (as was evident in Fig. 2), rural economic resilience, as measured by unemployment, differs quite considerably across localities. Both Beverley and York had a pre-regime shift stationary state, which was disrupted by changes during the 1989e1992

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14 Table 4 Simple correlations across localities. Bev U Bev Plant Tood York

Table 6 Employment dynamics e final model specification. Plant

Tood

York

0.485 (0.01) 0.582 (0.00)

Tood York E Bev Plant Tood York

1 0.05 (0.81) 0.622 (0.00) 0.704 (0.00)

1

1 0.398 (0.04) 0.576 (0.00) 0.695 (0.00)

a3 a4 a5

1 0.655 (0.00)

1

1 1 0.974 (0.00)

1

De Bev Plant Tood York W Bev Plant Tood York

1 0.272 (0.17) 0.545 (0.00) 0.825 (0.00) 1 0.928 (0.00) 0.934 (0.00) 0.925 (0.00)

1 0.661 (0.00) 0.539 (0.00)

1 0.681 (0.00)

1

1 0.880 (0.00) 0.827 (0.00)

1 0.925 (0.00)

1

Dw Bev Plant Tood York

1 0.868 (0.00) 0.787 (0.00) 0.841 (0.00)

1 0.707 (0.00) 0.726 (0.00)

1 0.782 (0.00)

1

Table 5 Unemployment rate dynamics e final model specification. Beverley

Plantagenet

b1 b2 a1 a2 a3 a4 a5 a6 R2 F(prob)

Toodyay

York

2.656** (1.043) 0.284** (0.090) 0.831** (0.243) 0.035** (0.009) 0.575 (0.001)

2.003** (0.128) 0.014** (0.006) 3.07** (0.977) 0.302** (0.092)

0.465 (0.003)

1.373** (0.414) 0.026** (0.008)

0.675 (0.000)

R2 F(prob)

Not Reported

7.684** (0.019)

6.339** (0.046) 0.048** (0.002)

6.915** (0.212)

0.161** (0.054) 0.011** (0.002) 0.606 (0.000)

Table 7 Wage rate dynamics e final model specification. Plantagenet

Toodyay

York

10.39** (0.022) 0.026** (0.002) 1.737** (0.256) 0.179** (0.024) 0.296** (0.072)

10.324** (0.022) 0.024** (0.002) 1.731** (0.209) 0.176** (0.019) 0.328** (0.040)

2.242 (0.158) 12.93** (1.658) 0.029** (0.004) 1.233** (0.256) 0.132** (0.024) 0.344** (0.065)

10.34** (0.019) 0.022** (0.002) 1.244** (0.266) 0.129** (0.017) 0.188** (0.035)

0.874 (0.000)

0.888 (0.000)

0.856 (0.000)

0.923 (0.000)

b1 b2 1.904** (0.132)

3.418** (1.093) 0.338** (0.103) 1.012** (0.103) 0.055** (0.009) 0.715 (0.000)

0.946 (0.000)

0.422** (0.057) 0.033** (0.002) 0.938 (0.000)

period, resulting in increasing trends in unemployment. This change in steady state dynamics was stronger in terms of the size of the trend parameter in York. Post-regime shift, both localities experience a reversal of fortunes, experiencing both step and trend shifts associated with a declining long run steady state unemployment rate in both localities. As a corollary, neither local economy is resilient in the sense of resisting the initial regime shift or returning to the pre-1989 unemployment rate stationary state. In other words, the economic and policy shock of the 1989e1992 period contributed to considerable economic disruption in York and Beverley as measured by unemployment. In both localities, the economies were tied closely to the fortunes of agriculture, which had experienced significant adjustment pressures associated with the cost-price squeeze, rising interest rates, and deregulation. Moreover, the collapse of the reserve price scheme in the wool industry in 1991 amplified the difficulties facing farmers, with both localities highly exposed to this industry. We would conjecture that the direct loss of agricultural employment, together with losses in allied industries (e.g. shearing, fencing contractors, transport contractors and agricultural suppliers) were important contributors to unemployment. In York, an additional set of impacts was felt as a result of broader economic upheaval in the Australian economy in the late 1980s and early 1990s. The collapse of the stock market in 1987, rising interest rates, and economic recession led to a period of difficulty for sectors related to tourism and amenity migration. While on the one hand we might consider these economies to lack resilience in that they were not able to withstand the effects the 1989e1992 shock, on the other they could be regarded as

Beverley

0.519** (0.158) 1.708** (0.127)

York

1

a6

0.671 (0.00) 0.693 (0.00)

Toodyay

a2 0.833 (0.00)

1 0.079 (0.69) 0.165 (0.412)

Plantagenet

1.001** (0.002)

b2 a1

1

Du Bev Plant

Beverley

b1 1 0.378 (0.04) 0.799 (0.00) 0.748 (0.00)

9

a1 a2 a3 a4 a5 a6

R2 F(prob)

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

10

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

resilient in the sense that post 1992 their performance improved and they shifted to a lower long run equilibrium decline in unemployment rates. In both locations, rural industries recovered somewhat after 1992, with many farmers transitioning their enterprises from a focus on mixed wool and cereals to a more intensive focus on crop production, particularly wheat, barley, canola and legumes and fat lamb production. While the outcome was an improvement in farm performance with some flow on effects for communities, it should be noted that the new industries were also less labour intensive. In both areas, part of the adjustment also involved out-migration of those that had lost jobs in the agricultural sector, with a strong out-migration in favour of coastal areas (Davies and Tonts, 2007). This contributed to a lowering of the unemployment rate. In York, an additional contributor to the falling unemployment rate was the recovery and expansion of sectors related to tourism and amenity migration, which were able to absorb some of the excess capacity in the local labour market (Davies and Tonts, 2007). In the case of Toodyay, there is no evidence of a qualitative change in unemployment dynamics either during or after the 1989e1992 regime shift. In this sense, Toodyay demonstrates resilience in that there was no discernible impact associated with the wider upheaval and restructuring of the wider economy. This tends to suggest that local factors were particularly important in enabling Toodyay to withstand broader socio-economic processes, and to continue to improve its overall performance across the study period. The reasons for this resilience require further investigation, but are likely to be linked to a relatively high degree of economic diversity, proximity and accessibility to the Perth metropolitan area, and a relatively small agricultural sector relative to the other localities (Table 2). Similar to Toodyay, pre-1989 Plantagenet was already experiencing a declining unemployment rate trend to which it returned post regime shift following the disruption of the 1989e1992 period. Accordingly, whilst Plantagenet did not display initial resilience to the regime shift, post 1992 it is considered to display rural economic resilience in that the underlying dynamic was one of falling unemployment. It is also apparent that the unemployment dynamics are significantly qualitatively different from the other localities. There is no evidence of deterministic shifts in the long run unemployment rate pre-, during or post- regime shift. Rather, there is evidence of out-of-equilibrium adjustments around a long run deterministic decline in the unemployment rate. This is indicative of resilience, although there is the possibility that the out of equilibrium adjustment process is non-stationary and, hence, displays persistence, with no tendency to return to the steady state unemployment rate. In other words, Plantagenet’s economy demonstrates a development trajectory that has consistently reduced the number of people who are unemployed, and this overall trend continued despite the 1989e1992 regime shift. In large part, this is likely associated with the increasing level of employment diversity within the economy over time. During the past few decades, Plantagenet’s agricultural economy has diversified beyond traditional mixed crop and livestock farming to incorporate a vibrant viticulture and associated winemaking sectors (Tonts et al., 2008). The area has also experienced the emergence of some lifestyle-led migration and tourism linked to the nearby Stirling Range National Park, and has gradually become an increasingly important subregional service centre for the surrounding region. While unemployment rates are an important indicator of economic performance in that they point to the relative ‘efficiency’ of a local labour market, equally important is the ability of an economy to create jobs. Table 6 and Fig. 4 show that employment dynamics across the localities are qualitatively different from those for unemployment rates. There is no evidence of a significant reaction

during the regime shift period, with employment trends continuing on their pre 1989 paths. In this sense, all localities are resilient in that they consistently generate jobs. In addition, for Toodyay and Beverley there is no evidence of structural shifts post 1992. Throughout the 1980e2010 period, employment in Toodyay fluctuated around an increasing long run steady state equilibrium and, hence, displayed economic resilience. Accordingly, despite the regime shift in 1989e1992, Toodyay’s capacity to generate jobs did not change considerably over the entire study period. In contrast to Toodyay, the employment dynamics in Beverley were driven by out-of-equilibrium adjustments, although these adjustments appear to be non-stationary, with no tendency to return to equilibrium following random shocks. Accordingly, employment dynamics in Beverley are characterized by persistence rather than resilience. Put differently, the number of jobs being created in Beverley is relatively low and this has not changed to any significant degree over the study period. In many respects, Beverley demonstrates a form of path dependence in so far as the 1989e 1992 regime shift did not contribute to major changes in the rate of job creation. In large part, this is linked to Beverley’s continuing dependence on agriculture, and while this sector did recover post1992 the structure of the industry was one that became increasingly capital intensive with lower labour requirements. Finally, employment dynamics in Plantagenet and York are qualitatively similar. Both pre 1989 and during the 1989e1992 regime shift employment levels were stationary, whereas post1992 both localities moved onto increasing long-run steady state growth paths, with the rate of increase being higher in York than Plantagenet. In the case of York, we would suggest that growth increased largely as a result of the displacement and subdivision of traditional farms for hobby farm development during the 1990s. Indeed, for some farmers exiting the industry by subdividing properties was a more attractive financial proposition than selling farms as going concerns. In the 1990s there were considerable efforts to expand the town’s tourism sector, with the combination of a major marketing campaign, new festivals and events, and new tourist investments contributing to an expansion of employment in the sector. In other words, Yorks’ post-shock employment performance is likely to be strongly influenced by commuting. In Plantagenet, an increasingly diversified economy as a result of the expansion of viticulture and winemaking, together with some lifestyle migration and a growing sub-regional service role, were important in contributing to job creation. While unemployment rates and employment dynamics provide important insights into the performance of local economies they have limitations. For example, employment dynamics do not readily account for the substitution of labour for capital, which has been particularly important in many rural industries where labour saving technologies have been employed. The employment data also do not fully account for economies where employment might remain stable, but incomes fall as a result of falling commodity prices, rising farm input costs, or both. Wage data also provide a different set of insights into rural wellbeing, by providing insights into how the incomes of individuals are affected by economic shocks. Table 7 and Fig. 5 show the actual and fitted time series paths followed by wage rates for each locality, which are simultaneously simpler and more complex to interpret. All localities display qualitatively similar dynamics and are not resilient to the changes that took place during the regime shift between 1989 and 1992. They all experienced an initial negative response over the regime shift period, moving from a rising to a falling steady state growth path. In addition, post 1992 all localities moved to an increasing long run steady state growth path. Although there was not a significant difference between the long run equilibrium trend in increasing

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

11

Fig. 3. Unemployment dynamics e final model specification.

incomes pre and post regime shift, post 1992 all localities experienced a step decrease in the long run equilibrium growth rate, with the reactions being greatest in Plantagenet and Toodyay. It is also clear that the recovery was extremely slow across the board. Indeed, it took until the early 2000s for York to return to pre regime shift levels, with Beverley and Toodyay attaining per wage rate levels in the mid-2000s, and Plantagenet only returning to per1989 levels at the end of the same decade. It is apparent from the analysis that the interaction between unemployment, employment growth, and incomes are complex,

with falling unemployment and job creation not necessarily linked in any immediate way with increasing incomes. This is likely to be because of the segmented nature of these local economies as they have experienced transition. In a number of these locations, the decline of agriculture and related sectors was accompanied by a rise in new sectors, including those related to amenity-led migration and tourism (e.g. the service sectors). It is likely that these new sectors were able to drive some employment growth and reduce unemployment, but that they were not able to replace or match the income levels earned prior to the regime shift. Moreover, despite a

Fig. 4. Employment dynamics e final model specification.

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

12

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

Fig. 5. Wage dynamics e final model specification.

recovery in the agricultural sector post regime shift it was not able to generate the income levels experienced prior to 1989. By the mid 2000s, however, the economic transition had gathered enough momentum to return to and then surpass the income levels prior to the regime shift. The next steps in understanding the ‘on the ground’ dynamics of the emergence of these new rural economies is detailed fieldwork to offer a closer assessment of the economic, social, cultural and political dynamics that contribute to resilience and adaptation. 7. Discussion and conclusion The analysis presented here begins to shed light on the potential value of an evolutionary approach to understanding local rural economic performance, and in particular how concepts such as path dependence and resilience might offer insights into how economies respond to processes operating at multiple spatial and temporal scales. For all four localities the 1980s and early 1990s were a period of considerable upheaval associated with significant challenges in the agricultural economy and major policy reforms. While one might argue that the 1980s were a decade long ‘slow burn’ for agricultural regions, the most intense shock was arguably experienced in the period 1989e1992. Indeed, this was one of the most traumatic periods of ‘rural adjustment’ since 1945 as many farmers exhorted by policy-makers to ‘get big or get out’ did both: borrowed large sums of capital on ‘easy terms’ to expand their operations, buying out the neighbouring farm, only to fall into irretrievably bad debt situations as interest rates escalated simultaneous with the collapse of the wool market. In the space of a few years one of the mainstays of many Australian agricultural regions e wool growing e became a niche industry. Of course, this not only affected traditional wool growers, but reverberated throughout local labour markets as farm hands, shearers, rouseabouts, and wool classers and so on lost regular and well-paying jobs, inducing severe negative multiplier effects across local rural economies. The evidence presented in this paper shows that all four case study local areas were deeply affected by these processes. However, it is also clear that post the regime shift all commenced a process of economic recovery. For example, all economies experienced a shift

to lower unemployment after the 1989e1992 shock. Similarly, employment growth is reasonably consistent post the regime shift, and incomes also recover across the board. In essence, it would appear that there is a set of broad based structural drivers at play, shaping rural economies. In the period post 1992, for example, a general economic recovery during the 1990s and early 2000s, the resources boom of the mid 2000s, and a relatively stable political environment were undoubtedly important in driving recovery. However, we would also argue that growth was constrained by ongoing adjustment in the deregulated wool industry, which never again reached the lofty, state-guaranteed, prices of the late 1980s and 1990s, triggering a sustained decline in the number of woolgrowers. It is also clear that these general trends are not able to fully account for the experiences of the different locations. In all four local areas the impact of the shock, and the time taken to recover, varies considerably. For example, prior to the 1989e1992 shock the stationary state for unemployment was disrupted, and led to increasing rates of unemployment post regime shift. In contrast, Plantagenet returned to its pre-1989 equilibrium unemployment rate post the regime shift. Similarly, there are notable differences between the localities in terms of real wage dynamics. While all four locations were slow to return to their pre-1989 levels, it took York until the early 2000s, Beverley and Toodyay the mid 2000s, and Plantagenet until nearly the end of the 2000s. All of this points to the importance of multiscalar processes that manifest themselves locally as distinctive development pathways. It is also clear that particular local conditions and processes are critical in moulding development. These include the amenity landscapes and the availability of former agricultural holdings in accessible locations like York and Toodyay contributing to counter-urbanisation led economic recovery, while relative remoteness and an ongoing dependence on agricultural commodities shaped performance in Plantagenet. Other factors that we would suggest are important include regional development policy, land use regulations and planning, local leadership and the structure of the economic base. This paper also has broader implications for how we think about rural economic development and transformation within the context of the theoretical architecture offered by evolutionary

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

economic geography. Drawing on this body of work, and in particular path dependence and resilience, the analysis was able to detect evidence of both path dependence and structural change, or regime shifts, across all localities. This suggests that rural localities may become locked into developmental trajectories from which it is difficult to de-lock, for good or ill. Indeed, for the local areas in question, being locked into a system of protected but simultaneously highly locally competitive agricultural industries generated enormous benefits. Yet, by the 1980s these rural economies had become problematic. Extending the evolutionary metaphor, the extent to which these place have been able to adapt to a changing regime of accumulation likely depends on the ways in which the changing nature of competition and selective pressures have played out over space. Increasingly, these multiscalar, inter and intra-sectoral, technological, and institutional selection pressures are working against the established agricultural industries in these locations. For example, in the case of the localities in question a range of competitive pressures are evident, including declining terms of trade, increased international economic exposure in the wake of deregulation, increasing competition from other commodities (e.g. cotton, artificial fibres) and competition from other regions around the world. Based on the analysis here, it is evident that local areas that have become locked into particular development trajectories may require relatively large shock to jump off the rails of path dependence. In the context of our case study rural economies, developmental trajectories are dominated by the transformational response of wage, employment and unemployment to the structural change of the early 1990s. While the capacity of any model to anticipate major regime shifts is questionable, it is apparent that there is capacity for robust local models to act as means of understanding recovery and resilience. We would argue that the next step is to advance these models in such a way that they provide policy tools in the face of effectively ex ante unpredictable changes in the socio-economic environment within which they are situated. Acknowledgements The authors thank the referees for their insightful comments on this paper. The work on which this paper is based was funded from Australian Research Council Discovery Grant DP0770460. References Adger, W.N., 2000. Social and ecological resilience: are they related? Prog. Hum. Geogr. 24, 347e364. Amable, B., Henry, J., Lordon, F., Topol, R., 1995. Hysteresis revisited: a methodological approach. In: Cross, R. (Ed.), The Natural Rate of Unemployment: Reflections on 25 Years of the Hypothesis. Cambridge University Press, Cambridge, pp. 153e179. Argent, N., 2011. Australian agriculture in the global economic mosaic. In: Tonts, M., Siddique, M.A. (Eds.), Globalisation, Agriculture and Development: Perspectives from the Asia-Pacific. Edward Elgar, Cheltenham, pp. 7e28. Argent, N., Walmsley, D.J., 2008. Rural youth migration trends in Australia. Geographical Research 46, 139e152. Argent, N., Smailes, P., Griffin, T., 2007. The amenity complex: towards a framework for analysing and predicting the emergence of a multifunctional countryside in Australia. Geographical Research 45, 217e232. Argent, N., Tonts, M., Jones, R., Holmes, J., 2011. Amenity-led migration in rural Australia: a new driver of local demographic and environmental change. In: Luck, G., Race, D., Black, R. (Eds.), Demographic Change in Australia’s Rural Landscapes. Springer, Dordrecht, pp. 23e44. Arthur, W.B., 1994. Increasing Returns and Path Dependence in the Economy. University of Michigan Press, Ann Arbor. Baddeley, M., Martin, R., Tyler, P., 1998. Transitory shock or structural shift? the impact of the early 1980’s recession on British regional unemployment. Appl. Econ. 30, 19e30. Banerjee, A., Doldo, J., Galbraith, J., Hendry, D., 1993. Co-integration, Errorcorrection, and the Econometric Analysis of Non-stationary Data. Oxford University Press, Oxford.

13

Barkley Rosser, J., 2000. From Catastrophe to Chaos: a General Theory of Economic Discontinuities. Kluwer Academic Publishers, London. Barnes, T.J., 1995. Logics of Dislocation: Models, Metaphors, and Meanings of Economic Space. The Guilford Press, New York. Casti, J., 1979. Connectivity, Complexity and Catastrophe in Large Scale Systems. Wiley & Sons, New York. Castle, J., Hendry, D., 2011. A tale of three cities: model selection in over-, exact, and under-specified equations. In: Kaldor, M., Vizard, P. (Eds.), Arguing About the World: the Work and Legacy of Meghnad Desai. Bloomsbury Academic, New York, pp. 75e98. Chatterjee, S., Hadi, A., 2006. Regression Analysis by Example. Wiley and Sons, New York. Charemza, W., Deadman, D., 1997. New Directions in Econometric Practice: General to Specific Modelling, Cointegration and Vector Autoregression. Edward Elgar, Cheltenham. Christopherson, S., Michie, J., Tyler, P., 2010. Regional resilience: theoretical and empirical perspectives. Cambridge J. Re. Econ. Soc. 3, 3e10. Davies, A., Tonts, M., 2007. Employment in the Wheatbelt: Regional Trends, Issues and Challenges, Geowest No. 33. Institute for Regional Development, The University of Western Australia, Crawley. David, P., 2001. Path dependence, its critics and the quest for ‘historical economics’. In: Garrouste, P., Ioannides, S. (Eds.), Evolution and Path Dependence in Economic Ideas. Edward Elgar, Cheltenham, pp. 15e40. Doornik, J., 2009. Autometrics. In: Castle, J., Shephard, N. (Eds.), The Methodology and Practice of Econometrics: a Festschrift in Honour of David F. Hendry. Oxford University Press, Oxford, pp. 88e122. Dopfer, K. (Ed.), 2005. The Evolutionary Foundations of Economics. Cambridge University Press, Cambridge. Elster, J., 1976. A note on hysteresis in the social sciences. Synthese 33, 371e391. Essletzbichler, J., Rigby, D., 2007. Exploring evolutionary economic geographies. J. Econ. Geogr. 7, 549e571. Fingleton, B., Garretsen, H., Martin, R., 2012. Recessionary shocks and regional employment: evidence on the resilience of U.K. regions. J. Reg. Sci. 1, 109e133. Foster, J., Metcalfe, S., 2012. Economic emergence: an evolutionary perspective. J. Econ. Behav. Organ. 82, 420e432. Franklin, A., Newton, J., McEntee, J., 2011. Moving beyond the alternative: sustainable communities, rural resilience and the mainstreaming of local food. Local Environ. 16, 771e788. Fujita, M., Mori, T., 2005. Frontiers of the new economic geography. Pap. Reg. Sci. 84, 377e405. Glover, J., 2012. Rural resilience through continued learning and innovation. Local Econ. 27, 355e372. Gocke, M., 2002. Various concepts of hysteresis applied to economics. J. Econ. Surv. 16, 167e188. Graugaard, J., 2012. A tool for building community resilience? A case study of the Lewes Pound. Local Environ. 17. http://dx.doi.org/10.1080/ 13549839.2012.660908. Harvey, A., 1993. Time Series Models. MIT Press, Cambridge. Harvey, D., 1982. Limits to Capital. Basil Blackwell, Oxford. Hassink, R., 2005. How to unlock regional economies from path dependency? from learning region to learning cluster. Eur. Plan. Stud. 13, 521e535. Hassink, R., 2010. Regional resilience: a promising concept to explain differences in regional economic adaptability. Cambridge Journal of Regions. Economy and Society 3, 45e58. Hendry, D., 2009. The methodology of empirical econometric modeling: applied econometrics through the looking glass. In: Mills, T., Patterson, K. (Eds.), Palgrave Handbook of Econometrics: Volume 2 Applied EconometricsPalgrave, London, pp. 4e67. Hendry, D., 1995. Dynamic Econometrics. Oxford University Press, Oxford. Hendry, D., Nielsen, B., 2007. Econometric Modelling: a Likelihood Approach. Princeton University Press, Princeton. Holling, C., 1996. Engineering resilience versus ecological resilience. In: Schultze, P. (Ed.), Engineering within Ecological Constraints. National Academy Press, Washington DC, pp. 51e66. Hudson, R., 2010. Resilient regions in an uncertain world: wishful thinking or a practical reality? Cambridge J. Reg. Econ. Soc. 3, 11e25. Hugo, G., Bell, M., 1998. The hypothesis of welfare-led migration to rural areas: the Australian case. In: Boyle, P., Halfacree, K. (Eds.), Migration into rural areas: theories and issues. Wiley, Chichester, pp. 107e133. Hugo, G., Smailes, P., 1985. Urbanerural migration in Australia: a process view of the turnaround. J. Rural Stud. 1, 10e11. Katzner, D., 1993. Some notes on the role of history and the definition of hysteresis and related concepts in economic analysis. J. Post-Keynesian Econ. 15, 323e345. Lawrence, G., 1987. Capitalism and The Countryside: the Rural Crisis in Australia. Pluto Press, Sydney. Lawrence, G., 1996. Contemporary agri-food restructuring: Australia and New Zealand. In: Burch, D., Rickson, R., Lawrence, G. (Eds.), Globalization and Agrifood Restructuring: Perspectives from the Australasia Region. Avebury, Aldershot, pp. 45e72. Mas, M., Hassink, R., 2013. An evolutionary perspective on tourism area development. Ann. Tour. Res. 41, 89e109. MacKinnon, D., 2012. Beyond strategic coupling: reassessing the firm-region nexus in global production networks. J. Econ. Geogr. 12, 227e245. Mackinnon, D., Driscoll Derickson, K., 2012. From resilience to resourcefulness: a critique of resilience policy and activism. Prog. Hum. Geogr. 37, 253e270.

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001

14

M. Tonts et al. / Journal of Rural Studies xxx (2014) 1e14

MacKinnon, D., Cumbers, A., Pike, A., Birch, K., McMaster, R., 2009. Evolution in economic geography: institutions, political economy and adaptation. Econ. Geogr. 85, 129e150. Marx, K., 1967. Capital, vol. 1. International Publishers, New York (Original published 1867). McAleer, M., 1994. Sherlock Holmes and the search for truth: a diagnostic tale. J. Econ. Surv. 8, 317e370. McManus, P., Walmsley, J., Argent, N., Baum, S., Bourke, L., Martin, J., Sorensen, A., 2012. Rural community and rural resilience: what is important to farmers in keeping their country towns alive? J. Rural Stud. 28, 20e29. Nelson, R., Winter, S., 1982. An Evolutionary Theory of Economic Change. Belknap Press, Cambridge MA. Page, S., 2006. Path dependence. Q. J. Political Sci. 1, 87e115. Perron, P., 1994. Trend, unit root and structural change in macroeconomic time series. In: Bhaskara Rao, B. (Ed.), Cointegration for the Applied Economist. St. Martin’s Press, London, pp. 1131e1146. Plummer, P., Tonts, M., 2013a. Do history and geography matter? Regional unemployment dynamics in a resource dependent economy: evidence from Western Australia, 1984e2010. Environ. Plan. 45, 2919e2938. Plummer, P., Tonts, M., 2013b. Income growth and employment in the Pilbara: an evolutionary analysis. Aust. Geogr. 44, 227e241. Plummer, P., Taylor, M., 2001a. Theories of local economic growth (part 1): concepts, models, and measurement. Environ. Plan. 33, 219e236. Plummer, P., Taylor, M., 2001b. Theories of local economic growth (part 2): model specification and empirical evaluation. Environ. Plan. 33, 385e398. Pritchard, B., 2000. Negotiating the Two-edged sword of agricultural trade liberalisation: trade policy and its protectionist discontents. In: Pritchard, B., McManus, P. (Eds.), Land of Discontent: The Dynamics of Change in Rural and Regional Australia. University of New South Wales Press, Kensington, pp. 90e 104. Pritchard, B., Burch, D., Lawrence, G., 2007. Neither ‘family’ nor ‘corporate’ farming: Australian tomato growers as farm family entrepreneurs. J. Rural Stud. 23, 75e 87. Pritchard, B., Tonts, M., 2011. Market efficiency, agriculture and prosperity in rural Australia. In: Tonts, M., Siddique, M.A.B. (Eds.), Globalisation, Agriculture and Development: Perspectives from the Asia-Pacific. Edward Elgar, Cheltenham, pp. 29e53. Rafiqui, P., 2009. Evolving economic landscapes: why new institutional economics matters for economic geography. J. Econ. Geogr. 9, 329e353.

Redding, S., 2010. The empirics of new economic geography. J. Regional Sci. 50, 297e311. Robinson, J., 1979. History versus equilibrium. Collect. Econ. Pap. 5, 48e58. Schumpeter, J., 1939. Business Cycles: a Theoretical, Historical, and Statistical Analysis of the Capitalist Process. McGraw-Hill, New York. Setterfield, M., 2010. Hysteresis. Trinity College, Department of Economics Working Paper 10-04 1-12, Dublin. Setterfield, M., 1988. Adjustment asymmetries and hysteresis in simple dynamic models. Manch. Sch. 3, 283e301. Setterfield, M., 1995. Historical time and economic theory. Rev. Political Econ. 7, 1e 27. Sheppard, E., Barnes, T., 1990. The Capitalist Space Economy: Geographical Analysis After Ricardo, Marx and Sraffa. Unwin Hyman, London. Skerratt, S., 2013. Enhancing the analysis of rural community resilience: evidence from community land ownership. J. Rural Stud. 31, 36e46. Smailes, P., 1996. Entrenched farm indebtedness and the process of agrarian change. In: Burch, D., Rickson, R., Lawrence, G. (Eds.), Globalisation and Agri-food Restructuring: Perspectives from the Australasia Region. Ashgate, Aldershot, pp. 301e322. Smailes, P., 1997. Socio-economic change and rural morale in South Australia, 1982e 1993. J. Rural Stud. 13, 19e42. Smith, N., 1984. Uneven Development: Nature, Capital and the Production of Space. Basil Blackwell, New York. Tong, H., 1990. Non-linear Time Series: a Dynamical Systems Perspective. Oxford University Press, Oxford. Tonts, M., Argent, N., Plummer, P., 2012. Evolutionary perspectives on rural Australia. Geogr. Res. 50, 291e303. Tonts, M., Davies, A., Haslam-McKenzie, F., 2008. Regional Workforce Futures: an Analysis of the Great Southern, South West and Wheatbelt Regions, Geowest No. 35. Institute for Regional Development, The University of Western Australia, Crawley. Veblen, T., 1898. Why is economics not an evolutionary science. Q. J. Econ. 12, 373e 397. Wilson, G., 2010. Multifunctional ‘quality’ and rural community resilience. Trans. Inst. Br. Geogr. 35, 364e381. Wilson, G., 2012. Community resilience, policy corridors and the policy challenge. Land Use Policy 31, 298e310.

Please cite this article in press as: Tonts, M., et al., Path dependence, resilience and the evolution of new rural economies: Perspectives from rural Western Australia, Journal of Rural Studies (2014), http://dx.doi.org/10.1016/j.jrurstud.2014.04.001