We also controlled for FRUNITSjt, the natural log- transformed number of franchisee-owned units of franchise i in year t. In addition, we included the incidence of shirk- ing by both parties. For franchisee shirking, we coded the binary variable UNPAID as 1 if unpaid dues by the fran- chisee were the reported reason for the conflict (0 other- wise); we coded the variable TRADEMK as 1 if the reason reported for the conflict was trademark misappropriation by the franchisee (0 otherwise). We then computed franchisee shirking (FESHjj) as the sum of UNPAID and TRADEMK. Similarly, we coded INFO as 1 if the franchisor's failure to provide adequate information disclosure or provision of misleading information was the cited reason for the litiga- tion and as 0 otherwise. We coded the variable BREACH as 1 if breach of contract was claimed by the franchisee and as 0 otherwise. We then computed franchisor shirking (FRSHij) as the sum of INFO and BREACH. Table 2 pro- vides the descriptive statistics and correlation matrix for all variables included in the study.
Model Specification
The incidence of litigation is by no means a random occur- rence. Rather, parties "select" litigation as a means to achieve an end. It is therefore necessary to account for the selection of litigation with a first-stage Heckman (1979) selection model. Conditiotiing on litigation incidence, we must then account for four additional characteristics of our data. First, we have a mix of binary (FRINIT, ADR, and OUTCOME) and continuous (EXPAN) dependent variables. Second, the litigation initia- tion and resolution decisions and their corresponding dyadic and systemwide outcomes are probably related, requiring the specification of a correlated error structure among the four outcomes of interest to us (Greene 2003). Third, to assess the impact of litigation initiation and resolution choices on both outcome favorability and the franchisor's achievement of its expansion goals, recognition of the endo- geneity of the former regressors is required. Fourth, we must also account for the clustering of individual observations (litigated confiicts) within franchise systems (Hsiao 2003). The first-stage sample selection model (see Equation 1) includes all cases of franchisor-franchisee htigation observed for all 75 firms in our sample, each observed over multiple years. Although this initial sample comprised 1,133 obser- vations, missing data on franchisors' presence across multi- ple states and ownership structures resulted in 622 complete observations for the selection regression estimation.^ Con- ditioning on litigation incidence enables us not only to account for litigation selection but also to estimate the probability that either party may initiate litigation; whereas p represents the probability of the franchisor initiating liti- gation against a franchisee, (1 - p) denotes the likelihood of the franchisee initiating litigation against the franchisor.^
เรายังควบคุมสำหรับ FRUNITSjt ธรรมชาติล็อก-แปลงจำนวนหน่วยที่เป็นเจ้าของแฟรนไชส์ของแฟรนไชส์ฉันในปีที นอกจากนี้ เรารวมเกิดหนี ing โดยทั้งสองฝ่าย สำหรับแฟรนไชส์ shirking เรารหัส UNPAID ตัวแปรฐานสองเป็น 1 ถ้ายังไม่ได้ชำระกสิ่ง โดย fran-chisee มีรายงานสาเหตุความขัดแย้ง (0 อื่น ๆ - wise); เรารหัส TRADEMK ตัวแปรเป็น 1 ถ้าเหตุผลรายงานความขัดแย้งใน ฐานยักยอกทรัพย์ของเครื่องหมายการค้า โดยผู้ได้รับสัมปทาน (0 อื่น) เราแล้วคำนวณแฟรนไชส์ shirking (FESHjj) เป็นผลรวมของ UNPAID TRADEMK ในทำนองเดียวกัน เรารหัสข้อมูลเป็น 1 ถ้าความล้มเหลวของ franchisor เพื่อให้เปิดเผยข้อมูลอย่างเพียงพอหรือข้อบัญญัติของน่ารักเป็นเหตุผลอ้างอิง สำหรับ litiga-สเตรชัน และ 0 เป็นอย่างอื่น เรามีโค้ดละเมิดตัวแปรเป็น 1 ถ้าละเมิดสัญญาถูกอ้าง โดยแฟรนไชส์ และ เป็น 0 เราแล้วคำนวณ franchisor shirking (FRSHij) เป็นผลรวมของข้อมูลและการละเมิด ตารางที่ 2 โปร - vides อธิบายสถิติและความสัมพันธ์ของเมตริกซ์สำหรับตัวแปรทั้งหมดที่รวมอยู่ในการศึกษาข้อมูลจำเพาะของโมเดลThe incidence of litigation is by no means a random occur- rence. Rather, parties "select" litigation as a means to achieve an end. It is therefore necessary to account for the selection of litigation with a first-stage Heckman (1979) selection model. Conditiotiing on litigation incidence, we must then account for four additional characteristics of our data. First, we have a mix of binary (FRINIT, ADR, and OUTCOME) and continuous (EXPAN) dependent variables. Second, the litigation initia- tion and resolution decisions and their corresponding dyadic and systemwide outcomes are probably related, requiring the specification of a correlated error structure among the four outcomes of interest to us (Greene 2003). Third, to assess the impact of litigation initiation and resolution choices on both outcome favorability and the franchisor's achievement of its expansion goals, recognition of the endo- geneity of the former regressors is required. Fourth, we must also account for the clustering of individual observations (litigated confiicts) within franchise systems (Hsiao 2003). The first-stage sample selection model (see Equation 1) includes all cases of franchisor-franchisee htigation observed for all 75 firms in our sample, each observed over multiple years. Although this initial sample comprised 1,133 obser- vations, missing data on franchisors' presence across multi- ple states and ownership structures resulted in 622 complete observations for the selection regression estimation.^ Con- ditioning on litigation incidence enables us not only to account for litigation selection but also to estimate the probability that either party may initiate litigation; whereas p represents the probability of the franchisor initiating liti- gation against a franchisee, (1 - p) denotes the likelihood of the franchisee initiating litigation against the franchisor.^
การแปล กรุณารอสักครู่..
