Trust and Control in Franchise Networks: A Dyadic,Multi-Referent Analysis on Franchisee Network Exit Intentions

Theoretical Backgrounds

Research on franchisee exit intentions has typically focused on either a relational/social exchange perspective, revealing trust as an important predictor (Chiou et al., 2004; Wu, 2015), or an agency perspective, demonstrating that franchisees may leave when the franchisor’s rules or controls restrict them to reach their own goals (Croonen & Brand, 2015; López-Fernández & López-Bayon, 2018). However, this research falls short by (a) adopting a unilateral trust perspective, (b) considering only the franchisor as trust referent, and (c) ignoring the simultaneous inclusion of trust and control (see Supplemental Material #1 for an overview; all our Supplemental Materials can be found online under the Contents tab on the article page).

In dealing with these limitations, our theoretical framework integrates social exchange theory (SET), trust (in)congruence literature, and trust–control literature. SET helps in building our dyadic and multi-referent trust perspective. It suggests that business partners like franchisees and franchisors evaluate their relationships based on both economic and social outcomes, with a significant emphasis on trust as a social outcome (Blau, 1964; Cropanzano & Mitchell, 2005; Lioukas & Reuer, 2015). Trust reflects a partner’s positive expectations that the other partner will reciprocate by being willing and able to fulfill its obligations in the exchange relationship (Connelly et al., 2018). It reduces partners’ perceived exchange risks (Das & Teng, 2001, 2002), and helps partners to bond emotionally (Rowley et al., 2005). In our analysis, we consider franchisee–franchisor trust (in)congruence (representing direct exchange) and franchisee trust in peers (representing generalized exchange). Finally, we build upon the trust–control literature (Das & Teng, 1998, 2001; Long & Sitkin, 2018) and introduce franchisee perceived network control as a contextual factor influencing how franchisees interpret and are affected by both types of trust (Figure 1).

OPEN IN VIEWER

Figure 1.

Conceptual model.

The “Trust Congruence Effect”: Trust Congruence Versus Incongruence

Franchisees face risks in the direct exchange with the franchisor as a result of the mixed and potentially conflicting motives between them and the impossibility to fully eliminate such risks via formal contracts (Combs et al., 2004; López-Fernández & López-Bayon, 2018). This risk is exacerbated by the franchisor’s asymmetrical control, placing franchisees in a vulnerable position and increasing their dependency on the franchisor’s actions (Altinay et al., 2014; Davies et al., 2011). Franchisee trust in the franchisor is crucial to maintain the relationship by reducing franchisee perceived direct exchange risks via stronger goal alignment and a greater likelihood of goal attainment (Das & Teng, 2001; Davies et al., 2011). Prior franchising studies have already shown that if franchisees trust their franchisor to be willing and able to fulfill its exchange obligations, they are more likely to anticipate positive future exchange outcomes and are more likely to stay in the franchise network (Chiou et al., 2004; Croonen & Brand, 2015; Wu, 2015). However, as argued, these studies have adopted a unilateral rather than a dyadic trust perspective and cannot account for the possible effects of trust (in)congruence.

Studies in other organizational fields have argued or found that trust congruence between partners—where trust levels are aligned—strongly influences relational behaviors and/or outcomes (Korsgaard et al., 2015; Tomlinson et al., 2009; L. Wang et al., 2023). Trust congruence enhances compatible views among partners on the nature of their exchange and how to behave in the relationship, regarding resource investments, risk taking, knowledge sharing, or communication routines (Korsgaard et al., 2015; L. Wang et al., 2023). This compatibility results in higher predictability and lower perceived exchange risks, even when the congruence results from partners both having low trust levels. In such a low trust environment, the exchange of resources becomes strictly transactional, focusing on the immediate utility and compensation rather than long-term relational synergy (L. Wang et al., 2023). Using a transaction-oriented approach, the franchisor and franchisee can then still sustain the relationship by ensuring that both of them adhere to agreed-upon standards and deliverables (Bradach & Eccles, 1989). Due to their low trust levels, partners are mutually cautious, but they understand where each of them is coming from (Tomlinson et al., 2009), enabling them to work better together than in case of trust incongruence; that is, when there is a difference in trust levels among the partners.

Trust incongruence leads to incompatible views among partners on the nature of their exchange. The unaligned expectations regarding desirable relationship behaviors (L. Wang et al., 2023) can lead to miscommunications, misconceptions, and conflicts (Korsgaard et al., 2015; Tomlinson et al., 2009) that destabilize the relationship. Trust incongruence can cause confusion and may lead to one party perceiving the other’s actions as too forward, invasive, or not committed enough, depending on the direction of the incongruence (L. Wang et al., 2023). For example, excessive trust from the franchisor might lead a franchisee to feel overwhelmed by unrealistic expectations, leading to a phenomenon known as “role overload” where the franchisee fears the potential consequences of failing to meet these expectations (cf. Tang & Vandenberghe, 2021). Conversely, when the franchisee bestows more trust in the franchisor than vice versa, this may also lead to unaligned expectations that destabilize the franchise relationship. A high-trusting franchisee may be more willing to take risks, such as engaging in innovation (Hadjielias et al., 2021) or opening additional units (Mignonac et al., 2015), based on the expectation of ongoing support and collaboration from the franchisor. However, if the franchisor lacks trust in the franchisee, it may be hesitant to take these risks and/or refrain from the necessary support. Again, this incongruence may lead to feelings of discomfort as the franchisee may feel unsupported or misled, whereas the franchisor may feel pressured and uneasy about the franchisee’s ambitious expectations, which may lead to a strained franchise relationship. Ultimately, when there is a mismatch in trust levels, misconceptions happen because parties infer a narrative based on how strongly they trust the other party, which may not align with the reality of the other party’s commitment, reliability, or openness (Tomlinson et al., 2009; Korsgaard et al., 2015). Each party may feel misunderstood or undervalued, and these feelings can escalate into conflicts, which might not only damage the relationship but also impede productive collaboration and the pursuit of mutual goals.

In sum, we propose that trust congruence promotes compatible views on risk taking, knowledge sharing, and communication in the relationship. Conversely, trust incongruence complicates relationships by making it difficult for both parties to develop a shared understanding and agree on common goals, leading to a less effective and more arduous relationship. This principle, where relational outcomes are more positive when trust is congruent than when it is incongruent, is referred to as the “congruence effect” (Humberg et al. 2019; L. Wang et al., 2023). Hence:

Moving on the Line of Congruence: From Low-Low to High-High Trust

In Hypothesis 1, we propose that trust congruence leads to lower exit intentions. However, we also posit that the effect of trust congruence on exit intentions varies according to the level of trust: it makes a difference whether partners are congruent because they both have low, medium, or high levels of trust on the line of congruence (LOC; for a discussion, see Tomlinson et al., 2009; L. Wang et al., 2023). In other words, the outcome (i.e., exit intention) depends on the level of the two examined variables (Edwards, 2007). As explained above, congruence at low trust levels (i.e., low-low trust) leads to very transactional interactions where partners are mutually cautious, resulting in low levels of resource investments, risk taking, knowledge sharing, and communication. Conversely, congruence at high trust levels (i.e., high-high trust) creates a fertile ground for positive interactions through shared norms, reciprocity, streamlined communication routines, and deeper comprehension of each other’s needs (Korsgaard et al., 2015). Hence, congruence at high trust levels encourages partners to take more risks, exchange vital instead of basic knowledge, and to work collaboratively toward achieving joint objectives when compared with congruence at lower trust levels (Tomlinson et al., 2009; L. Wang et al., 2023). Congruence at higher trust levels fosters a more committed investment in one another, setting the stage for more durable relationships that offer long-term benefits to both sides (Ferrin et al., 2008). While the congruence mechanisms still allow both parties to reach agreement regarding resource commitments, risk taking, knowledge exchange, and communication at low trust levels, we expect that trust congruence at higher levels of trust will enhance the franchisee’s expectation of positive outcomes from its network membership, increase commitment to strive for mutual benefits, and thereby reduce network exit intentions.

We also posit that the relationship between trust congruence and exit intentions is not linear, implying that the impact of trust increments on the LOC (i.e., the simultaneous and equal strengthening of trust among the franchisor and franchisee) on relationship outcomes may increase or decrease at different rates. Several trust studies suggest that positive relationship outcomes may be amplified, when two parties, who equally trust each other, also do so at higher levels of trust (Korsgaard et al., 2015), as trust congruence increments not only reduce risks (via greater predictability) but also foster the strengthening the positive side of relationships (via goodwill) (Hardy et al., 1998). The high trust setting allows for high mutual investments and an in-depth understanding of what the other party needs, providing a greater potential to reap relational rewards and enduring relationships. Applied to a franchise context, these findings suggest a nonlinear relationship in which exit intentions decrease more sharply for trust congruence increments at high trust levels. Hence:

Effect of Franchisee Trust in Peers

When entering a franchise network, franchisees inevitably become subject to a generalized exchange with their network peers. This means they engage in an interconnected network of indirect, multiparty relationships that benefit each franchisee not necessarily through direct quid-pro-quo transactions but through a web of mutual dependencies and collective value creation (cf. Das & Teng, 2002; Heidl et al., 2014; Li et al., 2012). Perceived generalized exchange risks arise in this setting because franchisees’ constructive or destructive behaviors impact the network (Combs et al., 2004; Kidwell & Nygaard, 2011). Since franchisor network control is inherently imperfect (Kashyap et al., 2012), franchisees must trust their peers to meet exchange obligations and adhere to franchisor guidelines and norms.

Despite previous studies on outcomes of franchisee peer cohesion or peer interactions (Brand et al., 2018; El Akremi et al., 2011; Hadjielias et al., 2021), the specific effects of franchisee peer trust on exit intentions have not been investigated. We hypothesize that peer trust reduces exit intentions because of a reduction of perceived risks of peer opportunistic behaviors, increased peer helping and knowledge sharing, and a cultivation of shared values and sense of community. We explain the three underlying arguments for our hypothesis below.

First, an increase in peer trust results in lower perceived risks of peer opportunistic behaviors, such as non-compliance or free-riding. Peer trust promotes compliance with the group’s behavioral norms through affective bonds (Cochet et al., 2008; Kidwell et al., 2007). When franchisees trust their peers, they expect them to comply with the franchisor’s guidelines, thereby reducing the expectancy that peers damage the network’s brand reputation (El Akremi et al., 2011; Kidwell et al., 2007).

Second, peer trust facilitates helping behaviors and knowledge sharing in franchise networks (El Akremi et al., 2011; Hadjielias et al., 2021). With greater peer trust, franchisees are more likely to provide and receive emotional support, share honest feedback, flag potential issues, and collaboratively tackle problems, thereby reducing the likelihood of problems and inefficiencies. High peer trust also facilitates open communication channels in which franchisees share knowledge that increase franchisees’ learning and expectations regarding the achievement of positive outcomes for their units from their network membership (Brand et al., 2018), which reduces their exit intentions.

Third, peer trust creates robust relationships based on shared values and interests, a psychologically safe environment, and engagement with the franchise network. Enhanced peer trust serves as a catalyst for an effective generalized exchange system by developing a collective sense of cohesion and brand community (El Akremi et al., 2011) in which franchisees do not expect to directly benefit from aiding another franchisee, but do so because they want to contribute to the network’s strength and operational efficiency. Franchisees might benefit from these positive peer behaviors via the generated positive word-of-mouth of their peers’ satisfied clients, or the increased brand reputation fueled by local advertising efforts from their peers. Hence:

Moderating Effects of Perceived Network Control

Building on trust–control literature, we hypothesize that franchisee perceived network control is a critical contextual factor that moderates the effects of trust congruence (H2) and peer trust (H3) on exit intentions. Existing studies highlight the significance of control in mitigating generalized exchange risks (Das & Teng, 2002; Heidl et al., 2014; Li et al., 2012) that manifest in forms like franchisee free-riding or non-compliance (Kashyap et al., 2012; Kidwell & Nygaard, 2011). These risks compel franchisors to implement control mechanisms to ensure consistent quality and uphold the network’s brand reputation (Kidwell et al., 2007). Franchisors typically carry out such control through establishing a formal contract and through the monitoring of franchisee behaviors, for example through regular field visits and audits (Antia & Frazier, 2001; Dada, 2018; Kashyap et al., 2012). In our focal franchise network, where formal contracts show minimal variation, we investigate franchisee perceived network control through franchisee perceived monitoring by the franchisor. Franchisees—even within the same network—may differ in terms of how they interpret franchisor network control depending on their own traits, experiences, and market circumstances (Cochet et al., 2008; Dada, 2018). Hence, we hypothesize that franchisee control interpretations vary across direct and generalized exchanges, that is, franchisee perceived control moderates the effects of franchisee–franchisor trust congruence and franchisee peer trust on exit intentions in distinct ways.

We hypothesize that franchisee perceived network control intensifies the negative relationship between trust congruence and exit intentions. In situations of high perceived control, franchisees can strongly benefit from the mutual understanding and alignment of objectives provided by trust congruence; trust congruence increments strongly reduce their exit intentions because control helps to more effectively reduce generalized exchange risks and deepen the franchisee–franchisor relationship. This amplified effect is rooted in the franchisee’s interpretation of control as a means to effectively reduce generalized exchange risks. A high perceived level of network control tends to be associated with lower performance network variability (cf. Das & Teng, 1998, 2001) via enabling quick identification of deviations and anomalies, and allowing timely corrections to prevent small problems from becoming major setbacks. The perception of frequent monitoring not only creates a secure and predictable operating environment but also acts as a safeguard for maintaining quality standards, serves as an early warning system, and ensures efficient utilization of resources shared with the franchisor (cf. Kidwell & Nygaard, 2011).

Additionally, in a high control setting, franchisees may perceive frequent franchisor interactions through field visits and audits as a form of deepening their relationship with the franchisor, rather than as mere compliance checks. This elevated level of control creates enhanced opportunities for mutual benefit and goal alignment (Bradach, 1997). Frequent interactions allow the franchisor to identify, communicate, and address the timely needs of individual franchisees in terms of financing, operational assistance, and growth opportunities. These interactions contribute to mutual understanding and a synergistic relationship where both parties can work toward shared objectives (Dada, 2018; Davies et al., 2011) and steers future value co-creation initiatives, in which both the franchisor and franchisee seek to share their valuable resources to improve the functioning of the franchise network. When high perceived network control is coupled with congruence at high trust levels, it creates a collaborative atmosphere that nurtures value creation and aligns objectives, making trust congruence particularly effective in reducing exit intentions due to enhanced relationship depth facilitated by franchisor monitoring. Hence:

Franchisee peer trust and perceived network control both serve to mitigate perceived generalized exchange risks within the network by discouraging opportunistic behavior, enhancing peer cooperation, and promoting peer knowledge sharing (El Akremi et al., 2011; Hadjielias et al., 2021). Yet, we posit that franchisee (horizontal) peer trust and the franchisor’s (vertical) network control are to a certain extent incompatible, such that their simultaneous presence creates antagonistic effects. Although peer trust leads to lower perceived risks of peer opportunistic behaviors, increased peer cooperation, and a cultivation of shared values and sense of community, strong franchisor network control—aimed hierarchically, unilaterally, and formally at reducing perceived generalized exchange risks—may diminish the positive impacts of peer trust, as it may create a wedge between the franchisor and franchisees. Franchisees, as independent entrepreneurs, are typically critical toward the control imposed on them (Dada, 2018; Heide et al., 2007). Enhanced peer trust is particularly effective in reducing perceived exchange risks in low control contexts, as it promotes strong bonds, shared values, and effective social control among franchisees (El Akremi et al., 2011). Conversely, in high-control situations, increased peer trust would make franchisees interpret the franchisor’s control activities as redundant, inappropriate, costly, and even excessive (Heidl et al., 2014), which may lead to their psychological reactance and noncooperative reactions (Heide et al., 2007). In such a scenario, franchisees may communicate their grievances about the franchisor’s strong but ineffective network control with other trusted peers creating an “us-vs-them” culture in the network that may ultimately lead to coalition formation and, in extreme cases, even lead to collective network exits (cf. Emerson, 1962). Hence:

Research Context, Sample, and Data Collection

To enhance internal validity (Davies et al., 2011), we collected data within a single franchise network (pseudonym: @Home), which provides home services to Dutch customers. The lion’s share of @Home franchisees worked alone or with very few employees. At the time of data collection (late 2011/early 2012), the @Home network had 209 franchisees—only male franchisees operating single units—who often knew each other personally through joint meetings and/or projects. The franchisor’s office included a management team of 5, led by a single owner-manager. The @Home network equipped its franchisees with a recognized brand, consolidated marketing, training, mentorship, and additional services like insurance and guarantee schemes. The franchisees also adhered to a uniform presentation style, including professional uniforms and company-branded vans. Back-office operations were less regulated, with no centralized procurement or logistics department, although @Home recommended specific suppliers that offered franchisees a standard discount.

The management team maintained regular contact with each individual franchisee through field visits to discuss performance, compliance, and challenges. Furthermore, the franchisor held annual network events and regional meetings to discuss updates and gather franchisee feedback.

To ensure honest and representative responses we: (a) cooperated with the network’s Franchise Advisory Council; (b) guaranteed confidentiality; (c) visited eight regional member meetings to establish respondents’ trust; (d) provided a small incentive to respondents; and (e) sent a paper-and-pencil survey to the personal addresses of the franchisees with a prepaid reply envelope administrated by our university. We sent reminders via email and phone calls.

We invited all franchisees to participate in the survey, receiving 135 responses, with 120 complete surveys. The net response rate (57%) is high, and helped to ensure the representativeness of our sample compared to the network’s franchisee population. In the same time period, during a meeting at the headquarters, the six members of @HOME’s management team jointly determined their trust in each franchisee, resulting in data on 120 franchisee–franchisor dyads. For a post hoc analysis, we also collected archival data from 2015 to analyze actual franchisee exits, and we conducted an interview with the network’s owner-manager to understand the reasons behind these actual exits (i.e., whether they were voluntary or involuntary).

Measures and Measurement Properties

We employed validated scales from earlier studies for all the constructs, adapting the items to fit our research context (see Appendix 1 for the full list of items). Unless specified otherwise, all items were measured on 5-point Likert scales.

Pre-Analysis Checks

We assessed the validity and reliability of our constructs using variance-based Structural Equation Modeling (SEM) utilizing SmartPLS 4 (Ringle et al., 2020) (see Appendix 1). All scales demonstrate convergent validity, as they have high standardized loadings. Furthermore, —with one exception of 0.48 for one of our control variables—the average variance extracted (AVE) of each construct is above the recommended threshold of 0.5 (Fornell & Larcker, 1981). There is evidence of discriminant validity in that the square root of the AVE is higher than each of the underlying correlations with the other constructs. Finally, based on the Cronbach’s alphas and composite reliabilities, the scales were deemed reliable. Table 1 outlines the descriptives and the correlation matrix.

OPEN IN VIEWER

Table 1.

Correlation Matrix and Discriminant Test.

Variables	Mean	SD	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
(1) Network exit intentions	2.513	1.090	(.831)
(2) Franchisee trust in franchisor	3.315	0.806	−.719***	(.794)
(3) Franchisee trust in peers	3.477	0.795	−.184	.315*	(.755)
(4) Franchisor trust in franchisee	3.432	0.670	−.458***	.395***	.013	(.812)
(5) Perceived network control	3.278	0.827	−.399***	.490***	.487***	.137	(.812)
(6) Economic satisfaction	3.575	0.793	−.311***	.265**	−.136	.277**	.120	(.800)
(7) Relationship duration	3.076	1.930	.076	−.153 Ψ	−.077	.125	−.069	.135	n.a.
(8) Alternatives	3.106	0.813	.725***	−.579***	−.222 Ψ	−.295***	−.406***	−.261**	.125	(.748)
(9) Switching costs	3.020	1.073	−.539***	.427***	.347**	.074	.351***	−.037	−.195*	−.510***	(.877)
(10) Strategic management	3.470	0.700	−.498***	.564***	.276*	.243**	.510***	.301***	−.061	−.478***	.396***	(.693)
(11) Quality assurance	3.050	0.861	−.466***	.656***	.491**	.230*	.561***	.076	−.234*	−.324***	.381***	.500***	(.825)
(12) ICT support	2.840	0.931	−.198*	.249*	.073	.193*	.117	.166 Ψ	−.082	−.109	.139	.207*	.399***	(.825)
(13) Industry experience	10.730	10.131	.012	−.078	−.112	−.098	−.059	−.043	−.106	−.011	.006	−.027	−.025	.013	n.a.

Note. N = 120. Means and standard deviations result from the unweighted, unstandardized composite scores of the constructs. The correlations between the latent constructs are based on PLS results with 5,000 bootstraps. The diagonal shows the square roots of the AVEs. Correlations in bold are the proxies for common method variance (see Supplemental Material #2). For the purpose of interpretability, the non-transformed means and standard deviations for all independent variables and control variables are shown. AVE = average variance extracted; n.a. = means not applicable; SEE = franchisee; SOR = franchisor.

^Ψp < .10, *p < .05, **p < .01, ***p < .001 based on two-tailed tests.

To assess possible non-response bias, we compare early and late respondents and find no significant differences regarding distance to the franchisor headquarters in both kilometers (M_early = 91.6; M_late = 90.7, t = 0.119, p = .475) and travel time in minutes (M_early = 66.4; M_late = 66.4, t = 0.005, p = .694), or in network exit intentions (M_early = 2.5; M_late = 2.6, t = −0.506, p = .971), which indicates that non-response bias is unlikely to be a problem. The late respondents tend to be slightly younger (M_early = 46.6; M_late = 43.8, t = 1.853, p = .026). Comparing the non-response with the response sample, we find no differences in terms of distance in kilometers (M_resp = 91.2; M_nonresp = 89.8, t = 0.227, p = .463), travel time (M_resp = 66.4; M_nonresp = 64.3, t = 0.555, p = .496), and age (M_resp = 45.2; M_nonresp = 43.8, t = 1.205, p = .535).

To reduce the danger of common method bias in the survey data among bivariate relationships, we followed Podsakoff et al.’s (2003) recommendations regarding survey design; we used reversed-coded items and prevented funneling in the survey. We also used multiple informants (franchisor and franchisees) to mitigate single-informant bias (Podsakoff & Organ, 1986). To assess the extent of common method variance (CMV), we used three tests. First, we assessed CMV using a full collinearity test (Kock, 2015). The full collinearity inflation factors, calculated using SmartPLS, were all below the 3.3 threshold. Second, we conducted a marker-variable test (Lindell & Withney, 2001), using the smallest (r = .006) and second smallest (r = .012) positive correlations as proxies for CMV. Results (see Supplemental Material #2) show that all partial correlations remained statistically significant, suggesting that CMV does not drive our results. Third, we assessed how much variance in the items was accounted for by a single latent CMV factor versus the respective constructs (Liang et al., 2007). The common factor accounted for less than 2% of the variance on average, whereas the constructs explained about 65%. Only 5 of the 43 paths from the CMV factor were significant. These tests collectively indicate that CMV is not a major concern.

Analysis Method

We test our hypotheses using PR and RSA, which models the effects of trust (in)congruence on relational outcomes (Edwards, 2002). Our goal was to assess if exit intentions are lower when franchisee and franchisor trust levels are congruent, as opposed to when they are incongruent (H1), following the criteria set by Humberg et al. (2019) for testing the acceptance (or rejection) of “the congruence effect.” To ascertain the effect of franchisee–franchisor trust congruence on exit intentions (H2), we rely on the coefficients of Model 2 in Table 2. Using the PR equations, we graphically plot these relationships and employ the RSA to test various features of the effect of trust congruence on franchisee exit intentions (Edwards, 2002; Edwards & Parry, 1993). To explore how trust congruence affects exit intentions under varying levels of perceived network control (H4), we categorized our data into low and high control groups using a median split (N = 51 and N = 48, respectively). ¹ Following prior research (L. Wang et al., 2023), we then ran Model 2 for both groups. Additionally, we examined the relationship between franchisee peer trust and exit intentions (H3) (Model 3, Table 2), and included an interaction term to assess the moderation effect of perceived network control (H5) (Model 3, Table 2).

OPEN IN VIEWER

Table 2.

Polynomial Regression Results.

Dependent Variable: Franchisee NEI	Coefficient	Model 0	Model 1	Model 2	Model 3
Intercept	b 0	2.881***(.730)(.000)	2.514***(.653)(.000)	2.470***(.653)(.000)	2.404***(.656)(.000)
Economic satisfaction (ES)		−.225*(.091)(.015)	−.110(.083)(.187)	−.124(.081)(.127)	−.115(.080)(.151)
Relationship duration (RD)		−.040(.035)(.255)	−.025(.032)(.430)	−.039(.031)(.217)	−.036(.031)(.251)
Alternatives (ALT)		.676***(.101)(.000)	.482***(.097)(.000)	.539***(.097)(.000)	.519***(.097)(.000)
Switching costs (SC)		−.169*(.074)(.025)	−.181**(.066)(.007)	−.203**(.065)(.002)	−.189**(.064)(.004)
Strategic management (SM)		−.030(.123)(.805)	.018(.110)(.866)	.031(.108)(.776)	.007(.108)(.946)
Quality assurance (QA)		−.312**(.096)(.002)	−.129(.105)(.221)	−.108(.101)(.288)	−.093(.104)(.375)
ICT support (IS)		.001(.077)(.986)	.021(.069)(.756)	−.027(.069)(.691)	−.007(.068)(.922)
Industry experience (IE)		.002(.006)(.697)	−.001(.006)(.903)	.000(.006)(.972)	−.000(.005)(.965)
Franchisee trust in peers (ETP)			.140Ψ(.084)(.098)	.114(.082)(.166)	.078(.085)(.365)
Franchisee trust in franchisor (ETO)	b 1		−.446***(.115)(.000)	−.572***(.118)(.000)	−.589***(.117)(.000)
Franchisor trust in franchisee (OTE)	b 2		−.298**(.097)(.003)	−.213*(.097)(.030)	−.230*(.097)(.019)
ETO2	b 3			.079(.080)(.324)	.032(.084)(.703)
ETO × OTE	b 4			.298*(.141)(.037)	.300*(.139)(.033)
OTE2	b 5			−.148(.104)(.158)	−.135(.104)(.196)
Perceived network control (PNC)					−.019(.102)(.856)
ETP × PNC					.184*(.083)(.030)
Lowest variance inflation factor		1.02	1.05	1.05	1.06
Highest variance inflation factor		1.69	2.66	3.14	3.14
Mean variance inflation factor		1.39	1.63	1.84	1.92
R 2 adjusted		.586***	.678***	.700***	.709***
ΔR2		—	.092***	.022**	.009*

Note. N = 120. Below the unstandardized regression coefficients, we present the standard errors and exact significance level (in italics) in parentheses. In terms of relative variance explained (ΔR²), each model is compared with its prior model. NEI = network exit intentions.

Unstandardized regression coefficients are shown with significance levels: ^Ψp < .10, *p < .05, **p < .01, ***p < .001.

Given the sensitivity of PR to multicollinearity and outliers, particularly with the inclusion of interaction and quadratic terms (Edwards, 2002), we scale-centered the variables before computing these terms. This step reduces the potential for multicollinearity and aids in interpreting response surface plots (Aiken et al., 1991). We also checked for influential outliers using Cook’s D and standardized residuals generated by the PR formulas (Bollen & Jackman, 1990). Although five potential outliers were identified, their exclusion did not significantly influence the results, leading us to include the complete dataset in our reported analysis.

Following Edwards and Parry (1993), we advanced from simple to complex regression models, favoring those that explained more variance significantly. We employed conventional PR equations (while including control variables) to represent our theoretical model (Edwards & Parry, 1993):

Z = b 0 + b 1 X + b 2 Y + e

(1)

Z = b 0 + b 1 X + b 2 Y + b 3 X 2 + b 4 XY + b 5 Y 2 + e

(2)

where Z = network exit intentions, X = franchisee trust in franchisor, and Y = franchisor trust in franchisee, and e = error.

The polynomial terms in equation (2) help calculate all elements of the response surface plot (slopes and curvatures of the lines of congruence and incongruence, as well as principal axes), which are needed to test the hypothesized congruence effect. Humberg et al. (2019) noted that these elements should not be interpreted separately, but alongside other elements of the RSA.

The beta coefficients resulting from equation (2) form the basis for the response surface plot, from which several key parameters can be derived. To identify the effect of franchisee–franchisee trust congruence, we are interested in parameters related to the LOC, along which the values of X and Y are equal (X = Y). To investigate the effect of trust incongruence, we look at increments on the line of incongruence (LOIC), where the values for X and Y differ (X = −Y). The key parameters regarding these lines are: the slope of the surface along the LOC (a₁, resulting from b₁ + b₂); the curvature of the surface along the LOC (a₂, resulting from b₃ + b₄ + b₅); the slope of the surface along the LOIC (a₃, resulting from b₁ − b₂), and the curvature of the surface along the LOIC (a₄, resulting from b₃ − b₄ + b₅).

Table 2 shows the results of the regression analyses predicting exit intentions. In Model 0, we insert our control variables. In Model 1, we test equation (1). As the results for Model 1 show, the linear terms for the three trust variables and the control variables explain 67.8% of the variance in exit intentions. Model 2 introduces the interaction and quadratic terms, as shown in equation (2), and explains statistically greater variance in exit intentions than Model 1 (Adj. R² = .700, ΔR² = .022, p = .009). Following Edwards (2002), we discard the linear model in favor of the quadratic model. Model 2 forms the basis of our analyses of Hypotheses 1, 2, and 4. Table 3 presents the coefficients for the response surface features for this model. Figure 2a shows the accompanying response surface plot for the entire sample. Figure 2b and 2c depict the cross-sections of Figure 2a, highlighting the LOIC and the LOC. Table 3 also presents the coefficients for the response surface features of Model 2 under low and high franchisee perceived network control, forming the basis for the response surface plots in Figure 3a and 3b. For Hypotheses 3 and 5, Model 3 in Table 2 incorporates the perceived network control variable and its interaction with peer trust. This model explains more variance in exit intentions (Adj. R² = .709, ΔR² = .009, p = .046), underscoring the importance of this interaction term.

OPEN IN VIEWER

Table 3.

Response Surface Features.

Models	Surface along the LOC (X = Y)		Surface along the LOIC (X = −Y)
	Linear slope a (a1)	Curvature (a2)	Linear slope b (a3)	Curvature (a4)
Model 2: Entire sample c	−.785***(.132)(.000)	.229*(.105)(.032)	−.359*(.171)(.038)	−.367(.266)(.171)
Model 2: Low perceived network control d	−.573*(.249)(.026)	.216(.268)(.424)	−.261(.272)(.342)	−.458(.363)(.213)
Model 2: High perceived network control d	−1.404**(.501)(.007)	.584(.513)(.260)	−.158(.513)(.759)	−.388(.568)(.498)

Note. LOC = line of congruence; LOIC = line of incongruence.

a

The linear slope of the response surface along the LOC is given by a₁ = (b₁ + b₂), where b₁ is the beta coefficient for franchisee trust in franchisor (ETO), and b₂ is the beta coefficient for franchisor trust in franchisee (OTE). The curvature of the slope along this line is given by a₂ = (b₃ + b₄ + b₅), where b₃ is beta coefficient for ETO², b₄ is the beta coefficient for the cross-product (ETO × OTE), b₅ is the beta coefficient for OTE².

b

The linear slope of the response surface along the LOIC is given by a₃ = (b₁ − b₂) and the curvature of the slope along this line is given by a₄ = (b₃ − b₄ + b₅).

c

N = 120 for the entire sample. Unstandardized regression coefficients, standard errors (in parentheses), and exact significance level (in italics and in parentheses) are shown using the following significance levels: ^Ψp < .10, *p < .05, **p < .01, ***p < .001. Significance levels of estimated coefficients are retrieved via inserting the corresponding coefficients of Model 2 (entire sample and after sample split) in the Excel file of Shanock et al. (2010).

d

We used a median split to group our data into a group of franchisees with values below the median (low perceived network control, N = 51) and above the median (high perceived network control, N = 48).

OPEN IN VIEWER

Figure 2.

(a) Response surface plot (entire sample); (b) Curvature for the LOIC; (c) Curvature for the LOC.

The “Trust Congruence Effect”

Regarding the “trust congruence effect” (H1), we follow the steps of Humberg et al. (2019) who delineate the conditions for testing the congruence effect. The first condition generally requires that the LOIC follows an inverted U, but because our DV represents a negative outcome, we should refer to the inverse: a U-curvature (Edwards & Parry, 1993). Our results should therefore demonstrate a positive and significant curvature on the LOIC, such that exit intentions increase when values for franchisor trust in the franchisee and franchisee trust in the franchisor deviate from each other in either direction (Baer et al., 2021; Humberg et al., 2019). In contrast with this condition, the curvature along the LOIC (X = −Y) was insignificant (a₄ [b₃ − b₄ + b₅] = −.367, p = .171, Table 3), suggesting that there is no U-shaped relationship wherein the values of exit intentions are higher when values deviate from congruence levels of trust.

In testing for the congruence effect, we consider the second principal axis instead of the first principal axis, because we test for the “reverse” congruence hypothesis that hypothesizes that network exit intentions are lower in the case of trust congruence compared to trust incongruence (Humberg et al., 2019). Upon further inspection of this second principal axis, we generated 10,000 bootstrapped samples to estimate 95% confidence intervals (CIs) for the intercept (p₂₀) and slope (p₂₁), following the methods of Edwards (2002) and Edwards & Parry (1993). For the congruence effect to exist, the CI for the intercept of the second principal axis should include zero, and the CI for the slope should include one. Our findings, however, reveal that the 95% CI for the intercept (p₂₀) excludes zero (2.118, 12.616) and that the slope (p₂₁) excludes one (−5.014, −1.011), disproving the congruence effect (Humberg et al., 2019). Hence, we reject Hypothesis 1 and conclude that there is no evidence for the “trust congruence effect.”

The absence of the “trust congruence effect” indicates that trust incongruence may not necessarily produce higher exit intentions. Surprisingly, we observed that the slope of the LOIC displays a significant negative linear association with exit intentions (a₃ = −.359, p = .038, Table 3), suggesting that, on average, higher levels of trust incongruence between the franchisor and the franchisee are associated with lower exit intentions. This somewhat counterintuitive result can be explained by the shape of the surface plot. Incongruence can refer to either negative or positive discrepancy (Edwards, 2008). In our findings, negative discrepancy means that franchisee trust in franchisor (X) falls short of franchisor trust in franchisee (Y), such that X < Y, whereas positive discrepancy means that X exceeds Y. An investigation of the LOIC (Figure 2b) reveals that incongruence does not enhance but reduces exit intentions—especially in case of positive discrepancy. Hence, when franchisee trust in the franchisor significantly exceeds the franchisor’s trust in franchisee, exit intentions tend to diminish. This finding challenges the traditional view of trust incongruence, and highlights the need to consider the direction of incongruence in future studies.

Effects of Trust Congruence for Low and High Franchisee Perceived Network Control

Next, we test Hypotheses 2 and 4, which investigate the effects of trust congruence on exit intentions (H2) and the moderating effect of perceived control on this relationship (H4). In line with Hypothesis 2, we find that trust congruence increments strongly reduce exit intentions (a₁ = −.785, p = .000, Table 3), such that exit intentions decrease when franchisor and franchisee trust are congruent at higher trust levels than at lower trust levels. While the significance of a₂ in Table 3 indicates that this negative association on the LOC is indeed nonlinear, the positive and significant curvature of (a₂ = .229, p = .032) and Figure 2c suggest that the reduction of exit intentions plateaus for higher values of trust on the LOC. Thus, Hypothesis 2 is only partially supported.

Hypothesis 4 predicts that perceived network control strengthens the relationship between trust congruence and exit intentions. Analyzing split samples, we find a more pronounced negative linear slope on the LOC for the high control (a_1high = −1.404, p = .007, Table 3) than for the low control sample (a_1low = −.573, p = .026). Furthermore, the response surface plots for low (Figure 3a) and high (Figure 3b) values of perceived control show that the slope is steeper, and more negative, along the LOC, supporting Hypothesis 4. Additionally, the insignificance of the curvature values for both groups (a_2low = .216, p = .424; a_2high = .584, p = .260) in Table 3 suggests that these effects are linear.

OPEN IN VIEWER

Figure 3.

(a) Low perceived network control; (b) High perceived network control.

Effects of Trust in Peers for Low and High Franchisee Perceived Network Control

Hypotheses 3 and 5 examine the effect of franchisee peer trust (H3) and its interaction with perceived network control (H5). Model 3 in Table 2 reveals no support for Hypothesis 3, which anticipated a simple linear negative effect of peer trust on exit intentions (B = .078, p = .365). However, for Hypothesis 5, the results confirm for the predicted interaction effect between peer trust and perceived control (B = .184, p = .030). Simple slope analyses demonstrate that peer trust positively affects under high control (B = .262 p = .018), but this effect is not significant under low control (B = −.106, p = .404), as depicted in Figure 4. Hence, peer trust only significantly influences exit intentions when perceived control is high.

OPEN IN VIEWER

Figure 4.

Interaction plot for franchisee trust in peers with perceived network control.

Robustness Checks and Post hoc Analyses

Theoretical Contributions and Implications

Practical Implications: Forging Bonds and Divide and Conquer

Our findings lead to two key practical implications at the dyadic and the network level. First, for the franchisee–franchisor dyad, our study highlights that franchisors can strongly reduce franchisees’ exit intentions by “forging bonds.” It is particularly effective for a franchisor to prioritize increasing franchisees’ trust by demonstrating both ability—by fulfilment of its strategic and operational duties to develop a robust business format—and goodwill—for example through openness in communication (Altinay et al., 2014; Croonen & Broekhuizen, 2019). Yet, since we found that trust-congruence increments display marginal returns, franchisees should be wary to not overinvest in improving franchisees’ trust for franchisees who already strongly trust them.

Second, from a network perspective, given the antagonistic effects we found between franchisee peer trust and perceived network control, franchisors should adjust their network control (i.e., monitoring) behaviors according to the level of franchisee peer trust. One simple way to reduce exit intentions is to adjust the level of monitoring; increases (decreases) in peer trust should be matched with lower (higher) monitoring activity to limit exit intentions, as long as the reduced monitoring does not lead to adverse effects regarding the network’s output quality and brand reputation. Another way to reduce exit intentions is to lower peer trust levels within the network. To counter the potential adverse effects of high levels of peer trust, franchisors may consider adopting a “divide and conquer” strategy to limit peer trust among their franchisees. However, the pursuit of divide and conquer can be questioned from an ethical point of view, as it refers to leaders deliberately generating or exploiting social divisions among members (Case & Maner, 2014). Such a strategy should therefore be carefully deployed and balanced against other franchise network benefits, since its pursuit may lower franchisee trust in the franchisor and reduce the positive effects of peer trust, such as peer learning and peer cooperation. Franchisors should thus carefully balance peer trust and network control by, for example, deploying “tournament strategies” (Gillis et al., 2011), which can boost franchisees’ innovativeness and output quality, while increasing peer competition.

Building on these dyadic and network perspectives, we argue that franchisors should engage in both strategies to prevent franchisee network exits, that is: “forging bonds” with individual franchisees in the direct exchange and“divide and conquer” to manage the generalized exchanges among peers.

Limitations and Further Research

Several limitations and future research directions are noteworthy. First, focusing on a single franchise network enhances internal validity by controlling for industry and network differences (Davies et al., 2011). However, it limits generalizability to other networks with varying network characteristics, such as centralization levels (Mumdžiev & Windsperger, 2011), or types of unit ownership (Bradach, 1997). Further research should examine the degree to which our findings apply across various franchise networks. Second, our data from 2011/2012, may not account for the impact of recent AI advancements and increased market data availability, altering how franchisors control and assess trust for individual franchisees. Future research could explore how this more customizable, technology- and data-driven franchisor control may affect exit intentions. Third, although our study considers time in linking intentions to exit behavior, our cross-sectional survey data did not allow to track trust dynamics in real time. More detailed longitudinal data would enhance our understanding of how franchisee trust and control ultimately influence exit intentions, behaviors, and network instability. Fourth, in measuring trust congruence, we favored the inclusion of role- and duty-specific trust items to capture the crucial aspects of the distinct roles and duties of the franchisees and the franchisor. Future research should further assess the use of more generic scales that display greater item similarity. Finally, our congruence approach helps to assess the effects of (in)congruence, but does not verify partners’ awareness of each other’s trust. Our approach reduces social desirability bias but overlooks the dynamics and relational depth of other types of trust, such as mutual and reciprocal trust (Korsgaard et al., 2015). Future research could investigate how mutual trust, in which the franchisor and franchisee at least trust each other, influences exit intentions and exits, and explore the role of reciprocal trust via analyzing specific franchisor interventions and franchisee responses.

In conclusion, we shed light on the complex trust–control interplay that determines why and under what control conditions franchisees intend to leave their networks. We encourage researchers to explore entrepreneurs’ exit intentions (and exits) in various strategic network organizations like cooperatives or R&D consortia (e.g., Neergaard & Ulhøi, 2006). Understanding how different network configurations and control conditions shape entrepreneurs’ exits could enhance our knowledge of effective network leadership strategies.