Contract Duration and Franchisors Performance. An empirical study on French Data

Transcription

1 Contract Duration and Franchisors Performance An empirical study on French Data Magali CHAUDEY Muriel FADAIRO Frédéric PERDREAU Abstract: This paper deals with the contractual design in franchising networks and its impact on the franchisors performance. The focus is on the variations of contract duration. Economic theory suggests a trade-off between long and short term franchise contracts. The modification of the contract duration reflects a greater reactivity of the franchisor to exogenous shocks and therefore should translates in a higher performance. This hypothesis is submitted to French data. Keywords: Franchising Network; Contract Design; Contract Duration; Franchisor Performance; Applied Econometrics. JEL Classification Numbers: C12; M21. Paper presented at the EMNet Conference 2011 December 1-3, 2011, Limassol, Cyprus ( Université de Lyon, F Saint Etienne, France CNRS, GATE Lyon-St Etienne, UMR n 5824, 69130, Ecully, France Université de Saint-Etienne, Jean Monnet, F Saint Etienne, France Université de Lyon, F Saint Etienne, France COACTIS Lyon-St Etienne, EA 4161, 6 rue Basse des Rives Saint-Etienne Cedex 2 Université de Saint-Etienne, Jean Monnet, F Saint Etienne, France

2 INTRODUCTION Franchising is one of the most significant forms of retailing. For the past forty years it has received important attention in varied fields of research (for a review, see: Chabowski et al., 2011; Dant et al., 2011). In the economic literature, investigations on franchise data are an interesting way to test the theory of contracts. As mentioned by Pénard et al. (2004), three main issues concerning the franchise relationship can be distinguished in the econometric literature: i) the choice franchising versus owning (for instance: Barthélémy, 2011; Perdreau et al., 2011); ii) the relevance and justifications of dual distribution that is the co-existence in the same distribution network of franchised and owned units (for instance: Pénard et al., 2011); iii) the contractual design ie the structure of franchise contracts (for instance: Chaudey and Fadairo, 2010). This paper deals with the contractual design in franchising networks and its impact on the franchisors performance. The focus is on the variations of contract duration. The originality of this research is threefold. First, the length of the contractual arrangement is an important provision in any kind of contract, yet it has received few attention in the empirical literature on franchise data. Second, there is still a little study of the evolution of the contractual design (on this specific point, see: Argyres et al., 2007). This paper draws attention on the variations of franchise contract duration, which implicate a learning process in contracting. Third, the impact of the contractual design is studied in terms of franchisors performance. For this purpose, financial data and managerial issues are introduced to enrich the economic analysis of the paper. The rest of the paper is organized as follows. Section 1 discusses the analytical framework and derives the research hypothesis. Section 2 presents the data and some empirical specifications. Section 3 contains the estimations. Some concluding remarks are provided in section ANALYTICAL FRAMEWORK AND HYPOTHESIS 1.1. Background Transaction cost theory is the main framework that has been relied upon to generate predictions concerning contract duration. The theory implies that contracts will be longer when firms have more specific investments, since the need to protect those investments is greater. They will be shorter, in contrast, when environments are more uncertain. 2

3 Different empirical contributions are interested in the contract duration issue. We can mention Joskow s papers (1985, 1987). This author finds strong support for the hypothesis that differences in relationship-specific investments determine the duration of electric-utility/coal contracts. Like Joskow, in their study of natural-gas, Crocker and Masten (1988) find that firms use longer-term contracts when they face a greater likelihood of hold up, for example, when they have fewer buyer, seller, or transportation options. Most of the past empirical research regarding the duration provision has focused on labor contracts. In the franchising context there are few studies of the determinants of contract duration. Brickley et al. (2006) analyze the factors affecting the duration of contracts. They find that better-established franchisors choice on longer contracts, as do those franchisors that require greater investment levels from their franchisees and those who face higher recontracting costs. They explain the result on franchisee investment using asset-specificity arguments, and interpret the franchisor-experience effect in terms of reduced uncertainty. The influence of fixed specific-assets on the duration of franchise contracts finds another empirical support in the recent study of Garcıa-Herrera and Llorca-Vivero (2010) on Spanish data. Thus, the study of the determinants of the contract s duration is a topic little explored in the context of the franchise. Certainly, the franchise systems do not change the duration of their contract each year (Brickley et al. 2006), and this provision s choice seems to reveal a high standardization across franchises (Garcıa-Herrera and Llorca-Vivero, 2010). Yet over a few years time span, the event is not so rare. Our preliminary figures on the French market show that between 1995 and 1999 as well as between 1999 and 2003, nearly 20% of networks vary their contract s duration at least once. Furthermore, as noted by Lafontaine and Slade (2008), while Transaction Costs framework has generated studies of incidence, or put differently duration decisions, there are no empirical analyses of the consequences of such decisions. The present contribution is a first step to fill this gap Testable prediction Economic theory suggests a trade-off between long and short term franchise contracts (Blair and Lafontaine, 2005; Brickley et al., 2006; Garcıa-Herrera and Llorca-Vivero, 2010). The long-term contracts are favorable to the franchisee as they give him more time to recover its investment and as they protect him against the potential franchisor opportunism. On the other side, such contracts are less flexible and prevent the franchisors to adapt to environmental 3

4 changes. The modification of the contract duration involves a greater reactivity of the franchisor to exogenous shocks and therefore a higher performance. Focusing on the variations in franchise contracts duration, we consider the following hypothesis: Hypothesis: The variations in contract duration favor the franchisor performance This hypothesis is submitted to data in order to study the consequences of the variations in contract duration - associated with other variables such as specific investment and experience on networks performance. 2. EMPIRICAL SPECIFICATIONS 2.1. The data Our French dataset is constructed from two sources: surveys from the French Franchise Federation and the French file on financial data DIANE. We first collected information on networks from the annual franchise guides edited by the ACFI and by the French Franchise Federation for Annual data collected are: contract duration, date of the franchisor s first incorporation, date of his first franchised outlet, number of outlets (franchised and owned), total investment required for an outlet, and total minimum investment a franchisee must invest. We complemented the networks information by financial and accounting data on franchisors extracted from the DIANE database between 1995 and As we have different sources on the period and on 2003, we decided to initially study two periods of 5 years: and then Finally we observe 95 networks The study variables As the study takes into account five years periods, the contracts and networks independent variables are measured at the beginning of the reference period of the dependant variables As accounting data are more volatile than contract data, we have decided to work with averaged variables relating to the performance on each 5 years period 2. 1 The annual franchise guide was previously edited by the ACFI (French Association of Chambers of Commerce); it is now edited by the French Franchise Federation, since Concerning the variables relating to the franchisor performance, an average is computed only if we have at least the information for 2 consecutive years during the 5 years period. 4

5 PERFORMANCE (Profitability of the franchisor) Table 1 Study variables VARIABLE DEFINITION MEASURE Average gross margin rate AGE EXPERIENCE SIZE Return on assets Age of the franchisor (years) Experience of the franchisor (years) Size of the network EBITDA/Turnover on each period: average on the 5 years period p. Operating income/total assets: average on the 5 years period p Number of years between the franchisor s first incorporation and the beginning of the 5 years period p. Number of years between the franchisor s first incorporation and the first franchised unit. Total number of owned and franchised units in the network. Measured at the beginning of the 5 years period p. OWNED UNITS RISK INTANGIBLE RETAILSERVICE CHANGE DURATION DURATION INVESTMENT CONTRIBUTION Percentage of company owned units in the network Franchisor s Conan- Holder score. This variable is a proxy of the probability of bankruptcy of the franchisor: the higher is the score the lower is the probability of default Proportion of intangibles in the total fixed assets of the franchisor. Dummy variable Dummy variable Contract duration (years) Initial investment per retailer (K ) Personal capital contribution per retailer (K ) Number of owned units in the network / size of the European network. Measured at the beginning of the 5 years period p. Conan-Holder score is based on financial structure ratios and leverage ratios. It is computed by DIANE and measured at the beginning of the 5 years period p. Measured at the beginning of the 5 years period p. 1: the network is a service company 0: it is a retail one 1: contract duration has changed (up or down) during the 5 years before the p period 0: if no contract duration changes during the same period Measured in p 1: five years before the beginning of the period p. Measured in p 1: five years before the beginning of the period p. Measured in p 1: five years before the beginning of the period p. 5

6 2.3. Descriptive statistics Table 2 presents descriptive statistics for the 95 observations used in our estimations models. Table 2 - Summary statistics Variable (n=95) Mean Std. Dev. Min Median Max PERFORMANCE DURATION CHANGE DURATION EXPERIENCE AGE SIZE INTANGIBLE RISK INVESTMENT CONTRIBUTION OWNED UNITS The median network is 14 years old and is in franchising since 11 years (it takes a median of 3 years before a franchisor start into franchising). It is composed of 67 units, of which 14% are owned. The median franchisee is required to make an up-front investment of K 83.5 and must have K 47.5 in initial equity. The median contract has 5 years duration (min 2, max 20). A little more than 20% of contracts have seen their duration changed in the period. These duration data are very different depending on the sector as shown by figure 1. 6

7 Figure 1 Sector-based distribution of durations and variations in contract duration YEARS duration in years downward changes % upward changes % At the beginning of the period (in 1995) mean sectorial contract duration ranges from less than 5 years in Textile and Personal service and Equipment sectors, to more than 10 years in the Hotel and Restaurant sector. There are more upward changes than downward changes on the period and mean contract duration slightly grow from 5.63 years in 1995 to 5.73 years in Once again these changes are not uniform across sectors. In most of the sectors upwards changes are more numerous than downward changes, but for the Hotel and Restaurant there is only downward changes, and for Real Estate and Automobile Service and Equipment there is no change. As our first concern is about the effect of renegotiation of contract duration, we compare financial and contractual features of networks that change contract duration to those that do not change. We first present in table 3 univariate t-test of difference in mean. 7

8 Variable Table 3 Comparisons between the two groups of networks Mean - did not change (75 obs.) Mean changed (20 obs.) Difference Difference s std. err. PERFORMANCE DURATION EXPERIENCE AGE * SIZE INTANGIBLE RISK INVESTMENT CONTRIBUTION OWNED UNITS H 0 : Difference=0; * p < 0.05, ** p < 0.01, *** p < Table 3 reveals that the mean differences for the performance variable do not reach the 5% significance threshold, although being closed to it 3. The only feature which differentiates the two groups of networks is the age. Networks that changed their contract s duration are younger than those that did not change their contract s duration. Table 4 - Correlation matrix CHANGE DURATION EXPERIENCE AGE SIZE OWNED UNITS INTANGIBLE RISK RETAILSERVICE _cons CHANGE DURATION EXPERIENCE AGE SIZE OWNED UNITS INTANGIBLE RISK RETAILSERVICE Constant If we measure performance with the economic return on assets, we barely reach the one sided 5% significance threshold: difference is (= ) and standard error (t=1,7). 8

9 3. ESTIMATIONS 3.1. Econometric model To estimate the influence of change in contract duration on the network performance, we retain a treatment-effects model. This model allows us to account for the bias in the choice of change in contract duration. The treatment-effects model considers the effect of an endogenously chosen binary treatment (change in contract duration) on another endogenous continuous variable (network performance), conditional on two sets of independent variables. The idea behind these models is to estimate two regressions simultaneously. The first is a probit regression predicting the probability of "treatment." The second is a linear regression for the outcome of interest as a function of the "treatment" variable, controlling for observable confounders. Selection bias arises because networks that changed their contract duration are not randomly selected, and this may have an impact on their performance: the "treatment" may be correlated with the error term in the outcome equation. (1) Equation for the choice change in contract duration versus no change (probit model): prob (CHANGE DURATION i = 1 Xi) = β 0 + β 1 DURATION i + β 2 AGE i + β 3 INVESTMENT i + β 4 CONTRIBUTION i + β 5 OWNED UNITS i + β 6 RETAILSERVI i + β 7 8 SECTOR i +µ i i = {1,, 95} s=1 (2) Equation for the franchisor performance (OLS): PERFORMANCE i = α 0 + α 1 CHANGE DURATION i + α 2 RISK i + α 3 RETAILSERVICE i + α 4 EXPERIENCE i +α 5 AGE i +α 6 SIZE I +α 7 INTANGIBLE INVESTMENT i +α 8 OWNED UNITS i + ε i i = {1,, 95} In equation (2), Performance is an average on a 5 years period, and each independent variable (except CHANGEDURATION) is measured at the beginning of the period. CHANGEDURATION relates to the preceding 5 years period. We expect a positive impact of change in contract duration on performance. For the franchisor, the possibility to modify contract duration allows a greater reactivity to environment change and then higher performances. The other variables are control variables. 9

10 Equation (1) is one period (5 years) lagged compare to equation (2). In equation 1, each independent variable is measured at the beginning of the CHANGEDURATION period, that is five years before the independent variables in equation (2) (except for OWNED UNITS which is present in each equation and measured only at the beginning of the performance variable). The variable CHANGEDURATION is the treatment variable in the model and is explained by the following variables: AGE refers to the age of the franchisor. The DURATION of the contract at the beginning of the period (p-1) is introduced as a control variable but could also be interpreted as a successful experience variable. Ceteris paribus, a sound and balanced relationship between franchisors and franchisee should translate in longer and more stable contract duration than an unbalanced relationship. Hence contract duration may have a negative impact on the probability of contract duration change. For franchisees, the minimum cash or equity investment to purchase is used to measure the importance of the financial stake in the relationship (CONTRIBUTION). For the franchisor, we used the minimum total investment to open an outlet (INVESTMENT). The more the investment at stake is important, the more stakeholders should carefully look at the duration and negotiate it. Hence, we expect a positive impact of these variables on the probability of renegotiation of contract duration. We used the proportion of OWNED UNITS as a proxy of the balance of power between franchisor and franchisees within the network. With a high proportion of owned units, franchisee point of view should be less taking into account and contract duration (shorter and) less often negotiated. Finally, 8 sector dummies are introduced as control variables The results The estimation results are reported in table 5. 10

11 Table 5 Estimation results for the franchisor performance (Dependant variable: profitability measured by the gross margin rate) (1) (2) (3) CHANGE DURATION * 0.175* 0.175* (0.0645) (0.0683) (0.0683) RISK ** ** ** ( ) ( ) ( ) RETAILSERVICE * * * (0.0362) (0.0370) (0.0371) EXPERIENCE ( ) ( ) AGE ( ) ( ) SIZE ( ) ( ) INTANGIBLE INVESTMENT (0.0498) (0.0510) OWNED UNITS (0.0670) _cons (0.0214) (0.0418) (0.0441) CHANGE DURATION DURATION ** ** ** (0.115) (0.114) (0.114) AGE (0.0284) (0.0285) (0.0285) 11

12 INVESTMENT ( ) ( ) ( ) CONTRIBUTION ( ) ( ) ( ) OWNED UNITS (0.773) (0.772) (0.772) Sector dummies yes yes yes RETAILSERVICE (0.618) (0.629) (0.629) _cons (1.015) (1.015) (1.015) hazard lambda * * (0.0425) (0.0439) (0.0439) N Prob > chi Standard errors are in brackets * p < 0.05, ** p < 0.01, *** p < The complete model is estimated in the third column. The models (1) and (2) confirm that the estimation results are robust. The good global significance level of the models suggests that the analytical framework is partially relevant. The last general comment is that the selection bias is confirmed, the hazard lambda is significant for the models (2) and (3). The change in contract duration, the risk and the nature of the sector (retail vs. service) are the main variables that explain the network performance. The other characteristics of the network, like experience, age or size, are not significant. In other words, two of three variables that explain performance are independent of the network strategy. In appendix 1, we propose another measure of performance based on return and we can mention that, even if estimations are less convincing, the change in contract duration estimated coefficient also significantly explains the franchisor performance. Concerning the probit equation, the initial contract duration is the only variable that explains the contract duration changes. But the coefficient of this variable is negative. In our case, 12

13 contract duration reduces the probability to change the contract duration. We can hypothesize that in our sample, as networks are quite old (15 years on average) and with experience (11 years in average), they are found the contract duration adapted with their strategy. So they are not incited to change the contract duration. This is consistent with Brickley et al. (2006, p. 179): Holding the age of the company constant, franchisors are likely to acquire more information about optimal contract terms in larger systems, since they can observe more units. If so, the economic view suggests that larger franchisors will tend to offer longer-term contracts. Thus, according to Brickley et al (2006), the best informed (the largest) franchisors choose long contract durations. Therefore it seems relevant that the franchisors that offer the longest contracts are also those that change the least the duration of their contracts. The nature of the sector has no significant influence on the contract duration change. We can only mention that networks in Hotel-restaurant sector differ from networks in Clothes-shoes sector (reference sector) concerning change in contract duration 4. A work on sub-samples could explain these sector-based specificities. At the present time, our sample is too small to allow these tests on sub-samples. 4. CONCLUDING REMARKS Few studies are devoted to the determinants of contract duration (Brickley et al. 2006, Garcia-Herrera A. and R. Llorca-Vivero 2010). This paper offers a first insight on the consequences of changes relating to this contractual provision. Consistent with our hypothesis, the estimation results show that changes in contract duration are associated with a better financial performance in the subsequent period. This result has managerial implications for the franchisors. It means that adapting the franchise contract duration turns out to enhance the financial performance compared to other franchisors. From a franchisee point of view, the issue requires additional tests as the benefits for franchisors can be a part of the network global surplus, or it can be extracted from franchisees. More generally, beyond this first result, many questions remain to explore, and hypotheses can be refined. In this paper we assume that networks change contract duration due to exogenous shocks. But specific hypotheses on the determinants of changes in contract duration can be developed. Our results furthermore reveal an impact of contract duration changes on the 4 The remark is the same concerning estimations in appendix 1. 13

14 franchisor performance, but didn t explore for sample size consideration- the issue of a different impact depending on whether the change is upward or downward. The first step provided by this paper could be completed in interesting ways.this enrichment requires a broadening of our data analyses and an increase in our sample size. Hence, our next step will be to construct a panel database. This panel would allow, on a broader sample, to implement other data analyses and to refine our results. In particular, it would offer more insights in the issues of what drives the duration changes 14

15 REFERENCES Argyres, N. S., Bercovitz J., Mayer K. J., 2007, Complementarity and Evolution of Contractual Provisions: An Empirical Study of IT Services Contracts, Organization Science, Vol. 18, No. 1, January February pp Barthélémy J., 2011, Agency and Institutional Influences on Franchising, Journal of Business Venturing, Vol. 26, p Blair, R. and Lafontaine, F., 2005, The Economics of Franchising, Cambridge University Press, United States of America. Brickley, J. A., Misra, S., & Van Horn, L., 2006, Contract duration: Evidence from franchising, Journal of Law and Economics, 49, Chabowski, B.R., Hult, T. and Menac, J.A., 2011, The Retailing Literature as a Basis for Franchising Research: Using Intellectual Structure to Advance Theory, Journal of Retailing, Vol. 87 No. 3, pp Chaudey M. and Fadairo M., 2010, Contractual Design and Networks Performance: Empirical Evidence from Franchising, Applied Economics, Vol. 42, pp Dant R. P., Grünhagen M., Windsperger J., 2011, Franchising Research Frontiers for the Twenty-First Century, Journal of Retailing, Vol. 87 No. 3, pp Garcia-Herrera A. and R. Llorca-Vivero, 2010, How time influences franchise contracts: the Spanish case, European Journal of Law and Economics, 30, Pénard T., Raynaud E., Saussier S., 2004, Théorie des Contrats et Réseaux de Franchise, Revue française d économie, Vol. 18, No. 4, pp Pénard, T., Raynaud, E. and Saussier, S., 2011, Monitoring Policy and Organizational Forms in Franchised Chains, International Journal of the Economics of Business, forthcoming. Perdreau F., Le Nadant M.L., Cliquet G., 2011, Plural Form and Franchisor Performance : Early Empirical Findings from Europe, in Tuunanen M., Windsperger J., Cliquet G., Hendrikse G., New Developments in the Theory of Networks, Springer,

16 APPENDIX 1 For this estimation, we retain return as a measure of performance. The variable Rentab is the return on assets and is computed as Earnings before interest taxes depreciation and amortization (EBITDA) / (Shareholder's equity + financial debt). Treatment-effects model -- two-step estimates Number of obs = 94 Wald chi2(11) = Prob > chi2 = Coef. Std. Err. z P> z [95% Conf. Interval] rentab1 Change durat Risk Retailserv Experience Age Size Intang inv Owned units Change durat (omitted) _cons Change durat Duration Age Invest contrib

17 Owned unit alimentati hotelresta real estate leisure serviceseq~e serviceseq~o serviceseq~n comser _cons hazard lambda rho sigma lambda