Mimetic isomorphism, offshore outsourcing and backshoring decisions among micro and small enterprises

ABSTRACT

This paper exploits novel data on Italian manufacturing firms to investigate how mimetic isomorphism influences offshore outsourcing and backshoring decisions among micro and small enterprises (MSEs). The results show that offshore outsourcing by peers located in the same home region positively affects the likelihood that an MSE will engage in the same global sourcing strategy. When these dynamics emerge, firms are more likely to engage in long-lasting foreign subcontracting relationships. Mimetic isomorphism is only detected during the early stages of the internationalization process, while backshoring practices appear to be the result of more independent decisions.

KEYWORDS:

• offshoring
• network
• micro and small enterprises (MSEs)
• backshoring
• mimetic isomorphism
• reshoring
• internationalization

JEL:

• F23
• L14
• L24
• R12

INTRODUCTION

In the last decades, the number of businesses engaged in international activities has increased dramatically. In this context, relocation of business functions abroad (offshoring) is becoming particularly significant: firms invest in new or existing production facilities (captive offshoring) or establish relationships with foreign suppliers (offshore outsourcing) in an attempt to exploit the benefits arising from lower labour costs, access to natural resources and the possibility to leverage diverse sources of knowledge (Kinkel & Maloca, 2009). The core of the debate on offshoring has initially revolved around large firms, which tend to choose forms of direct participation in local companies producing abroad (Dunning, 1980). However, recent contributions have argued that these organizations are not alone in their journey (Oviatt & McDougall, 1994), highlighting that production relocation strategies are increasingly spreading from large producers to smaller firms (Di Gregorio et al., 2009).

Previous research has reported a higher propensity among smaller producers to relocate their production activities abroad through offshore outsourcing (Cutrini, 2011; Kinkel & Maloca, 2009). Despite its potential advantages, this global sourcing mode carries significant business risks and is characterized by high levels of complexity: when interacting with foreign suppliers, managers often face unexpected challenges, such as coordination or quality issues, which tend to become manifest only after relocation decisions are made (Larsen, 2016). Therefore, the ability to anticipate and limit potential challenges is critical for the success of offshore outsourcing decisions. Internal assessment and strategic planning procedures are often too costly for small firms (Zhu et al., 2012), which often rely on knowledge acquired by peers already active in international markets (Baum et al., 2015; Di Gregorio et al., 2009): this practice is commonly known as mimetic isomorphism (DiMaggio & Powell, 1983). Previous research has shown that, for smaller firms, the most common reference group is represented by peers located in the same home region (Fernhaber & Li, 2013; Milanov & Fernhaber, 2014). Given the relevant role played by local interactions in the domestic environment, it is surprising that there is so little empirical work inspecting these dynamics.

The present article contributes to advance the current debate by exploring how domestic co-location of firms affects the likelihood of offshore outsourcing decisions and the choice of foreign locations among micro and small enterprises (MSEs). The proposed empirical analysis also aims to evaluate the role of domestic interactions among neighbouring peers on the duration of offshore outsourcing, analysing MSEs’ backshoring decisions: in the following subsections, backshoring is intended as the voluntary choice to relocate activities back to the firm’s home country (Bettiol et al., 2019).¹ The aim is to address the following three research questions:

• To what extent do domestic peers located in the same home region influence MSEs’ decisions to engage in offshore outsourcing?

• Does mimetic isomorphism affect the duration of international subcontracting relationships among MSEs?

• What is the role of peers located in the same home region in influencing backshoring strategies among MSEs?

This piece of work adds new insights to the current debate on small firms’ internationalization, emphasizing the impact of domestic regional relationships on MSEs’ offshore outsourcing and backshoring decisions. The investigation exploits unique data from the Italian Ministry of Economy and Finance Annual Survey (IMEFAS), including an extensive sample of MSEs operating in the clothing and footwear industries between 2006 and 2012. The main database is integrated with province-level information on captive offshoring extracted from Bureau van Dijk’s Analisi Informatizzata delle Aziende Italiane (AIDA), in an attempt to evaluate both the influence of small neighbouring producers and that of larger firms operating in the same local environment and engaged in different offshoring strategies. The territorial level used to identify local interactions is the Italian province, which is equivalent to a NUTS-3 region according to the European Union classification.

The contribution of this work to the literature is threefold. First, the main findings advance the current understanding of internationalization processes by providing a more comprehensive overview of the role of local domestic peers as learning sources for MSEs. This topic has only recently gained attention among researchers (Chevassus-Lozza & Galliano, 2003; Fernhaber & Li, 2013; Milanov & Fernhaber, 2014) and has often been limited to the analysis of formal interactions. Second, the peculiar structure of the IMEFAS database allows to detect mimetic isomorphism associated with both offshore outsourcing and backshoring, thus identifying the role of these mechanisms during the different stages of the internationalization process. Third, the analysis is implemented on an extensive longitudinal sample of MSEs that permits greater generalization than case studies. To the best of our knowledge, this is the first work that provides a comprehensive analysis of offshore outsourcing and backshoring dynamics among MSEs in a specific sector for an entire country. In this respect, Italy represents a particularly interesting setting for the analysis, considering the widespread diffusion of subcontracting activities in several manufacturing sectors, especially within industrial districts (Capasso et al., 2013; Giunta et al., 2012): inside these local systems, firms continuously share information through voluntary and involuntary mechanisms, stimulating knowledge diffusion processes where imitation and learning lead to the emergence of best-practices that are shared among the local community (Canello & Pavone, 2016).

THEORETICAL FRAMEWORK AND HYPOTHESES DEVELOPMENT

Mimetic isomorphism and offshore outsourcing decisions among MSEs

In the literature, the decision to engage in international activities has been studied by numerous researchers. Early theories describe internationalization as the result of a rational process, where firms acquire knowledge and experiences independently before moving to a specific foreign location (Dunning, 1980; Johanson & Vahlne, 1977). Firm-centric approaches are generally more appropriate for larger businesses, as they highlight the importance of possessing unique resources and capabilities, as well as the critical role of experiential knowledge (Casillas et al., 2009). However, these views overlook the fact that most economic agents are bounded rational and make decisions under limited information.

The assumption of bounded rationality is particularly relevant for smaller firms, which are increasingly active in global markets thanks to the recent advances in transportation and communication technologies (Lloyd-Reason & Sear, 2007). Despite the increasing opportunities, these businesses still face significant resource and size constraints relative to their larger competitors and are characterized by higher levels of risk aversion (Di Gregorio et al., 2009): therefore, their success is critically associated with the ability to access relevant external knowledge and overcome the liability of foreignness and smallness (Bruneel et al., 2010).

The recent debate on internationalization strategies by bounded rational actors have been framed by two main theories, the network and the neo-institutional approach. In both cases, the common assumption is that, when resources for ex-ante planning are scarce and entrepreneurs are risk averse, internationalization is rarely the result of an isolated decision. Network theories postulate that firms are sets of interlinked relationships that become interdependent through cooperation and competition (Ruzzier et al., 2006): therefore, the economic agent is viewed as an exchange unit rather than an isolated entity (Johanson & Vahlne, 2009; Musteen et al., 2014). According to the network approach, formal and informal interactions with the external environment can increase awareness of international opportunities, facilitating the process of knowledge acquisition (Dabić et al., 2019; Johanson & Mattsson, 2015). In this respect, both business networks and social ties can represent valuable sources of information, allowing MSEs to overcome resource constraints and reduce risks and uncertainty (Costa et al., 2015; Ellis, 2011). Network theories contribute to explain why internationalization rates are higher among businesses that rely on global relationships since their inception (Coviello & McAuley, 1999; Coviello & Munro, 1997; Romanello & Chiarvesio, 2019).

The neo-institutional approach on locational mimicry takes a slightly different perspective, suggesting that even when formal relationships are not developed, organizations facing uncertainty tend to observe and imitate internationalization practices perceived to be successful (Belderbos et al., 2011; Casillas et al., 2009; Lévesque et al., 2009). As highlighted by the theoretical and empirical literature, mimetic isomorphism is often used by potential entrants to address the lack of experiential knowledge (Bruneel et al., 2010). In the case of small internationalizing firms, imitation is particularly beneficial as it increases managerial confidence and the ability to make complex decisions through reciprocal observation, repetition and experimentation (Lévesque et al., 2009). Such dynamics explain why foreign expansion in specific sectors or countries is often the result of ‘herding’ or ‘bandwagon’ effects (Abrahamson & Rosenkopf, 1993), where ‘birds of a feather flock together’ (Zhu et al., 2012).

Both network and neo-institutional theories emphasize the role of specific reference groups in influencing internationalization decisions of late movers: indeed, the knowledge acquisition process is often oriented towards peers with similar history, experience or location (Guillén, 2003). Geographically proximate peers often represent the most immediate source of information for smaller firms, whose interactions tend to be confined at the local level (Fernhaber & Li, 2013; Manolova et al., 2010). Local mimetic behaviour allows small firms to mobilize the complementary resources and capabilities embedded in local networks (Davenport, 2005), promoting the diffusion of strategic information regarding the customer and competitor base in the foreign country, the institutional framework and the resources needed to succeed abroad (Eriksson et al., 2000).

Mimetic isomorphism has been observed in captive offshoring (Chung & Song, 2004; Rose & Ito, 2008; Xia et al., 2008) and exporting activities (Koenig et al., 2010). However, one can expect the same dynamics to be at play in offshore outsourcing for multiple reasons. First, inter-organizational learning is especially critical in production relocation activities, given they require a substantial organizational transformation compared to that needed for sales-oriented internationalization (Mariotti et al., 2008; Maskell et al., 2007). Second, an increasing number of MSEs are involved in offshore outsourcing (Kinkel & Maloca, 2009), and often begin their global journey on the sourcing side rather than through direct and indirect exports (Di Gregorio et al., 2009). As discussed, during the early internationalization stages these firms often lack experiential knowledge and face significant entry barriers (Kinkel & Maloca, 2009). Third, bandwagon effects have already been observed for domestic outsourcing in the 1990s (Hätönen & Eriksson, 2009) and it is reasonable to expect that similar dynamics occur for international subcontracting decisions.

Given the above considerations, it is hypothesized that when MSEs perceive that neighbouring peers are achieving efficiency gains through offshore outsourcing, they will be more willing to use these experiences to justify and legitimate their own global sourcing decisions. When mimetic isomorphism occurs, it is expected that, with all else equal, the probability for an MSE to start offshore outsourcing will be higher when more industry peers in the same home region are engaged in the same type of internationalization activities. Hence, the following hypothesis is posited:

Hypothesis 1: Offshore outsourcing by industry peers located in the same home region increases MSEs’ propensity to initiate offshore outsourcing.

Mimetic isomorphism and backshoring decisions among MSEs

In the recent past, consensus has been reached on the fact that production internationalization cannot be simply described as a linear and unidirectional process, involving the relocation of low value-added activities from developed to developing countries. Evidence shows that an increasing number of offshoring firms are voluntarily choosing to partly or fully relocate their production activities back to their home countries, triggering a process described as ‘backshoring’ (Rasel et al., 2020) or ‘reshoring’. This phenomenon is now widespread in most developed countries and involves both large and small firms (Ancarani et al., 2015; Bailey & De Propris, 2014; Gylling et al., 2015).

According to the literature, backshoring can be explained by three main conceptual frameworks. First, this strategy can be the consequence of exogenous or endogenous changes in the competitive environment of the host country, which reduce the attractiveness of the foreign location (Martínez-Mora & Merino, 2014). Second, backshoring can be the effect of a strategic shift of the offshoring firm, aimed at repositioning its brand (Boffelli et al., 2020; Di Mauro et al., 2018), increasing synergies between manufacturing activities and research and development (Di Mauro et al., 2018), pursuing the triple bottom line (Barbieri et al., 2018; Fratocchi & Di Stefano, 2019) or adopting new production technologies (Ancarani et al., 2019; Ancarani & Di Mauro, 2018; Dachs et al., 2019). Third, backshoring can configure as an error correction of previous managerial mistakes (Gylling et al., 2015; Kinkel, 2014; Kinkel & Maloca, 2009). Indeed, offshore outsourcing decisions are often supported by simple heuristics based on cost reduction, which tend to overlook important soft factors and are based on assumptions that are not reliable in the medium and long term (Gray et al., 2017; Kinkel & Maloca, 2009). In several cases, firms fail to effectively evaluate the trade-off between lower labour costs and the large friction-related costs associated with quality problems, loss of knowledge and extended lead times (Kinkel & Maloca, 2009).

Large and small firms often face different challenges in their global activities and this reflects on their backshoring strategies. Misjudgements are more common among smaller firms, which often rely on incomplete and inefficient assessment methods and lack systematic location planning and forecasting capabilities (Kinkel, 2014; Kinkel et al., 2007). These aspects are often coupled with the inability to implement direct monitoring and control procedures (Manning, 2014; Nujen et al., 2018), increasing the exposure to opportunistic behaviours from foreign suppliers and leading to higher coordination costs and extended delivery times. This combination of factors contributes to explain why smaller firms repatriate manufacturing activities earlier compared to large producers and are generally more satisfied with their backshoring experiences (Barbieri et al., 2018). Given the above considerations, it seems reasonable to assume that the duration of offshore outsourcing can be a reliable proxy to measure the success of these initiatives. Indeed, when backshoring is observed shortly after the production relocation process, this decision is more likely to reflect the inability of the MSE to anticipate substantial hidden and unexpected costs of international subcontracting.

In this paper, one of the key assumptions is that mimetic isomorphism has an impact not only on the decision to start offshore outsourcing, but also on how this activity is initiated (Ancarani et al., 2015; Kamal & Sundaram, 2016). This assumption is consistent with the theoretical framework proposed by Rauch and Watson (2003), which assumes that internationalization activities occur through search, investment and rematch. For offshore outsourcing, it is expected that a buyer will face significant search costs whenever (s)he tries to identify the most suitable supplier among a potential pool of foreign partners. The initial cost of matching often represents the most significant share of the total matching costs (Eslava et al., 2014; Obashi, 2010). Given the uncertainty regarding the supplier’s characteristics, one option for the buyer is to start the subcontracting relationship with small orders, in an attempt to learn about the partner’s quality. However, low levels of commitment reduce the likelihood that the buyer will invest resources to cooperate with the supplier when issues arise (Córcoles et al., 2015; Rauch & Watson, 2003): in such circumstances, the optimal strategy for the buyer is to abandon the subcontracting relationship instead of making relationship-specific investments. This aspect contributes to explain why a significant share of trade relationships are short lived (Besedeš & Prusa, 2006a; Hess & Persson, 2011; Nitsch, 2009).

We claim that the presence of pioneers in the focal firm’s domestic region reduces the informational frictions underlying the search process. During their exploration activities, pioneers acquire critical information regarding the requirements and challenges associated with establishing subcontracting links in a foreign country (Wagner & Zahler, 2015). By observing first movers’ decisions, followers can acquire direct or indirect information on the reliability and strength of potential suppliers (Al-Laham & Souitaris, 2008; Besedeš & Prusa, 2006b), including their ability to fulfil delivery deadlines, quality requirements and technical needs (Albornoz et al., 2016). By increasing buyers’ confidence, local interactions influence the initial transaction size of late movers, positively affecting the duration of offshore outsourcing. Hence, the following hypothesis is posited:

Hypothesis 2: The duration of offshore outsourcing among MSEs is higher when offshore outsourcing is influenced by domestic industry peers located in the same home region.

Finally, the role of bandwagon effects in backshoring decisions is also investigated. Despite the core of the literature has focused attention on host country and firm-specific determinants, recent contributions have highlighted the key role of home country factors in explaining backshoring dynamics (Baraldi et al., 2018; Wiesmann et al., 2017). In this paper, the above-mentioned approach is further developed, claiming that the transformations of the domestic regional environment can contribute to explain backshoring strategies among MSEs. This claim is based on the assumption that, when manufacturing activities are relocated abroad and domestic subcontractors are displaced, the subcontracting infrastructure in the home region can be severely affected. In such cases, the domestic subcontracting base can collapse, disrupting the entire local supply chain and negatively affecting the stability of informal social networks (Baraldi et al., 2018). In several cases, restoring product and process competences that have been outsourced years before can be extremely challenging (Kinkel, 2014; Pisano & Shih, 2009). Even when domestic suppliers are still active, reshoring firms might face resistance, lack of commitment or distrust, as they might be perceived as unfaithful to the local domestic community (Baraldi et al., 2018).

When such reshoring barriers are present (Engström et al., 2018a, 2018b), being a late mover might be a rational strategy for an MSE that considers moving production back to its own region. If domestic neighbouring industry peers succeed in re-establishing the subcontracting network in the home region, reshoring barriers can be substantially reduced and the reshoring strategy can be more effective. This approach can be particularly convenient when the repatriation process is implemented through ‘outsourced backshoring’ (Mlody, 2016), which requires re-embedding activities in domestic subcontracting networks (Baraldi et al., 2018). In such cases, the successful outcome of backshoring typically depends on how the reshored activities fit in the local resource constellation and activity pattern (Baraldi et al., 2018). By choosing a late mover strategy, the focal firm can free-ride on the investments made by first movers to recreate suitable resource interfaces and a new subcontracting infrastructure. For these reasons, we can expect MSEs to imitate backshoring decisions of industry peers located in the same domestic region, leading to the same local bandwagon effects observed for offshore outsourcing. Despite the presence of mimetic isomorphism in backshoring has been regarded as plausible (Gray et al., 2013), no empirical evidence currently exists to support these theoretical claims. Therefore, in this paper the following final hypothesis is also tested:

Hypothesis 3: Backshoring decisions of industry peers located in the same home region increase the likelihood of backshoring strategies among MSEs.

METHODOLOGY

Data sources and sample

The primary data source used to implement the empirical analysis is the IMEFAS database. The dataset includes all Italian firms with a turnover lower than €7.5 million that are required to report their activities to the Italian Tax Revenue Agency. The IMEFAS database is functional for this study for two reasons. First, it contains a significant share of the Italian manufacturing firm population, considering the fragmented structure of the economy in this country: according to Eurostat’s Structural Business Statistics, 83% of the Italian manufacturing businesses in 2015 employed fewer than 10 workers. Despite representing a large share of the secondary sector, the wide majority of these micro producers are not part of the most commonly available firm-level datasets (Canello, 2016; Canello et al., 2017). Second, the structure of the IMEFAS database allows one to track the intensity and evolution of offshore outsourcing over time: indeed, the survey contains firm-level information on the costs associated with domestic and foreign subcontracting agreements and identifies when a specific manufacturing activity is transferred to an external third party abroad and also when the activity is moved back to Italy.

Despite its benefits, the IMEFAS database does not include information on internationalization activities conducted by larger firms. As discussed in the introductory section, these producers tend to prefer captive offshoring for their global sourcing strategies. To evaluate the presence of imitation across different internationalization modes, information on medium and large Italian companies (turnover higher than €7.5 million) having controlling interests in foreign production units was extracted from AIDA and used to calculate some of the variables described in the following section.² Data on captive offshoring is used to evaluate whether mimetic isomorphism stems from neighbouring MSEs engaged in offshore outsourcing or if the same mechanisms are generated by larger local firms engaged in different production internationalization modes.

The sample extracted from the IMEFAS dataset for the first part of the empirical analysis includes 15,902 MSEs (equivalent to 63,608 firm–region observations) operating in the footwear and clothing industry³ in all years of the period 2006–12. In the second part of the investigation, the focus is moved to a subsample of 1966 producers that established international subcontracting links in any of the years between 2007 and 2011 and have been active in offshore outsourcing for at least one year during the period considered. The choice of footwear and clothing production is motivated by the fact that, in both industries, the impact of production internationalization has been particularly widespread during the recent decades: according to United Nations Conference on Trade and Development (UNCTAD) (2011), garments and footwear account for one fourth of total offshore outsourcing in the world, generating significant employment opportunities in developing countries. Recently, the Italian clothing and the footwear industry has experienced an increasing diffusion of backshoring practices, following a generalized trend observable in both labour-intensive and high-tech sectors (Fratocchi et al., 2014).

Descriptive analysis

The descriptive analysis of offshore outsourcing presented in this section provides some preliminary indications on the international subcontracting strategies implemented by the MSEs included in the considered sample. Table 1 reports the dynamics of offshore outsourcing and backshoring between 2006 and 2012, highlighting several interesting patterns.

Table 1. Offshore outsourcing and backshoring strategies among micro and small enterprises (MSEs) in the clothing and footwear industry, 2006–12.

	2006	2007	2008	2009	2010	2011	2012	Average
Total MSE population
Total number of MSEs	19,896	19,852	19,063	16,827	16,661	16,181	15,809	17,755
Turnover (€ thousands)	809	827	848	848	897	934	929	870
MSEs engaged in offshore outsourcing
Number of MSEs engaged in offshore outsourcing	1571	1287	1218	1199	1204	1130	1079	1241
Number of new offshore outsourcing MSEs	270	252	167	141	234	140	115	188
% of new offshore outsourcing MSEs with subcontracting relationships in Italy	91.2%	84.6%	87.8%	95.1%	91.0%	88.3%	90.9%	90.7%
Turnover (€ thousands)	703	704	697	643	682	725	703	694
Average cost of subcontracting production in foreign countries (€ thousands)	228	298	321	332	370	397	373	331
Average cost of subcontracting production in Italy (€ thousands)	271	301	344	322	341	353	355	327
Destination
Europe	64.4%	66.8%	66.0%	66.7%	68.4%	70.1%	70.8%	67.6%
North Africa	10.8%	11.5%	12.6%	12.4%	12.7%	13.9%	14.4%	12.6%
East Asia	11.8%	10.9%	10.6%	11.4%	10.5%	9.4%	8.7%	10.5%
Other world regions	13.0%	10.8%	10.8%	9.5%	8.3%	6.7%	6.1%	9.3%
Backshoring MSEs
Number of backshoring MSEs	250	484	261	216	256	184	200	264
% of backshoring MSEs with domestic subcontracting relationships at t − 1 and t	80%	68%	74%	75%	70%	74%	68%	72.7%

Sources: Author’s elaborations of the Italian Ministry of Economy and Finance Annual Survey (IMEFAS) and Bureau van Dijk’s Analisi Informatizzata delle Aziende Italiane (AIDA) data.

The first relevant aspect relates to the number of MSEs (1241 on average) involved in offshore outsourcing: despite representing a relatively small share (6.9%) of the entire MSEs population, this number is significant considering that only 174 medium and large firms internationalized through captive offshoring during the same period. This trend confirms that, when the analysis of production internationalization is limited to captive offshoring, a significant part of the entire offshoring phenomenon is inevitably overlooked. The second aspect relates to the temporal dynamics reported in Table 1: the data show that, despite the presence of a declining pattern, MSEs have been engaged in offshore outsourcing during the entire time span considered. Backshoring activities seem also to be distributed across the entire time span, with a more significant concentration in 2007. The third interesting finding is that, when production activities are relocated abroad, most MSEs do not seem to disrupt their domestic subcontracting relationships: this attitude is visible from the balanced distribution of domestic and international subcontracting costs among offshore outsourcing firms (€327,000 and €331,000, respectively), as well as from the share of MSEs that maintain their domestic subcontracting relationships when production activities are moved abroad (90.7% on average). Interestingly, the wide majority of backshoring firms (72.7%) maintain their relationships with domestic subcontractors even after production is backshored.

The geographical patterns reported in Table 1⁴ highlight a clear tendency among most micro and small producers to choose Europe over more distant destinations for international subcontracting activities. The choice of European destinations seems to have become more popular over time and is in line with previous findings (Hätönen & Eriksson, 2009). This pattern is possibly driven by the increasing attractiveness of Eastern Europe as an offshoring destination during the considered period (Di Mauro et al., 2018; Kinkel & Maloca, 2009).

Analytical methods

The empirical investigation presented in this paper aims to evaluate the extent to which the patterns identified in the descriptive analysis can be explained by mimetic isomorphism. The analysis exploits the panel structure of the IMEFAS database and is composed of two main parts. In the first part, the antecedents of offshore outsourcing decisions are inspected using a dynamic probit approach: this model is aimed to test Hypothesis 1. In the second part of the paper, the determinants of backshoring are analysed using the Cox Proportional Hazard model: in this case, the focus is on Hypotheses 2 and 3.

Part 1: Antecedents of offshore outsourcing decisions

The first part of the empirical analysis aims to evaluate the role of imitation in influencing an MSE’s decision to engage in offshore outsourcing. The methodology used to test Hypothesis 1 is a dynamic probit regression with random effects. The general specification of the empirical model is the following: (1) $y_{ijk{l}_t}={\text{β}}_0+{\text{β}}_1{y}_{ijk{l}_{t-1}}+{\mathit{\gamma }}^{{}^{\mathrm{\prime }}}{\mathit{x}}_{ijk{l}_{t-1}}+{\mathit{\delta }}^{{}^{\mathrm{\prime }}}{\mathit{x}}_{i_{t-1}}+{\alpha }_i+{ϵ}_{it}$(1) The set of regressors includes the lagged value of the dependent variable, capturing the autoregressive nature of this process, and a set of firm-level and contextual variables that are also lagged at time t−1 to address potential reverse causality bias.⁵ In order to address the initial condition problem of non-linear dynamic models, the approach proposed by Wooldridge (2005) is used.

In this model, the dependent variable $y_{ijk{l}_t}$ is a dummy equal to 1 if at time $t$ the ith firm located in the jth province and operating in the kth industry transfers one or more production activities to a supplier located in the lth foreign region.

The key independent variables are identified as follows:

• Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region: this variable is expected to capture imitation effects generated by MSEs operating in province j and industry k and engaged in offshore outsourcing in the same foreign region l. It is expected that a larger presence of domestic neighbouring peers engaged in international subcontracting in a specific destination region positively affects the decision of an MSE to internationalize production in the same world region.

• Number of medium and large peers located in the same domestic region and engaged in captive offshoring in the same foreign region: the model aims to evaluate whether bandwagon effects are restricted to offshore outsourcing by MSEs or if captive offshoring strategies of medium and large producers operating in province j and industry k have the same triggering effect. This aspect is captured using the number of medium and large firms in the province owning or controlling production units in the ith foreign region (data extracted from AIDA) and applying the same criterion used to compute the preceding variable.

The model contains two additional variables to verify whether imitation occurs between firms operating in different foreign destinations:

• Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in other foreign regions.

• Number of medium and large peers located in the same domestic region and engaged in captive offshoring in other foreign regions.

All the four above mentioned variables are computed by taking the average number of firms relocating production abroad in the two years immediately preceding each time period $t$. This choice is motivated by the observation that the knowledge acquired by the more recent entrants is generally more valuable compared to that of long-term incumbents: in most cases, the experiences of these peers are directly applicable to address the challenges faced by the MSEs that are planning to internationalize (Chung & Song, 2004).

The specification includes an extensive set of firm- and regional-level covariates to control for the main determinants that are believed to influence offshore outsourcing decisions among MSEs. A detailed explanation of the data sources and the criteria used to construct these variables is provided in Table A1 in the supplemental data online. Organizational-level factors are evaluated discriminating between standard firm-level characteristics, such as size, labour productivity, use of capital-intensive technology, presence of financial constraints, and aspects related to relevant prior knowledge acquired by the firm, such as previous experience with exporting (number of years) and domestic outsourcing (relative costs associated with these activities). The choice of this set of control variables is motivated by recent findings in the literature regarding the significant relationship between the probability of internationalizing and factors including business size (Fillis, 2001), productivity (Helpman et al., 2004), buyer–supplier relationships (Martin et al., 1998), access to credit (Fillis, 2001) and experiential knowledge in exporting (Kinkel & Maloca, 2009) and domestic outsourcing (Kotabe et al., 2003).

Regarding environmental factors, the two variables used to measure the quality of local institutions are Government, indicating the ability of the local government to effectively formulate and implement policy measures, and Corruption, measuring the degree of corruption found in those who perform public functions. Both variables are taken from the Italian Institutional Quality Database (Nifo & Vecchione, 2014). The specification also includes other regional-level variables such as Average employee wages, proxy for the cost of labour, the Share of migrant firms operating in province j, industry k, the population size of province j and a measure of the quality of transport and digital infrastructure of the province. Finally, the model includes a set of regional, sectoral and cohort dummies to capture time-invariant effects associated with the specific location of the firm, the industry in which the firm is operating and the specific year in which the event is observed. A descriptive overview of the variables together with the correlation matrix is presented in Table A2 in the supplemental data online.

Part 2: Determinants of backshoring decisions

The second part of the empirical investigation focuses on the duration patterns of offshore outsourcing among the Italian manufacturing firms included in the sample. The main aim in this case is twofold:

• To evaluate whether MSEs which learn vicariously before relocating production abroad are those characterized by more durable offshore outsourcing activities (Hypothesis 2).

• To assess whether the duration of MSEs’ offshore outsourcing is affected by backshoring decisions adopted by peers located in the focal firm’s home region (Hypothesis 3).

The methodology used to test the above statements is the Cox Proportional Hazard model, which represents the most popular approach for the analysis of survival data. In this model, backshoring takes place when the firm ceases to outsource one or more parts of the production process to foreign suppliers: therefore, we consider both the situation in which outsourced production is moved from foreign suppliers back to domestic suppliers and also the case in which production is moved back to the firm’s home country and performed internally. The model can be specified as follows: (2) $h\left(t_i\right)=\mathrm{e}\mathrm{x}\mathrm{p}({\text{β}}_0+{\text{β}}^{\mathrm{\prime }}{x}_i)h_0\left(t_i\right)i=1,\dots ,n$(2) where the hazard $h\left(t_i\right)$ for firm $i$ at time $t$ is the product of the baseline hazard rate, which is left unspecified, and a linear function of a set of covariates that are expected to influence the probability of backshoring.

The three key regressors in the Cox PH model are the following:

• Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing before the focal firm's internationalization decision; this variable is expected to capture the role of mimetic isomorphism in affecting the duration of MSEs’ offshore outsourcing activities. It is expected that a larger average number of MSEs’ engaged in international subcontracting in the period preceding the beginning of offshore outsourcing can positively affect the likelihood of durable production internationalization strategies (Hypothesis 2).

• Number of medium and large peers located in the same domestic region and engaged in captive offshoring in the same foreign region before the focal firm’s internationalization decision: this regressor measures the extent to which the effects described for the previous variable are also generated by captive offshoring decisions by medium and large firms. The effect is evaluated by including the average number of medium and large firms engaged in captive offshoring in the province where the firm is located in the period preceding the beginning of offshore outsourcing. The variable is calculated using data from AIDA.

• Number of micro and small peers located in the same domestic region and backshoring from the same foreign region: this variable allows to evaluate whether MSEs’ backshoring practices are the result of an autonomous decision or are rather affected by the backshoring strategies adopted by domestic neighbouring producers (Hypothesis 3). It is calculated as the average number of MSEs engaged in backshoring in the province where the firm is located in the period preceding time t.

The first and the second variable are computed by taking the average number of firms relocating production abroad in the two years immediately preceding the relevant time period under consideration. The rationale behind this approach is discussed for the dynamic probit model.

The set of control variables includes time-varying and time-invariant factors that are believed to influence the duration of offshore outsourcing activities, following the indications provided by the recent literature (Fratocchi et al., 2016). The list of firm-level control variables includes firm size, the relative costs of domestic outsourcing, the share of revenues from the main client firm, the experience as exporter, the use of capital-intensive technology and the number of migrant firms operating in the same province and industry. In addition to the set of industry, territorial and cohort dummies used in the dynamic probit model, all the specifications include a set of dummies capturing the fixed effects associated with the specific destination region of offshore outsourcing activities. Additional information on the criteria used to construct the above-mentioned variables is provided in Table A3 in the supplemental data online.

RESULTS

Main models

The results of the dynamic probit model are reported in Table 2, focusing on three different specifications. The estimates in model [A] confirm that offshore outsourcing decisions among MSEs are affected by a combination of firm-specific and contextual factors. As far as the former aspects are concerned, the choice to subcontract production activities to foreign suppliers is more frequent among local actors who have developed superior human, financial and technological skills. Experiential knowledge seems to play an equally eminent role in increasing the likelihood of MSEs to choose production relocation activities. Contextual factors also contribute to a significant extent to explain production internationalization decisions: indeed, areas characterized by lower institutional quality and higher degrees of corruption foster MSEs to look for outsourcing opportunities abroad.

Table 2. Dynamic probit model: probability to engage in offshore outsourcing in foreign region l at time t, 2007–12, industries: clothing and footwear production.

Variable	[A]		[B]		[C]
	Coefficient	SE	Coefficient	SE	Coefficient	SE
yt−1	2.434***	0.055	2.363***	0.055	2.335***	0.055
yt0	1.763***	0.105	1.583***	0.096	1.581***	0.096
Internal factors
Firm-level characteristics
Size	0.005***	0.001	0.005***	0.001	0.005***	0.001
Labour productivity	0.002***	0.000	0.002***	0.000	0.002***	0.000
Supplier	−0.171***	0.032	−0.159***	0.032	−0.162***	0.032
Presence of financial constraints	0.266***	0.044	0.264***	0.043	0.264***	0.043
Use of capital-intensive technology	0.262***	0.033	0.265***	0.032	0.267***	0.033
Experiential knowledge
Exporting experience	0.046***	0.005	0.044***	0.005	0.045***	0.005
Domestic outsourcing experience	0.071***	0.011	0.070***	0.011	0.070***	0.011
External factors
Province-level factors
Government	−0.284	0.235	−0.342	0.227	−0.337	0.228
Corruption	0.357***	0.116	0.357***	0.114	0.359***	0.115
Quality of infrastructure	0.001*	0.000	0.001**	0.000	0.001**	0.000
Average employee wages	0.007*	0.004	0.006*	0.004	0.007*	0.004
Population (100,000)	0.004*	0.002	0.004*	0.002	0.004*	0.002
Share of migrant firms in province j and industry k	−0.016***	0.005	−0.015***	0.004	−0.015***	0.004
Mimetic isomorphism
Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region			0.044***	0.005	0.039***	0.006
Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in other foreign regions			−0.029***	0.007	−0.025***	0.007
Number of medium and large peers located in the same domestic region and engaged in captive offshoring in the same foreign region					0.044***	0.010
Number of medium and large peers located in the same domestic region and engaged in captive offshoring in other foreign regions					−0.031**	0.013
Industry dummies	Yes		Yes		Yes
Territorial dummies	Yes		Yes		Yes
Cohort dummies	Yes		Yes		Yes
Observations	381,632		381,632		381,632
Firm–region combinations for each year	63,608		63,608		63,608
Log-likelihood	−7609.11		−7533.50		−7523.11

Sources: Author’s elaborations of the Italian Ministry of Economy and Finance Annual Survey (IMEFAS) and Bureau van Dijk’s Analisi Informatizzata delle Aziende Italiane (AIDA) data.

In models [B] and [C], information on the local pool of internationalizing peers is added to the initial set of control variables, allowing to directly assess the role played by mimetic isomorphism. The results provide evidence in support of the construct validity of Hypothesis 1, showing higher probabilities of starting offshore outsourcing activities in those areas characterized by a higher number of neighbouring MSEs already engaged in the same internationalization modes. Interestingly, the dynamics changes as far as different world regions are concerned: indeed, the variable Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in other foreign regions has a negative impact on the probability to initiate offshore outsourcing in the target destination region. The results also show that captive offshoring strategies of larger firms affect MSEs’ relocation decisions: the variable Number of medium and large peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region is positive and significant, whereas negative effects are detected for different destination regions.

Further relevant insights emerge from the duration analysis reported in Table 3: the three models provide an overview of the influence of internal and external factors with respect to backshoring decisions. The coefficients of the main control variables are significant and the sign is consistent with our expectations. The analysis of mimetic isomorphism (models II and III) provides support for the validity of Hypothesis 2 for smaller local peers, suggesting that, when internationalization decisions are driven by previous relocation strategies by micro and small neighbouring producers, MSEs are more likely to engage in long-lasting international subcontracting relationships. Model III shows that such dynamics do not occur across different internationalization modes.

Table 3. Cox PH model: relative risks of backshoring among manufacturing firms in Italy, 2007–12, industries: clothing and footwear production.

Variable	[I]		[II]		[III]		[IV]
	Hazard ratio	SE	Hazard ratio	SE	Hazard ratio	SE	Hazard ratio	SE
Internal factors
Size	0.783***	0.088	0.786***	0.087	0.778***	0.089	0.778***	0.091
Domestic outsourcing costs	1.338**	0.138	1.332**	0.142	1.344**	0.143	1.356**	0.144
Share of revenues from main client firm	0.983***	0.006	0.983***	0.006	0.984***	0.006	0.984***	0.006
Previous experience as exporter	0.750	0.245	0.707	0.244	0.712	0.243	0.734	0.244
Use of capital-intensive technology	9.100*	1.156	8.512*	1.151	8.565*	1.151	8.533*	1.149
External factors
Share of migrant firms in province j and industry k	18.499***	0.866	24.167***	0.870	24.162***	0.873	24.767***	0.880
Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region before the offshore outsourcing decision			0.746***	0.081	0.751***	0.082	0.755***	0.082
Number of medium and large peers located in the same domestic region and engaged in captive offshoring in the same foreign region before the offshore outsourcing decision					0.816	0.208	0.778	0.211
Number of micro and small peers located in the same domestic region and backshoring from the same foreign region							1.398	1.642
Number of micro and small peers located in the same domestic region and backshoring from other foreign regions							2.643	1.601
Number of medium and large peers located in the same domestic region and backshoring from the same foreign region							0.922	0.125
Number of medium and large peers located in the same domestic region and backshoring from other foreign regions							0.923	0.093
Industry dummies	Yes		Yes		Yes		Yes
Domestic region dummies	Yes		Yes		Yes		Yes
Destination region dummies	Yes		Yes		Yes		Yes
Cohort dummies	Yes		Yes		Yes		Yes
Observations	1966		1966		1966		1966

Sources: Author’s elaborations of the Italian Ministry of Economy and Finance Annual Survey (IMEFAS) and Bureau van Dijk’s Analisi Informatizzata delle Aziende Italiane (AIDA) data.

Finally, the validity of Hypothesis 3 is tested in model IV by incorporating variables measuring mimetic isomorphism in backshoring: the results do not provide evidence in support of the validity of the stated hypothesis. More specifically, backshoring decisions among MSEs are not significantly affected by the choices previously made by either larger neighbouring peers or smaller local producers: in this respect, the decision to move production back to the home country seem the result of an individual choice rather than the effect of mimicking behaviour.

Identification and robustness checks

The identification strategy proposed in the previous section is potentially affected by the fact that the present analysis is based on secondary data: therefore, mimetic isomorphism can only be assumed rather than directly demonstrated. The consequences of such an approach are particularly relevant in the first part of the empirical investigation, where the results might potentially be affected by omitted variable bias. Two economic dynamics might influence the bandwagon effects identified in the previous section. First, a larger number of firms might choose the same destination region because it is objectively the best location for the offshored job. For example, several MSEs in an industrial district specialized in the footwear industry might choose to relocate production to North Africa because suppliers in this region can offer better products at lower prices. In such cases, it is expected that it will take some time for all firms to relocate production, but yet any existing offshoring activity in that domestic region will predict further offshoring. The results of the dynamic probit model could also be explained as evidence of learning processes occurring in the destination region: indeed, more offshoring can help the location to develop specific expertise or to improve the local institutional environment and/or the quality of the infrastructures, thus increasing the attractiveness of the foreign location to other would-be offshorers.

In an attempt to verify the robustness of the main findings, the original specification of the dynamic probit model was modified as follows:

• Number of micro and small peers located in the same home country and engaged in offshore outsourcing in the same foreign region was included as additional control variable (model D, Table 4). This variable is expected to capture the industry-level push effect generated by early movers exploiting the advantages of international subcontracting (Jacobides & Winter, 2005), as well as controlling for any effects associated with the emergence of optimal locations for offshore outsourcing activities.

• A set of interaction terms between the industry dummies and the year dummies was added to the model (model E, Table 4).

• Number of medium and large peers located in the same home country and engaged in captive offshoring in the same foreign region was included as additional control variable⁶ (model F, Table 4).

• The model was estimated only considering significant values of offshore outsourcing activity in the province⁷ (model G, Table 4).

• The model was estimated removing the observations located in provinces where the number of footwear and clothing producers was lower than 20 (model H, Table 4).

• The model was estimated removing the observations located in provinces where the number of footwear and clothing producers was lower than 40 (model I, Table 4).

Table 4. Dynamic probit model: robustness checks.

Variable	[D]		[E]		[F]		[G]		[H]		[I]
	Coefficient	SE	Coefficient	SE	Coefficient	SE	Coefficient	SE	Coefficient	SE	Coefficient	SE
yt−1	2.354***	0.055	2.357***	0.055	2.363***	0.054	2.368***	0.054	2.355***	0.056	2.390***	0.056
yt0	1.548***	0.097	1.549***	0.097	1.541***	0.097	1.544***	0.097	1.546***	0.098	1.501***	0.099
Firm-level characteristics
Size	0.005***	0.001	0.005***	0.001	0.005***	0.001	0.005***	0.001	0.005***	0.001	0.005***	0.001
Labour productivity	0.002***	0.000	0.002***	0.000	0.002***	0.000	0.002***	0.000	0.002***	0.000	0.002***	0.000
Supplier	−0.177***	0.033	−0.177***	0.033	−0.176***	0.033	−0.177***	0.033	−0.164***	0.034	−0.147***	0.034
Presence of:
financial constraints	0.278***	0.047	0.278***	0.048	0.278***	0.045	0.278***	0.045	0.281***	0.045	0.326***	0.048
Use of capital-intensive technology	0.278***	0.034	0.278***	0.034	0.278***	0.024	0.277***	0.034	0.283***	0.035	0.274***	0.035
Exporting experience	0.048***	0.005	0.048***	0.005	0.048***	0.005	0.048***	0.005	0.049***	0.005	0.047***	0.005
Domestic outsourcing experience	0.070***	0.010	0.070***	0.011	0.070***	0.011	0.070***	0.011	0.069***	0.011	0.072***	0.012
Province-level factors
Government	−0.290	0.239	−0.214	0.241	−0.225	0.240	−0.227	0.241	−0.245	0.248	−0.461*	0.275
Corruption	0.335***	0.120	0.313***	0.120	0.312***	0.0121	0.327***	0.120	0.320**	0.123	0.367***	0.126
Quality of infrastructure	0.001**	0.000	0.001*	0.000	0.001*	0.000	0.001*	0.000	0.001**	0.000	0.002***	0.000
Average employee wages	0.007*	0.004	0.008*	0.004	0.008*	0.004	0.008*	0.004	0.006	0.004	0.004	0.006
Population (100,000)	0.004	0.002	0.004	0.002	0.004	0.002	0.004	0.002	0.005**	0.002	0.005**	0.002
Share of migrant firms in province j and industry k	−0.016***	0.005	−0.016***	0.005	−0.016**	0.005	−0.017***	0.005	−0.021***	0.005	−0.021***	0.005
Mimetic isomorphism
Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region	0.015**	0.007	0.015**	0.007	0.016**	0.007	0.010*	0.005	0.025**	0.009	0.026**	0.011
Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in other foreign regions	−0.001	0.007	−0.002	0.007	−0.001	0.007	−0.003	0.006	−0.007	0.009	−0.015	0.011
Number of medium and large peers located in the same domestic region and engaged in captive offshoring in the same foreign region	0.013	0.012	0.014	0.013	0.014	0.013	0.015	0.012	0.024	0.018	0.032	0.021
Number of medium and large peers located in the same domestic region and engaged in captive offshoring in other foreign regions	0.004	0.011	0.004	0.011	0.005	0.010	0.005	0.010	0.013	0.012	0.027	0.011
Industry-level factors
Number of micro and small peers located in the same home country and engaged in offshore outsourcing in the same foreign region	0.231***	0.016	0.230***	0.016	0.665***	0.124	0.673***	0.132	0.682***	0.136	0.662***	0.138
Number of medium and large peers located in the same home country and engaged in captive offshoring in the same foreign region					−0.142***	0.128	−0.427***	0.129	−0.459***	0.131	−0.449***	0.134
Industry, territorial and cohort dummies	Yes		Yes		Yes		Yes		Yes		Yes
Industry × Cohort dummies	No		Yes		Yes		Yes		Yes		Yes
Observations	381,632		381,632		381,632		369,632		369,632		350,096
Firm–region combinations for each year	63,608		63,608		63,608		61,608		61,608		58,352
Log-likelihood	−7393.92		−7384.04		−7369.32		−7379.95		−7070.83		−6666.06

Sources: Author’s elaborations of the Italian Ministry of Economy and Finance Annual Survey (IMEFAS) and Bureau van Dijk’s Analisi Informatizzata delle Aziende Italiane (AIDA) data.

The results of all robustness checks are consistent for the Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region, whose impact remains significant and positive. As far as medium and large firms are concerned, the additional checks suggest that the findings of the main specification might be explained by the omitted variable bias. Indeed, the variable Number of medium and large peers located in the same domestic region and engaged in captive offshoring in the same foreign region becomes insignificant when new controls are added to the model.

Finally, a two-stage approach (Boehmke et al., 2006) was implemented to control for the possible presence of selection bias in the duration model. The model allows to control for the factors that may explain the decision to engage in offshore outsourcing and the duration of this strategy. The results of this robustness check are available upon request and confirm the main findings reported in Table 3.

DISCUSSION AND CONCLUSIONS

This paper provides evidence of offshore outsourcing patterns in the manufacturing sector, focusing on the impact of mimetic isomorphism and drawing on longitudinal data on Italian MSEs specialized in clothing and footwear production. The empirical analysis uses an integrated approach, combining different data sources and providing a more extensive assessment of the dynamics behind equity and non-equity internationalization modes. An additional benefit of this empirical analysis is the possibility to inspect backshoring decisions in relation to home country factors, investigating the role of inter-organizational interactions occurring in the domestic business network of the focal firm (Baraldi et al., 2018).

The analysis of the determinants of offshore outsourcing shows that MSEs have higher probabilities to choose these global sourcing strategies when they are surrounded by a higher number of neighbouring MSEs engaged in the same activities. In this respect, relocation decisions by local peers in the home region seem to help MSEs to overcome the initial inertia associated with lateral rigidity, providing guidelines for complex decisions with uncertain outcomes (Fernhaber & Li, 2013). The duration model provides further evidence of how mimetic isomorphism affects offshore outsourcing, suggesting that when internationalization decisions are driven by imitation of micro and small neighbouring producers, MSEs are more likely to engage in long-lasting foreign subcontracting relationships. As highlighted throughout the paper, offshore outsourcing is a complex and risky strategy, involving a number of operational and organizational challenges: when ex-ante evaluations of the benefits and costs are not implemented effectively, firms often face unexpected problems and are forced to revise their decisions in the short term (Ancarani et al., 2015; Manning, 2014). The empirical investigation suggests that mimetic isomorphism decreases the risk of cost estimation errors, driving MSEs to make better locational choices and to establish more durable relationships with foreign suppliers. In this respect, the evidence contradicts previous findings on the detrimental effects of mimetic isomorphism (Mitsuhashi, 2011), suggesting that Italian MSEs operating in the fashion industry do not engage in ‘inertial imitation’ (Albertoni et al., 2019), overlooking latent factors that can lead to negative organizational outcomes (Heimeriks, 2010; Zollo, 2009).

The second key finding of the empirical analysis is that contrary to the initial expectations, backshoring decisions are not significantly affected by bandwagon effects. In this respect, the decision of the focal firm to move production back to its home country seems the result of an individual choice rather than the consequence of previous decisions of neighbouring industry peers in the domestic region. There are two possible explanations of this result. First, experiential learning might become more relevant as the MSE starts operating abroad (Bruneel et al., 2010; De Clercq et al., 2012): according to Manning (2014), managers are not only motivated to mitigate operational challenges, but they also gradually develop the capability to address these issues over time. Mitigation capabilities are often developed semi-automatically through experience and problem-driven organizational learning (Zollo & Winter, 2002). Not infrequently, firms might find new and more effective sources of vicarious learning in the host country, exploiting ongoing interactions with foreign suppliers and other partner firms (Di Gregorio et al., 2009). Therefore, the beneficial effects associated with vicarious learning tend to become less relevant in the late stages of the internationalization process (Bruneel et al., 2010). The second possible explanation is related with the descriptive findings presented in Table 1. The data show that almost every offshore outsourcing firm in the considered sample maintains domestic subcontracting relationships when production is moved abroad. Additionally, backshoring decisions seem to be rarely associated with insourcing, contrary to the findings of other studies (Boffelli et al., 2020; Dachs et al., 2019): these patterns suggest that, in the clothing and footwear industry, most returning MSEs subcontract previously offshored production activities to domestic suppliers, possibly exploiting their existing subcontracting relationships. In this respect, we can argue that the decision to maintain domestic subcontracting relationships allows MSEs to take their backshoring decisions independently, rather than waiting for neighbouring backshoring peers to recreate the domestic subcontracting network in the home region (Baraldi et al., 2018).

Overall, the results of the investigation confirm that the motivations driving backshoring strategies among MSEs do not mirror those affecting offshore outsourcing (Di Mauro et al., 2018; Gray et al., 2017). More specifically, MSEs seem to address ex-ante perceptions of potential problems by observation, solving ex-post challenges through the experiential learning developed out of global sourcing activities. Therefore, the effect of direct experience and experiential learning versus vicarious learning seem to be different depending on the specific stages of the internationalization process (Lévesque et al., 2009). On the one hand, our findings on offshore outsourcing are consistent with previous evidence reported for other internationalization modes (Fernandes & Tang, 2014; Koenig et al., 2010; Xia et al., 2008), suggesting that MSEs rely on neighbouring domestic peers to engage in internationalization activities. In this respect, the capacity to exploit different learning sources may be a key factor influencing the ability of MSEs to overcome the liabilities of smallness and foreignness (Bruneel et al., 2010). On the other hand, the results of the backshoring analysis suggest that the decision-making process following offshore outsourcing is based on the actual performance of suppliers, and not on imitation of domestic industry peers.

This work has important practical implications from the standpoint of practitioners and policymakers. Our findings suggest that to be successful in offshore outsourcing, MSEs should secure access to external assets providing useful knowledge about potential foreign suppliers. However, relevant information about new partners should not necessarily be acquired using distant sources or costly national or international strategic alliances: entrepreneurs can also exploit their embeddedness in local networks, learning vicariously from small local peers located in close proximity and already engaging in the same activities. By making the best use of existing knowledge available in situ, managers can increase their propensity to anticipate the ‘hidden costs’ of offshore outsourcing (Larsen, 2016; Manning, 2014), moderating the risk of facing unexpected challenges in the short term. Policymakers can enhance these mechanisms, leveraging the role of local industrial associations and promoting both the dissemination of opportunities and the establishment of new formal and informal relationships inside the region (Costa et al., 2015). These measures can be combined with the provision of technical tools that could support MSEs in the development of more complete and less biased heuristics (Gray et al., 2017).

This empirical study has some limitations that may offer promising lines for future research. First, this study focuses on a specific sector, that is, the footwear and clothing industry, and therefore might suffer from some generalizability issues. Second, the structure of the database prevents the possibility to measure captive offshoring for MSEs and to detect offshore outsourcing among medium and large firms: therefore, it is not possible to assess whether interactions between larger firms and MSEs occur through the same internationalization mode. As far as captive offshoring is concerned, the data from AIDA show that this strategy is relatively uncommon among MSEs. This pattern is consistent with previous findings in the literature (Cutrini, 2011; Mariotti et al., 2008), confirming that the establishment of a subsidiary in a foreign country is typically not within the means of MSEs. The gap in the data is likely to be more problematic for offshore outsourcing, and a possible consequence is that bandwagon effects initiated by larger firms can be underestimated in the present analysis. Finally, the IMEFAS dataset does not permit to identify the nature of the interactions driving the presence of bandwagon effects among MSEs. More specifically, this database does not contain information on the direct ties between MSEs: therefore, one cannot assess whether mimetic isomorphism is associated with the mere observation of local early movers or if these dynamics are rather the result of conscious interactions between neighbouring actors. Furthermore, the role of other factors (such as employee-poaching, job-hopping or the emergence of intermediaries) in influencing international opportunity recognition among MSEs can only be assumed rather than directly demonstrated. A questionnaire survey among a sample of internationalizing MSEs might represent an effective tool to investigate the sources of mimetic behaviour through a more direct approach.

ACKNOWLEDGEMENT

The author thanks Sjoerd Beugelsdijk, Steven Brakman, Harry Garretsen, Nicoletta Giudice, Bart Los, Paolo Pavone, three anonymous referees and the participants at the annual conferences of the Regional Studies Association and the European Regional Science Association for helpful comments and discussions. The author is also grateful to the Italian Tax Revenue Agency for providing the data used for this research. The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of the agency.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author.

Notes

1. Given the firms in our sample always backshore from foreign suppliers, the focus of the paper is on outsource backshoring.

2. For more information on the AIDA BvD database, see https://aida.bvdinfo.com. A preliminary selection process was implemented on the dataset to restrict the analysis of captive offshoring to manufacturing subsidiaries of clothing and footwear producers: more specifically, information on the economic activity of the affiliate was extracted from the ORBIS BvD or retrieved from the internet when the subsidiary was not present in the main database.

3. These manufacturing industries are proxied in this empirical analysis by the sector studies D07A, D07B and D08U. For a list of ATECO (Classificazione delle Attività Economiche) codes associated with these sector studies, see the Italian Tax Revenue Agency at: https://www.agenziaentrate.gov.it/portale/archivio/archivio-studi-di-settore/modelli-comunicazione-annualita-pregresse/modelli-sds-2010/tabella-di-raccordo-ateco-2007.

4. In this investigation, four main destination regions are considered: Europe, North Africa, East Asia and other countries. The decision of using this territorial disaggregation is motivated by the nature of the data at our disposal. Indeed, information on offshore outsourcing activities is only available at this level of territorial detail.

5. A similar approach was proposed by Koenig et al. (2010) in their analysis of local spillovers in exporting activities.

6. The variable is computed by taking the average share of firms in the two years immediately preceding time t. The model also contains an additional variable to verify whether mimetic isomorphism also exists across different foreign destinations. The aim is to verify whether learning about industry peers’ production internationalization strategies in a particular world region helps the focal firm’s understanding of the opportunities to outsource to suppliers in a different foreign destination.

7. In this specification, when the average number of offshore outsourcing peers in the two years preceding t is < 2, the variables Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in the same foreign region and Number of micro and small peers located in the same domestic region and engaged in offshore outsourcing in other foreign regions are each assigned a value of 0.