LOAD FACTOR, BAGGAGE FEES, AND MERGER AND ACQUIS I TION IN THE U.S. AIRLINE INDUSTRY By W enyi K uang A DISSERTATION Submitted to Michigan State University in partial fulfil l ment of the requirements for the degree of Business Administration Logistics Doctor of Philosophy 2020 ABSTRACT LOAD FACTOR , BAGGAGE FEES, AND MERGER AND ACQUI SI TION IN THE U.S. AIRLINE INDUSTRY By Wenyi Kuang The relationships between load factor , operational performance , and financial performance present inconsistent findings in extant literature. As such, Chapter O ne aims to reconcile the mixed findings by delineating these relationships at more nuanced levels thought statistical within and betwee n specification, which has not been adopted in previous literature. The findings strongly support the crucial importance of within and between specification, indicating that between carriers, load factor demonstrates an inverted U - shaped relationship with financial performance. Within carriers, the higher the average load factor, the more negative impact on financial performance with the increase of load factor. Building on the mixed findings from previous literature as well as leveraging on cognitive appr aisal theory, Chapter Two investigates how the implementation of the new baggage fee - time arrivals, and consumer complaints. Utilizing discontinuous growth modeling, our analysis shows that the effect of this policy is twofold. Fi nancial performance dropped immediately upon the policy implementation but improved for about 3.5 years before facing a diminishing return. On - time arrivals improved immediately upon the policy implementation and kept improving for another 4 years before t he effect diminishes. Although there was no immediate impact on consumer complaint, the trend of consumer complaint, in the long run, demonstrates an inverted - U shaped curve with time passing since the policy implementation. Drawing on organizational learn ing framework and building on discontinuous growth curve modeling , Chapter Three investigates the impact of mergers on operational performance and financial performance at two distinctive stages: the immediate transition stage and the long - term recovery st age. Operational performance was found to deteriorate immediately while financial performance was found to increase immediately upon mergers. No long - term impact was found with regard to both operational performance and financial performance. However, carr - merger performance moderates the performance during the transition stage in that low - performing carriers, rather than high - performing carriers, benefit more in both operational performance and financial performance. Keywords: Load fact or, baggage fee policy, merger and acquisition, on - time performance, operating profit, operating revenue, U . S . airline industry . iv A CKNOWLEDGEMENTS I would like to extend my gratitude to my dissertation committee (Dr. Miller, Dr. Griffis, Dr. Whipple, and Dr. Bo lu mole) for supporting me during my dissertation development stage and for mentoring me during my whole PhD life at Michigan State University . I would also like to thank my family who always support s me in my life . v TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ........................ vii LIST OF FIGURES ................................ ................................ ................................ ..................... viii KEY TO ABBREVIATIONS ................................ ................................ ................................ ........ ix CHAPTER ONE LOAD FACTOR AND FIRM PERFORMANCE ................................ .............. 1 1.1 INTRODUCTION ................................ ................................ ................................ ................. 1 1.2 HYPOTHESES DEVELOPMENT ................................ ................................ ....................... 3 1.2.1 Lo ad Factor and On Time Performance: Queuing Theory and the Concept of Asset Frontier ................................ ................................ ................................ ................................ .... 4 1.2.2 Load Factor and Financial Performance: T he Concept of Slack ................................ ... 8 1.3 DATA ................................ ................................ ................................ ................................ .. 12 1.3.1 Airlines ................................ ................................ ................................ ......................... 13 1.3.2 Dependent Va riables ................................ ................................ ................................ .... 14 1.3.3 Independent Variables ................................ ................................ ................................ . 16 1.3.4 Control Variables ................................ ................................ ................................ ......... 17 1.3.5 Summ ary Statistics ................................ ................................ ................................ ...... 21 1.4 ANALYSIS AND RESULTS ................................ ................................ .............................. 22 1.4.1 Within and Between Specification ................................ ................................ .............. 23 1.4.2 Methodology ................................ ................................ ................................ ................ 25 1.4.3 Results ................................ ................................ ................................ .......................... 27 1.5 MANAGERIAL INSIGHTS ................................ ................................ ............................... 32 1.5.1 Insi ghts for Decision and Policy Makers ................................ ................................ ..... 32 1.5.2 Insights for Operations Managers ................................ ................................ ................ 34 1.6 CONCLUSION ................................ ................................ ................................ .................... 35 1.6.1 Theoretical Contributions ................................ ................................ ............................ 35 1.6.2 Limitation and future research ................................ ................................ ..................... 37 CHAPTER TWO BAGGAGE FEES AND FIRM PERFORMANCE ................................ ......... 40 2.1 INTRODUCTION ................................ ................................ ................................ ............... 40 2.2 HYPOTHESES DEVELOPMENT ................................ ................................ ..................... 42 2.2.1 New Baggage Fee Policy and Carrier Financial Performance ................................ ..... 43 2.2.2 New Baggage Fee Policy and On - Time Performance ................................ ................. 46 2.2.3 New Baggage Fee Policy and Consumer Comp laints ................................ ................. 50 2.3 DATA ................................ ................................ ................................ ................................ .. 53 2.3.1 Airlines ................................ ................................ ................................ ......................... 53 2.3.2 Dependent Variables ................................ ................................ ................................ .... 54 2.3.3 Independent Variables and Coding of Time ................................ ................................ 56 2.3.4 Control Variables ................................ ................................ ................................ ......... 58 2.3.5 Summary Statistics ................................ ................................ ................................ ...... 62 2.4 ANALYSIS AND RESULTS ................................ ................................ .............................. 64 vi 2.4.1 Between and Within Specification ................................ ................................ .............. 64 2.4.2 Model Testing Procedures ................................ ................................ ........................... 64 2.4.3 Hypotheses Testing Results ................................ ................................ ......................... 67 2.5 ROBUSTNESS TEST ................................ ................................ ................................ ......... 71 2.6 MANAGERIAL INSIGHTS ................................ ................................ ............................... 73 2.6.1 For Strategic Decision Makers ................................ ................................ ..................... 73 2.6.2 For Operations Managers ................................ ................................ .............................. 74 2.7 CONCLUSION ................................ ................................ ................................ .................... 75 2.7.1 Theoretical Contributions ................................ ................................ ............................. 75 2.7.2 Limitation and Future Research ................................ ................................ .................... 77 CHAPTER THREE MERGER AND ACQUIS I TION AND FIRM PERFORMANCE .............. 79 3.1 INTRODUCTION ................................ ................................ ................................ ............... 79 3.2 LITERATURE R EVIEW ................................ ................................ ................................ .... 82 3.2.1 Overview of Research on Merger ................................ ................................ ................. 82 3.2.2 Merger in the U.S. Airline Industry Background Information ................................ .. 83 3.2.3 Performance Implications of Merger in the U.S. Airline Industry ............................... 84 3.3 THEORY AND HYPOTHESES ................................ ................................ ......................... 87 3.3.1 Theoretical Foundation ................................ ................................ ................................ . 87 3.3.2 The Impact of Merger on Operational Performance ................................ ..................... 89 3.3.3 The Impact of Merger on Financial Performance ................................ ......................... 93 3.3.4 Moderating Effect of the Immediate and Long - term Impact of Merger ....................... 94 3.4 METHOD ................................ ................................ ................................ ............................ 97 3.4.1 Data Source and Sample ................................ ................................ ............................... 97 3.4.2 Measures ................................ ................................ ................................ ....................... 98 3.4.3 Descriptive Statistics ................................ ................................ ................................ ... 105 3.4.4 Analytical Method ................................ ................................ ................................ ...... 105 3.5 RESULTS ................................ ................................ ................................ .......................... 107 3.6 DISCUSSION ................................ ................................ ................................ .................... 111 3.6.1 Theoretical Contributions ................................ ................................ ........................... 111 3.6.2 Managerial Implications ................................ ................................ ............................. 113 3.6.3 Limitation and Future Research ................................ ................................ .................. 113 3.7 CONCLUSION ................................ ................................ ................................ .................. 115 APPENDICES ................................ ................................ ................................ ............................. 117 A PPENDIX A Comparison of Current Research with Selective Literature ............................ 118 A PPENDIX B Summary Statistics and Correlation Matrix ................................ .................... 119 A PPENDIX C Hypothesized Relationships for H1, H2, and H3 ................................ ............ 120 A PPENDIX D Summary Statistics and Correlation Matrix ................................ .................... 121 A PPENDIX E Summary of U.S. Airline Merger and Acquisition Research .......................... 122 REFERENCES ................................ ................................ ................................ ............................ 126 vii LIST OF TABLES Table 1 Airlines in the Dataset ................................ ................................ ................................ ...... 14 Table 2 Variable Used in Analysis ................................ ................................ ............................... 22 Table 3 Random Intercept Model 1 and Random Intercept and Slope Model 2 .......................... 26 Table 4 OTP as the Dependent Variable ................................ ................................ ....................... 28 Table 5 OPOR as the Dependent Variable ................................ ................................ ................... 29 Table 6 Airlines in Dataset ................................ ................................ ................................ ........... 54 Table 7 Coding Time Using Alaska Airline as an Example ................................ ......................... 57 Table 8 Variables Used in Analysis ................................ ................................ .............................. 63 Table 9 Random Intercept Model to Calculate ICC ................................ ................................ ..... 65 Table 10 Select Model Random Effects ................................ ................................ ....................... 66 Table 11 Final Model to Test Hypotheses ................................ ................................ .................... 68 Table 12 Airlines in Dataset ................................ ................................ ................................ ......... 98 Table 13 Coding Transition, Recovery, and Time Using Delta Airline as an Example ............. 101 Table 14 Define High - performing and Low - performing Acquirers ................................ ........... 102 Table 15 Variables Used in Analysis ................................ ................................ .......................... 104 T able 16 Summary Statistics and Correlation Matrix ................................ ................................ . 105 Table 17 Hypotheses Testing Results ................................ ................................ ......................... 108 Table 18 Comparison of Current Research with Selective Literature ................................ ........ 118 Table 19 Summary Statistics and Correlation Matrix ................................ ................................ . 119 Table 20 Summary of Statistics and Correlation M atrix ................................ ............................ 121 Table 21 Summary of U.S. Airline Merger Literature ................................ ................................ 122 viii LIST OF FIGURES Figure 1 The Effect of Average Load Factor on OPOR ................................ ............................... 31 Figure 2 Moderating effect of Average Load Factor ................................ ................................ .... 32 Figure 3 Hypothesis 1 OPOR Graph (American Airline as an Example) ................................ .... 69 Figure 4 Hypothesis 2 OTP Graph (Frontier as an Example) ................................ ....................... 70 Figure 5 Hypothesis 3 Complaint Graph (Frontier as an Example) ................................ ............. 71 Figure 6 Moderating Effect on OTP at Transition ................................ ................................ ...... 110 Figure 7 Moderating Effect on Financial Performance at Transition ................................ ......... 110 Figure 8 Hypothesized Relationships for H1, H2, and H3 ................................ ......................... 120 ix KEY TO ABBREVIATIONS DO J Department of Justice DO T Department of Transportation OLS Ordinary Least Squares OPOR Operating Profit over Operating Revenue OTP On - Time Performance U.S. United States 1 CHAPTER ONE LOAD FACTOR AND FIRM PERFORMANCE 1.1 INTRODUCTION Firm financial performance has been an evergreen research focus across all disciplines because 1995). The criticality o f firm financial performance is accentuated in the U.S. airline industry by the Department of Transportation (DOT), the U.S. airline industry demonstrated substantial flu ctuations in the past decade, ranging from 2.9 billion net income in 2009 up to 24.8 billion net income in 2015. Without sound financial achievement, airlines will face financial stress and eventually file for bankruptcy or go out of business (Alan and La pré 2018). One equally important airline performance measure on the operations side is on - time performance (OTP), which demonstrates its importance both internally and externally. Internally, OTP affects 2006); exter nally, OTP is the service characteristic most 2001). passengers. The ratio of number of passe ngers to total available seats is accordingly known as - oriented passengers brings in the same time, imposes more operational challenges, leading to potential deteriorated OTP. In fact, research on the relationship s performance ha ve been flourishing in operations management literature. Interestingly, the results of both relationships present conflicting findings. Load factor has been found to worsen OTP 2 (Bratu and B arnhart 2006; McCartney 2010; Scotti and Dresner 2015) as well as have no impact financial performance due to the increased capacity utilization (Behn and Riley 199 9; Tsikriktsis (Collins et al. 2011) or have no impact on it (Belobaba 2005). Intrigued by these conflicting findings, we aim to delineate the relationships between load factor, OTP, and financial performance at more nuanced levels. Specifically, we distinguish between - carrier differences from within - carrier variations by resorting to between - within specification (Hoffman 2015; Bell and Jones 2015), in contras t to previously adopted methodologies, such as ordinary least squares (Behn and Riley 1999; Shaffer et al. 2000), fixed effect approach (Ramdas and Williams 2008; Sim et al. 2010; Atkinson et al. 2013), and random effect approach (Saranga and Nagpal 2016; Zou and Chen 2017). Between - within specification allows us to partition longitudinal within - carrier variations from cross - sectional between - carrier differences so that more nuanced levels of phenomena can be examined. Our new methodological approach also a Using between - within specification as well as drawing theories and findings from extant literature, we hypothesize that between carriers, the effect of load factor on OTP demonstrates a diminishing return curve while the effect of load factor on financial performance expects an inverted U - shaped relationship. However, within carriers, the hypothesized relationships depend will differ for carriers who consistently operate at higher load factors against carriers who consistently operate at lower load factors. Although our results did not lend support to the 3 relationship between load factor and OTP, the relationship between load factor and financial performance is strongly validated. Our findings illustrate that between carriers, the relatio nship between load factor and financial performance is an inverted U - shaped relationship. Within carriers, increasing load factor will hurt financial performance for carriers whose average load factor is high but enhance financial performance for carriers whose average load factor is low. Our research contributes to knowledge accumulation in airline research in several ways. Frist, we leverage extant operations and management theories to reconcile the conflicting findings in literature. Second, our betwee n - within specification reveals more nuanced relationships compared with extant fixed effect or random effect models applied in previous research. Third, our findings provide substantial guidance for airline strategic decision makers as well as related prac titioners. A brief comparison of our research and the related research could be found in Appendix A . Our article starts with hypotheses development based on literature review in Section 2. Data collection and variable construction are described in Section 3. Section 4 outlines the key steps we conducted to test our hypotheses and reports the results. Managerial insights were presented in Section 5 following analysis and results. Lastly, Section 6 concludes our research by summarizing the article with limit ations and directions for future research. 1.2 HYPOTHESES DEVELOPMENT This section aims to leverage the extant literature and theory from operations and management field to reconcile the conflicting findings previously discussed in the introduction section. We 4 - ai rport segment - airport segment multiplied by the number of seats Load factor in our manuscript specifically refers to passenger load factor. 1.2.1 Load Factor and On Time Performance: Queuing Theory and the Concept of Asset Frontier - time arrival or the opposite of on - time arrival arrival delays. DOT defines an on - The literature presents contradictory findings regar ding the relationship between load factor and OTP. The majority of literature reveals that load factor negatively impacts on - time arrivals, presenting two categories of explanations. First, higher load factors imply greater number of passengers and greater number of passengers impose greater challenges on passenger flows starting from check in, security check, boarding, up to deplaning (Bratu and Barnhart 2006). The overcrowded cabin with passengers squeezing around fighting for spaces for their carry - on lu ggage was reported as one major reason for flight delays (Tuttle 2014). This effect is also intrinsically indicates proportionally greater number of luggage, which c onsequently prolong s the ground handling time (McCartney 2010; Scotti and Dresner 2015). This effect is termed as non - to longer system processing time, resulting in late arrivals. Other scholars have also argued that the worse on - time performance is a 5 deliberate action planned out by carriers to achieve higher revenues. For example, Atkinson et al. legacy (low - cost) carrier may trade off an increase of 1% delay (>15 min) sed a very limited time frame of data, such as August 2000 of one US major airline (Bratu and Barnhart 2006), Quarter 1 only from 2007 to 2010 (Scotti and Dresner, 2015), or simply a snapshot in time (Tuttle 2014; McCartney 2010). Other literature, howeve r, found no significant relationship between load factor and on - time arrivals. For example, Ozment and Morash (1994), using a panel data compiled from DOT arrival performance at all. To reconcile the conflicting findings of the relationship between load factor and OT P , we resort to queuing theory to explain this relationship at a more nuanced level. According to Kleinrock (1975, 1976) and Kelton (2002), a queuing system can be summarized to have the following a If the demand of service requested by the increasing number of incoming entities exceeds the system handling capacity, waiting time will become prolonged and eventually increase at an increa sing rate when the system approaches 100% capacity utilization. Airline operations can be visualized as a typical queuing system (Ramdas and Williams 2008). Customers arrive at airports to take flights thus an arrival process is generated. Individual - in counters and boarding gates, and onboard crew, are all 6 present to provide services to customers. Customers, on the other hand, compete for services from resources, such as service from agents and crew, space in the overhea d cabin, and seat assignments etc. We accordingly apply queuing theory to reconcile the findings of extant literature where load factor demonstrates conflicting impact on OTP. Based on queuing theory, it is reasonable to expect that with everything else b eing equal, higher load factor will exacerbate on - time performance. However, on - time performance will keep a steady trend for a while before it turns worse. The reasons are as follows. Airlines plan and schedule a fixed number of flights at each airport an d subsequently allocate appropriate resources to accommodate customers (Brueckner 2004; Papadakos 2009). When the number of customers is still within the handling capacity of the system, the system will operate with minimum waiting time at checking in, boa rding, and up to deplaning. Thus, on - time arriv als will not be impacted. Consequently, the steady performance of on - capacity. However, the resources allocated at each a irport such as agents and cabin crew are normally fixed in the near term (Brueckner 2004; Papadakos 2009). Once these fixed resources can no longer handle the services demanded by increased numbers of customers, the waiting time will be prolonged at all st ages in the queuing system, which will inevitably cause delays and as a result, on - time performance will start to diminish. The discussion so far applies when we view airline industry operations holistically across all carriers at any given airport, i.e., the relationship discussed here is a cross - sectional relationship between load factor and OTP across airlines regardless of where each airline stands in terms of their overall average load factor. We term this between - carrier effect and accordingly, we pro pose: 7 H1a: The effect of load factor on OTP remains constant before starting to face diminishing returns . In reality, some carriers consistently operate at high load factors while other carriers consistently operate at low load factors. Thus, the effect of 1% increase of load factor on their respective on - time performance will most likely differ. This is where we turn our attention to the concept of asset frontier and the discussion of within - carrier effect. In examining why some manufacturing plants out perform others, Schmenner and Swink (1998) frontier is made up of operat ing frontier (frontiers formed by choices in plant operation) and asset frontier (frontiers formed by choices in plant design and investment). Schmenner and Swink (1998) proposed that if a firm operates close to its asset frontier, the firm will be likely to operate under the law of trade - offs while if a firm operates away from its asset frontier, the firm will be likely to operate under the law of cumulative capabilities. When a firm operates close to its asset frontier under the law of trade - gle plant can provide superior performance in The Theory of Performance Frontier has subsequently been applied to airline - related research. Using fleet utilization (total block hours divided by total aircraft hours) as the proxy for asset frontiers, Lapré and Scudder (2004) investigated the relationship between quality (consumer complaints) and cost (operating expenses divide d by available seat miles) where airlines with higher fleet utilization are assumed to be closer to their asset frontier. Lapré and Scudder (2004) found that those airlines operating closer to their asset frontier are operating under the law of trade - offs, i.e., they were only able to improve either cost or quality but not simultaneously on 8 referred to aircraft utilization (flight time and taxi time divided by total time s cheduled) as the moves towards its asset frontier; 2) increasing load factor has a worse impact on highly utilized aircrafts than for less utilized aircrafts. Exte nding the Theory of Performance Frontier and the findings of the literature to the current research, we similarly use load factor (revenue passenger miles divided by available seat miles) as the proxy for asset frontier. Airlines with higher load factor ar e accordingly assumed to be closer to their asset frontier. Consequently, we argue that when an airline operates close to its asset frontier (at higher load factor), it will operate under the law of trade - offs, i.e., it cannot increase both its on - time per formance and load factor simultaneously. In other words, when an - time performance will deteriorate. However, if an airline operates away from its asset frontier at a lower load factor, it will operate under the law o f cumulative capabilities. In this case, it will be able to improve both dimensions simultaneously. We term this as within carrier effect and present: load factor such that OTP will become worse for carriers who operate closer to their asset frontier but become better for carriers who operate away from their asset frontier. 1.2.2 Load Factor and Financial Performance: The Concept of Slack Financial performance in airlin e literature has been operationalized in a variety of ways, such as return on assets (Ramaswamy et al. 1994), operating profit over operating revenue (OPOR) (Tsikriktsis, 2007), and profitability (Collins et al. 2011; Zou and Chen 2017). 9 There are also two distinctive findings regarding the impact of load factor on financial performance. One stream of findings reveals that load factor positively contribute to airline financial performance. In general, Wyckoff and Maister (1977) found that 1% of difference i n load factors could lead to as high as 5% differences in profitability. Ramaswamy et al. (1994) also confirmed that 5% greater load factor translates into 7% greater return on assets. More specifically, the impact of load factor on profitability can be ju stified from two perspectives: increased capacity utilization and greater number of passengers. From capacity utilization perspective, Behn and Riley (1999) found that load factor is positively associated with contemporaneous operating income; Tsikriktsis (2007) concluded that 1% increase in passenger load factor result in a 0.63% increase in OPOR; Zou and Chen (2017) also found that higher From the effec t of greater number of perspective, Schefczyk (1993) observed that passenger - focused airlines achieved higher profitability compared with non - passenger focused airlines. On the other hand, another stream of literature equally found that higher load factor does not lead to increased financial performance. Belobaba (2005) analyzed DOT data from 2001 to 2004 and concluded that although price cuts in airline ticketing stimulated record high load factors, the high load factors, however, do not impro ve revenues. Collins et al. (2011), analyzing 14 top profit margin in their generalized least squares ( GLS ) models using both quarterly and annual data. To exp lain the mixed findings regarding the relationship between load factor and financial performance, we leverage the concept of slack to explore this relationship. Bourgeois (1981) 10 defines organizational slack at four different levels: strategic, individual, subunit, and process level. The process level slack is the most relevant to operations and supply chain management in that it buffers between organization processes, such as raw materials and finished goods. Overall, helps to solve internal problems as well as to pursue external goals (Bourgeois, 1981, p. 29). financial slack, operational slack, customer relational slack, and human resource slack. Operational slack itself, in the operations management field, is operationalized in a variety of ways, including excess capacity (Steele and Papke - Shields 1993; Bourland and Y ano 1994), days of inventory (Hendricks et al. 2009; Azadegan et al. 2013; Kovach et al. 2015), ratio of sales to property, plant and equipment (Hendricks et al. 2009; Kovach et al. 2015), and cash to cash cycle (Hendricks et al. 2009; Kovach et al. 2015). In our current research, we adopt the excess capacity perspective to develop our hypotheses. Excess capacity slack in our research settings refers to the percentage of empty seats in an aircraft. i.e., high load factor indicates less slack while low load factor means more slack. The relationship between slack and firm performance has been extensively explored in different disciplines. Mishina et al. (2014) investigated the relationship between financial slack and firm growth in 112 manufacturing firms and Yu (2016) explored the impact of physical capacity utilization on actual and long run minimum costs in 13 international low - cost airlines and subsequently concluded that it is better for carriers to bear some idle capacity rather than to operate at full capacity. Tan and Peng (2003) examined the relationship between slack and firm financial performance and su bsequently found that this 11 not whether slack is uniformly good o r bad for performance, but rather, what range of slack is Extending the concept of slack and the findings from literature to the relationship between load ue that either too high or too low lots of empty seats which directly translates to lost revenues, so , the bottom line will suffer as a result. When the load factor is high, everything else being equal, airlines will have to utilize more resources to cope with greater number of passengers, which will generate more costs. Excessive costs, on the other hand, will also hurt the bottom line. Similar to hypothesis 1a and corresponding to slack literature where cross - firm relationships were examined, we term this as between - carrier effect and posit: e demonstrates an inverted U - shape relationship. As with hypothesis 1b, we also argue that this relationship should also demonstrate different effects within carriers for carriers who operate at high load factor versus carriers who operate at low load fact or. To explain the within - carrier relationship, we turn our attention to the law of diminishing returns and the law of diminishing synergy (Schmenner and Swink 1998). Law of uring plant nearer and nearer to its operating frontier or its asset frontier, more and more resources must be 1998, he strength of the synergistic effects 12 predicted by the law of cumulative capabilities diminishes as a manufacturing plant approaches To illustrate how these two laws work in airline industry, let us assume we have two carriers operating at the load factor of 70% and 90% respectively and both want to increase their load factor by an absolute 5%. Then , according to the law of diminishing returns, we can expect that it w ill require less resources to increase load factor from 70% to 75% than from 90% to 95%. In a financial performance) the two carriers can expect will also be dif ferent, i.e., the benefit of increasing load factor from 70% to 75% is expected to be greater than that of increasing from 90% to 95%. Hence, we present: on carriers who operate closer to their asset frontier but become better for carriers who operate away from their asset frontier. 1.3 DATA We collected our data from the Depa least one percent of total domestic scheduled - performance measures. DOT airline data has been widely explored for publication in various discipl ines such as management (Schefczyk 1993), economics (Atkinson et al. 2013), and operations research (Lapré and Scudder 2004). operational performance was compiled 13 Financial reports are a conglomerate of six regions: Atlantic, Domestic, International, Latin America, Pacific, and System while monthly consumer report consists of US domestic flights only. So only domestic financial figures were kept in our data to match the domestic data in monthly consumer report. At the time of accessing DOT site, financial performance is available from 1990 to 201 9 while monthly consumer report spans from 1998 to 201 9 . 1.3.1 Airlines we elected to choose our data starting point as the first quarter of 2004. Our key financial measures are in quarterly format while operational measures are in mo nthly format, which was subsequently aggregated into quarterly level - - nd tracking the name changes of some airlines, our data consists of 25 carriers from 2004Q1 to 201 9 Q 2 . Some carriers span the whole 62 quarters while others report fewer quarters either due to their revenues falling below the one percent reporting threshol d or due to merger and acquisition. A detailed summary of airlines in our data analysis is presented in Table 1. 14 Table 1 Airlines in the Dataset No. Airline First quarter in the sample Last quarter in the sample Total quarters in the sample 1 AIRTRAN 2004 Q1 2013 Q4 40 2 ALASKA 2004 Q1 2019 Q2 62 3 ALOHA 2006 Q2 2008 Q1 8 4 AMERICA WEST 2004 Q1 2005 Q4 8 5 AMERICAN 2004 Q1 2019 Q2 62 6 ATA 2004 Q1 2006 Q4 12 7 ATLANTIC SOUTHEAST 2004 Q1 2012 Q3 35 8 COMAIR 2004 Q1 2010 Q4 28 9 CONTINENTAL 2004 Q1 2011 Q4 32 10 DELTA 2004 Q1 2019 Q2 62 11 ENDEAVOR 2007 Q1 2013 Q4 26 12 ENVOY 2004 Q1 2015 Q4 54 13 EXPRESSJET 2004 Q1 2018 Q4 60 14 FRONTIER 2005 Q2 2019 Q2 57 15 HAWAIIAN 2004 Q1 2019 Q2 62 16 INDEPENDENCE 2004 Q1 2005 Q4 8 17 JETBLUE 2004 Q1 2019 Q2 62 18 MESA 2006 Q1 2013 Q4 38 19 NORTHWEST 2004 Q1 2009 Q4 24 20 SKYWEST 2004 Q1 2019 Q2 62 21 SOUTHWEST 2004 Q1 2019 Q2 62 22 SPIRIT 2005 Q1 2017 Q4 18 23 UNITED 2004 Q1 2019 Q2 62 24 US AIRWAYS 2004 Q1 2013 Q4 40 25 VIRGIN AMERICA 2012 Q1 2017 Q4 24 Notes: 1. RU code was used from October 2003 to June 2006 by DOT to code ExpressJet. Effective July 2006, the carrier code for ExpressJet Airlines changed in the report from RU to XE. In our dataset, RU was changed to XE. 2. American Eagle Airlines changed to Envoy effective April 2014 report. Both Envoy and American Eagle were treated as ENVOY in our data. 3. Atlantic Coast Airlines changed to Independence Airline since 2004 November in the report. Both airlines were treated as Independence in our data. 4. Endeavor Air, formerly Pinnacle Airlines, was ranked for the first time in January 2013 . Both Pinnacle and Endeavor were treated as Endeavor in the data. 5. Atlantic Southeast (EV) was acquired by ExpressJet and changed to XE since. 1.3.2 Dependent Variables Our first dependent variable is financial performance. In airline literature, three categories of measures were used regarding airline financial performance: absolute measures, predicted values and relative measures. Absolute measures take the form of profitability (Kalemba and Campa - Planas 2017 a (Ramdas et al. 2013). Relative measures are calculated either as operating profit over operating 15 cost (Steven et al. 2012) or as operating profit over operating revenue (Tsikriktsis 2007; Mellat - Parast et al. 2015). We e lected to use operating profit over operating revenue (OPOR in our models) as our financial performance measures due to two reasons. First, profitability has some variations over the years and a considerable amount of profitability values are negative. If taken natural logarithm, those negative profitability values will become missing data points , which is not a true reflection of airline financial status. Second, the excessive variance of profitability comes from the different sizes of carriers. Ratio meas ures, in this case, can help to account for the size differences among carriers than other financial measures (Dresner and Xu 1995) as well as to overcome the difficulty to discern owned versus leased aircrafts (Tsikriktsis 2007; Mellat - Parast et al. 2015) . Operating profit and operating revenue were retrieved from DOT Schedule P1.2. Our second dependent variable is operational performance. Our main research interest is to investigate how each airline performs in terms of their on - time arrivals. DOT define s an on - time - time performance, four closely related measures were adopted in literature. On e stream of airline literature exactly follows the definition of DOT by calculating the overall percentage of fights arriving within15 minutes of scheduled arrival time (Suzuki 2000; Rupp et al. 2006; Peterson et al. 2013; Kalemba and Campa - Planas 2017). A nother stream of literature adopts the opposite of DOT definition by calculating percentage of delays, such as delay% > 15 minutes (Tsikriktsis 2007; Forbes 2008 a ; Prince and Simon 2009; Ramdas et al. 2013; Mellat - Parast et al. 2015). Averaged minutes of d elay or the actual duration of delay were also used in some research (Rupp et al. 2006; Forbes 2008 a ; Prince and Simon 2009; Cook et al. 2012; Yimga 2017). Finally, Scotti et al. (2016) 16 examined the percentage of a specific type of delay, i.e., airline - cau sed flight delays. We chose - time performance by calculating the percentage of flights that variable is denoted as OTP in our model s and complied from DOT Air Travel Monthly Consumer Report. 1.3.3 Independent Variables Our main independent variable is load factor definition of passenger load factor (revenue passenger miles divided by avail able seat miles) was followed strictly in airline literature (Behn and Riley 1999; Shaffer et al. 2000; Atkinson et al. 2013; Dana and Orlov 2014). We also adopted the same formula when calculating load factor. Relevant data were retrieved from DOT Schedul and was subsequently collapsed into quarterly format. There are four different forms of load factor in our models: the between - effect load factor, the qu adratic term of load factor, the within - effect of load factor, and the cross - level interaction term of load factor. Denote each carrier by i and the measurement occasions by t , then the between - effect load factor can be calculated by taking the group mean of load factor for each carrier, denoted as in our models. The square term of the between effect of load factor is constructed to investigate the non - linear relationship, denoted a s . The within - . This variable is denoted as in our models. 17 Lastly, to investigate the mod erating effect, we created a cross - level interaction between and . 1.3.4 Control Variables Fuel Cost Fuel as a control variable in airline literature falls into two categories. Ramdas et al. (2013) adopted fuel price (price per gallon) as a control to investigate the relationship between service quality and airline financial performance while Dana and Orlov (2014) used fuel cost to control for cost shocks to examine how internet penetration impacts load factor. We used fuel cost , rather than fuel price, as a control because fuel cost is more relevant to our research question given that some carriers hedged their fuel requirements at much lower cost compared with the market fuel price. For example, the average fuel price fluctuate d from $44.6 per barrel (2016) to $111.8 per barrel (2012) in our sample data period (IATA 2018 ). But the actual fuel cost per barrel varies greatly for each carrier depends on how well they have hedged their fuel requirement. For example, in the second ha lf of 2005, Southwest hedged 85% of its fuel requirements at the equivalent of $26 while the industry average is $72.35 (Alexander 2006). To performance. Fuel cost was reported in monthly basis in Schedule P12(a), which was also aggregated to quarterly data. Number of Enplaned Passengers Number of enplaned passengers was included as a control for two considerations. As was discussed before, greater number of passengers will impose greater challenges in the system which consequently leads to longer processing time, resulting in potential delays. On the other hand, carriers can also expect higher revenue with increased number of passengers, which then 18 impact DOT Monthly Consumer Report . Market Share Market share is a common control variable in airline research but was operationalized slight differently. Behn and Riley (1999) and Suzuki (2000) defined market share as the ratio of the number of passengers of the sampled airline to the total passengers of 10 largest airlines. Shaffer r miles of scheduled flights on that route. Collins et al. (2 011) created an annual market share index by squaring revenue passenger miles to the sum of revenue passenger miles of the total 2 5 carriers in that quarter. Revenue passenger miles were retrieved from DOT Schedule T1. Firm Size market Three categories of measures were used to proxy firm size in airline literature: financial measures, capacity measures, and human resource measures. Financial measures take the form of total revenue, total sales or total assets (Mishina et al. 2004; Collins et al. 2011). Capacity measures demonstrate themselves as revenue passenger miles (Shaffer et al. 2000) as well as available seat miles (Steven et al. 2012). Number of employees was used to proxy human resource measures (Tan and Peng 2003; Kalemba and Campa - Planas 2017 a ). Because our 19 dependent variable is operating profit over operating revenue while our load factor is revenue passenger miles over available seat miles, we hence avoid using financial measures and capacity measures to proxy firm size. We elect to use number of employees to proxy firm size because it is more relevant in our research given the fact that it can both impact on - time perfor mance and operating profit. More employees, especially employees at the airport, will be helpful to fasten ground operations processes thus improving OTP. However, more employees also indicate more expenses which will negatively impact operating profit giv en that employee expenses are the 2018 ). Number of employees was taken from DOT Schedule P1(a). Total Delays DOT defines delays as flights that arrived 15 minutes after the scheduled time shown in the carriers' Computerized Reservations Systems. Delays at departure gates contribute directly to on - time arrival performance. Late arrivals at arrival gates subsequently impact the departure time of the following scheduled flights. Delays, either airborne or ground, result in significant costs to We thus include delay as a control variable in our models. Recent airline literature made stringent efforts to control for weather - related delays regarding on - time performance (Ramdas et al. 2013; Nicolae et al. 2017). However, DOT breaks down delays into seven categories in their report since October 2003. These include cancelled flights, diverted fligh ts, aircraft delay, extreme weather delay, national aviation system delay, security delay, and late arriving aircraft delay. With our 1 6 years data inclusive of 25 carriers, the total delayed flight is 22% out of 90,235,491 total flights while among the 22 % delayed flights, the distribution is as follows: cancelled flights (7.8%), diverted flights (1%), air craft delay (25.7%), extreme weather 20 delay (3.2%), national aviation system delay (30.4%), security delay (0.2%), and late arriving aircraft delay (31.6 %). Based on this analysis, we decided to keep the total number of delays in our model to reflect the holistic picture of the impact of delay rather than focus on weather only which only accounts for 3.2% of total delays. Another reason to use total number of delays is that no matter what kind of delay it is, delay will eventually impact on - time performance as well as financial performance. LCC researchers have noticed the different impact of different airline groups. To this end, the classic distinction between airline groups, also the terminology adopted by DOT, is low cost carriers (LCC) and legacy carriers (Rupp et al. 2006; Atkinson et al. 2 013; Yimga 2017). Low cost carriers are also referred to as focused carriers (Tsikriktsis 2007; Mellat - Parast et al. 2015) or geographic specialists (Lapré and Scudder 2004). The main characteristics of LCC are that LCC fly point to point within limited ge ographic areas with fewer aircraft types targeting price - sensitive customers (Mellat - Parast et al. 2015). Legacy carriers are also referred to as network carriers (Collins et al. 2011; Garrow et al. 2012), full service carriers (Tsikriktsis 2007; Ramdas an d Williams 2008; Mantin and Wang 2012), non - focused carriers (Mellat - Parast et al. 2015), and geographic generalists ( Lapré LCC and legacy carriers throughout our manuscript. On - time performance wise, Rupp et al. (2006) found that everything else being equal, worse on - time performance occurs on those routes where there is more competition from LCC, indicating that LCC is the potential contributor to worse on - time performance. Financial performance wise, Collins et al. (2011) showed that the legacy carriers tend to achieve more persistent profit 21 margins and asset turnover ratios than LCC. Mantin and Wang (2012) also confirmed that the profitability of legacy carriers improved faster than that of LCC after 9/11. However, Tsikriktsis (2007) revealed a different story by concluding that LCC outperformed the rest of the industry in terms of profitability by focusing their resources on limited point to point network operations. To account for the different impact of the two categories of carriers on OTP and financial performance, we included LCC as dummy variable in our models. In the latest DOT report, six carries were classified as low - cost carriers (LCC): Allegiant Air, Frontier, JetBlue, S outhwest, Spirit , and Virgin America. 1.3.5 Summary Statistics To provide a better view of the variables used in this manuscript, Table 2 provides all the variable names, how they are constructed, and their data sources. Appendix B presents correlation matrices for the two outcome variables with load factor and other control variables. We report correlations for each variable constructed as both within and between effect, which is to be further explained in the following section. 22 Table 2 Variable Used in Analysis Variable Formula/Definition Data Source On - T i me Performance Quarterly overall percentage of fights arriving within 15 minutes of scheduled arrival time. DOT Air Travel Monthly Consumer Report OPOR Operating profit divided by operating revenue at quarterly level . DOT Schedule P1.2 Load Factor Quarterly revenue passenger miles divided by available seat miles. DOT Schedule T1 Load Factor Between Group mean of load factor for each carrier . DOT Schedule T1 Load Factor Within n. DOT Schedule T1 Fuel Cost Between Group mean of quarterly fuel cost for each carrier . DOT Schedule P12(a) Fuel Cost Within . DOT Schedule P12(a) Enplaned Passengers Between Group mean of quarterly enplaned passengers for each carrier . DOT Air Travel Monthly Consumer Report Enplaned Passengers Within mean . DOT Air Travel Month ly Consumer Report Market Share Between The ratio of a carrier's quarterly revenue passenger miles to the sum of revenue passenger miles of the total carriers in that quarter. Take the mean to construct between variables . DOT Schedule T1 Market Share Within . DOT Schedule T1 Number of Employees Between Group mean of quarterly number of FTEs for each carrier . DOT Schedule P1(a) Number of Employees Within employees from its mean . DOT Schedule P1(a) Total Delay Between The sum of delays caused by "cancelled flights, diverted flights, aircraft delay, extreme weather delay, national aviation system delay, security delay, and late arriving aircraft delay. Take the mean to construct between variables". DOT Air Travel Monthly Consumer Report Total Delay Within . DOT Air Travel Monthly Consumer Report LCC Low cost carrier defined by DOT DOT Air Travel Monthly Consumer Report 1.4 ANALYSIS AND RESULTS Before we present the steps taken to build our models and test hypotheses, we first briefly discuss the concepts of within - carrier effect and between - carrier effect from a more statistical perspective. 23 1.4.1 Within and Between Specification As discussed in the introduction section, estimation method in extant airline research evolves from OLS to fixed effect approach then to random effect approach when estimating parameters using panel data. However, fixed effect ap proach and random effect approach demonstrate their own limitations in extant literature. model panel data for many researchers, which is also reflected in airline related literature (Ramdas and Williams 2008; Sim et al. 2010; Atkinson et al. 2013). However, Bell and Jones esults leading to misleading results interpretations (p. 134). Fixed effect in this sense can only estimate within group variations over time (Bell and Jones, 2015). Further, time varying covariates (load factor in our case) contain two parts: one part spe cific to higher - level entity (carrier in our case) which does not change between measurement occasions and the other part that changes over time representing the differences between measurement occasions (Bell and Jones 2015; Hoffman 2015). The two parts a ccordingly have their own different effects in a model and are subsequently called between and within effects are compressed together by asking one single variab le to account for both within and between effects, which results in removing all between - firm variances on the variable (Bell and Jones 2015; Hoffman 2015). Correctly specifying within and between effect is separate between - and within - person sources of variation when modeling repeated measures data can lead to biased results and potentially 24 incorrect conclusions about within - p. 119). To solve the problems associated with fixed effect modeling , random effect modeling (also referred to as multilevel modeling , hierarchical linear modeli ng , or mixed models) is preferable. Multilevel modeling has the following advantages: 1) it accounts for differences between groups (carriers in our case) by partitioning variances between them; 2) slopes of different groups are allowed to vary at differen t magnitudes; 3) variances at measurement occasion level can also be modeled, allowing specifics of occasion level to be retained in the model while still having the ability for generalization (Bell and Jones 2015; Hoffman 2015). To correctly specify a mul tilevel model, two new variables for a time varying covariate are constructed: one variable accounting for the between effect using the mean (Mundlak 1978) and one variable accounting for the within effect using the deviation from the mean (Berlin et al. 1 999; Bells and Jones 2015). Reviewing the random effect models adopted in extant airline research, we can see that 1) between and within effects were not modelled separately (Saranga and Nagpal 2016; Zou and Chen 2017); 2) Hausman test was used to conclude that random effect is preferred over fixed 2015, p. 144). Since our panel data set is hierarchically constructed, i.e., it consists of repeated measures over time t nested within multiple carriers i , within and between specification can be readily applied. We follow Bell and Jones (2015) and Hoffman (2015) to construct ou r variables and specify our models. 25 1.4.2 Methodology The first step of our analysis is to build the level one model. The importance of a correct level one model specification can be illustrated from two perspectives. First, since our main research interest is to investigate between - carrier effect as well as within - carrier effect, we need to make 6 years (Bliese and Ployhart 2002). Second, the correct specification of the level one model is critical in that the validity of the full model depends on the correct specification of the level one model, especially how time is defined relative to the outcome (Raudenbush 2001). Our level one model - building decomposes into two steps. Step 1 is to calculate Intraclass Correlation (ICC) to make sure our dataset is indeed longitudinal. Step 2 is to select the correct random effect. A random intercept model was fitted for step 1 and a random inter cept and random slope model was fitted to compare the model fit in step 2 (Bliese and Ployhart 2002; Fitzmaurice et al. 2011; Hoffman 2015). The comparison between the two models is summarized in T able 3 . The ICC of OTP as dependent variable model is 0.44, revealing 44% variation between carriers and the remaining 56% variation within carriers. Similarly, in the OPOR as dependent variable model, the ICC yields a value of 0.5391, indicating that 53.91% of the variation is between carriers and the remaining 4 6.09% variation is within carriers. Both ICCs strongly indicate the nature of longitudinal data, providing further support to our theoretical hypotheses on between - carrier and within - carrier effects. A likelihood ratio test between model 2 (random intercept and random slope model) and model 1 (random intercept model) for both OTP 2 2 = 72.38, p = 0.000 respectively), which serves as random effect specification all throughout our analysis. 26 Ta ble 3 Random Intercept Model 1 and Random Intercept and Slope Model 2 On Time Performance OPOR Parameter Model 1 Model 2 Model 1 Model 2 Fixed Effects Intercept 0.78** (93.36) 52.29** (59.67) 0.78** (93.36) 52.29** (59.67) Occasion 0.0008** (5.60) 0.0007* (2.20) 0.0008** (5.60) 0.0007* (2.20) Random Effects Level 2: Carriers 0.0017** 0.004** 0.016** 0.029** Variance 0.000003** 0.000027** Covariance - 0.00008 * - 0.0008** Level 1: Occasion 0.0022** 0.0019** 0.014** 0.012** Measures of Fit - 2 Log Likelihood - 2925.6 - 3006.1 - 1220.2 - 1292.6 2 30.76** 80.52** 25.75** 72.38** ICC 44.00% 53.91% - tailed). Z - tests are reported in parentheses for the fixed effects parameters The full specification of our model is illustrated in equation 1 follow ing Bell et al. (2018) and R square calculation is presented in equation 2 follow ing Nakagawa and Schielzeth (2013). Individual carriers are denoted as i (level 2) which are measured/reported on multiple occasions t odel both within and between individual effects concurrently, and also explicitly models heterogeneity in the effect of predictor variables at the . Equation 1 Full Model Specification 27 Equation 2 R square calculation 1.4.3 Results To test our hypotheses, we fitted a series of models based on our full model specification in equation 1. The results are summarized in Table 4 and Table 5 (with OTP and OPOR as different dependent variables). Model 1 is our baseline model where no focal p redictors were added, only operational performance and financial performance. Model 2 is the model where the main predictor was added followed by model 3 which further adds the square ter m of load factor ( ) to test the quadratic operational performance and financial performance . Built on model 3, model 4 adds the between effect of load factor to simultaneously test the main effects of betwe en - carrier differences and within - carrier trends. Finally, model 5 is the fully specified model to test all the main effects and interaction effects by adding the interaction term of within effect and between effect. 28 Table 4 OTP as the Dependent Variable OTP Parameter Model 1 Model 2 Model 3 Model 4 Model 5 Fixed Effect Intercept 0.49* (2.52) 0.44* (2.09) 0.45* (2.13) 0.33 (1.77) 0.34 (1.78) - 0.12 ( - 0.77) 0.89 (0.40) 0.38 (0.18) 0.34 (0.17) 0.34 (0.17) - 0.65 ( - 0.46) - 0.40 ( - 0.30) - 0.38 ( - 0.28) Load Factor_within - 0.26** (11.08) - 0.26** (10.31) * Load Factor_within 0.14 (0.43) - 0.009 ( - 1.41) - 0.009 ( - 1.51) - 0.009 ( - 1.43) - 0.009 ( - 1.68) - 0.009 ( - 1.68) Fuel Cost_within - 0.002 ( - 1.32) - 0.002 ( - 1.34) - 0.002 ( - 1.32) - 0.002 ( - 1.66) - 0.002 ( - 1.65) 0.11** (4.99) 0.11** (4.78) 0.11** (4.69) 0.12** (5.47) 0.12** (5.46) Enplaned Passengers_within 0.17** (27.02) 0.17** (27.03) 0.17** (26.97) 0.17** (30.05) 0.17** (30.06) 0.30 (1.58) 0.30 (1.56) 0.31 (1.62) 0.19 (1.13) 0.20 (1.15) Market Share_within 0.06 (0.85) 0.05 (0.81) 0.05 (0.82) 0.10 (1.61) 0.10 (1.61) - 0.05** ( - 2.57) - 0.05** ( - 2.66) - 0.05** ( - 2.68) - 0.04* ( - 2.47) - 0.04* ( - 2.49) Total Employees_within 0.02** (3.02) 0.02** (3.04) 0.02** (3.04) 0.01 (1.62) 0.01 (1.61) - 0.07** ( - 7.78) - 0.08** ( - 6.60) - 0.08** ( - 6.59) - 0.08** ( - 7.81) - 0.08** ( - 7.80) Total Delay_within - 0.18** ( - 64.51) - 0.18** ( - 64.44) - 0.18** ( - 64.44) - 0.18** ( - 69.43) - 0.18** ( - 69.44) LCC - 0.03* ( - 2.45) - 0.03* ( - 2.34) - 0.03** ( - 2.28) - 0.03** ( - 2.71) - 0.03** ( - 2.70) Year and Quarter Fixed Effect (Included) Random Effect Level 2: Carriers 0.001 0.001 0.001 0.001 0.001 Variance 0.00 0.00 0.00 0.00 0.00 Covariance - 0.0002 - 0.0002 - 0.0002 - 0.0002 - 0.0002 Level 1: Occasion 0.0002 0.0002 0.0002 0.0002 0.0002 Measures of Fit - 2 Log Likelihood - 4630.5 - 4631.1 - 4631.3 - 4745.7 - 4745.9 AIC - 4566.5 - 4565.1 - 4563.3 - 4675.7 - 4673.9 BIC - 4413.8 - 4407.6 - 4401.0 - 4508.7 - 4502.1 R 2 (MVP) 94.20% 94.22% 94.18% 94.83% 94.83% 2 (MVP) 0.03% - 0.05% 0.65% 0.00% 2 (LRT) 0.57 0.20 114.39** 0.18 - tailed). Z - tests are reported in parentheses for the fixed effects parameters 29 Table 5 OPOR as the D ependent V ariable OPOR Parameter Model 1 Model 2 Model 3 Model 4 Model 5 Fixed Effect Intercept - 2.16** ( - 6.97) - 2.14** ( - 6.78) - 2.14** ( - 6.84) - 1.87** ( - 5.91) - 1.90** ( - 6.07) 0.13 (0.40) 11.95* (2.14) 13.28* (2.37) 15.84** (2.80) - 7.33* ( - 2.11) - 7.97* ( - 2.29) - 9.47** ( - 2.70) Load Factor_within 0.74** (4.61) 0.57** (3.40) * Load Factor_within - 7.71** ( - 3.49) 0.05** (6.94) 0.06** (6.84) 0.06** (7.27) 0.06** (6.69) 0.06** (6.98) Fuel Cost_within 0.001 (0.06) 0.001 (0.13) 0.004 (0.47) 0.005 (0.56) 0.005 (0.51) 0.15** (4.51) 0.15** (4.14) 0.15** (4.32) 0.15** (4.29) 0.14** (3.97) Enplaned Passengers_within - 0.004 ( - 0.10) - 0.005 ( - 0.12) - 0.002 ( - 0.14) - 0.03 ( - 0.69) - 0.02 ( - 0.66) - 1.19** ( - 3.87) - 1.19** ( - 3.84) - 1.12** ( - 3.68) - 0.87** ( - 2.83) - 1.03** ( - 3.34) Market Share_within - 0.13 ( - 0.31) - 0.13 ( - 0.32) - 0.07 ( - 0.16) - 0.21 ( - 0.51) - 0.25 ( - 0.62) - 0.05 ( - 1.58) - 0.06 ( - 1.64) - 0.07* ( - 2.16) - 0.07* ( - 2.16) - 0.07* ( - 2.09) Total Employees_within - 0.12** ( - 3.09) - 0.12** ( - 3.02) - 0.13** ( - 3.26) - 0.10** ( - 2.37) - 0.10** ( - 2.35) - 0.07** ( - 4.25) - 0.06** ( - 3.09) - 0.06** ( - 3.04) - 0.06** ( - 2.95) - 0.05** ( - 2.72) Total Delay_within 0.009 (0.52) 0.009 (0.51) 0.009 (0.53) 0.02 (1.20) 0.02 ( - .96) LCC - 0.07** ( - 3.55) - 0.07** ( - 3.50) - 0.08** ( - 3.79) - 0.08** ( - 3.79) - 0.08** ( - 3.86) Year and Quarter Fixed Effect (Included) Random Effect Level 2: Carriers 0.03 0.03 0.03 0.03 0.03 Variance 0.00002 0.00002 0.00001 0.00001 0.00001 Covariance - 0.0007 - 0.0007 - 0.0006 - 0.0006 - 0.0006 Level 1: Occasion 0.01 0.01 0.01 0.01 0.01 Measures of Fit - 2 Log Likelihood - 1349.0 - 1349.1 - 1353.3 - 1374.3 - 1386.3 AIC - 1284.9 - 1291.1 - 1285.3 - 1304.3 - 1314.3 BIC - 1132.3 - 1152.7 - 1123.0 - 1137.3 - 1142.5 R 2 (MVP) 42.34% 42.58% 46.01% 48.82% 51.43% R 2 (MVP) 0.24% 3.43% 2.81% 2.60% 2 (LRT) - 4.16 21.02** 12.05** - tailed). Z - tests are reported in parentheses for the fixed effects parameters. 30 Turning our attention to the model fit, we see that o ur fully specified model (model 5) explains 94.83% of variation for OTP and 51.43% of variation for OPOR. The consistent reduction of - 2 loglikelihood value through model 1 to model 5 indicates improved model fit. The smaller AIC value for model 5 also pro ves that our fully specified model is the better fit. predicts that the effect of load factor on OTP will start with a constant return curve then transit into a diminishing return curve. H1a is jointly tested by the coefficients of and , neither of which is significant, failing to support H1a. Hypothesis 1b posits that the within - carrier effect of load factor on OTP is moderated by a c H1b is subsequently tested by the coefficient of the interaction term, which is also non - significant, thus H1b is not supported either. ial performance, proxied by OPOR in our research. Hypothesis 2a predicts an inverted U - shaped relationship between average load factor and firm profitability while Hypothesis 2b suggests that verage load factor. Turning to the results of OPOR, the coefficient of is positive ( 15.84, p < 0.01 ) while that of is negative ( - 9.47, p <0.01 ), both of which are statistically significant, suggesting an inverted U - shaped relati onship. To better visualize this relationship, we plotted this relationship in Figure 1. To plot the graph, all significant control variables were taken at their mean values and load factor ( ) was taken at the range of 65.4% to 87.7% according to the data summary statistics. The solid line represents the range of values within our dataset while the dotted line stands for the extrapolated load factor up to 94% (the highest 31 individual carrier level load factor). We can clearly see an inverted U shape from the graph. A calculation of the turning point also proves that the turning point is at the load factor of 8 4.2 %, within the range of our dataset. Hence H2a is supported. Figure 1 The Effect of Average Load Factor on OPOR The interaction term between average load factor and the within effect from its average load factor is also statistically significant ( - 7.71, p <0.001 ), providing strong support for H2b performance strongly depends on its average load factor. To illustrate this relationship, we also plotted the interaction effect in Figure 2, following the procedures de scribed by Dawson (2014). High deviation and low deviation represent the degree to which a carrier increase s its load factor. financial performance while inc is high, further increment in its load factor will only hurt its financial performance. -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 OPOR Load Factor The Effect of Load Factor on OPOR 32 Figure 2 Moderating effect of Average Load Factor 1.5 MANAGERIAL INSIGHTS We discuss the potential managerial contributions from both strategic and operational perspective to provide possible guidance for strategic decision makers as well as operational practitione rs. 1.5.1 Insights for Decision and Policy Makers While Hypotheses 1a and 1b were not supported, it is worthwhile to check the relevant coefficients. Although not statistically significant, and respectively show a positive and a negative sign on OTP, indicating a potential inverted U - shaped relationship. Plotting this relationship in a graph shows that OTP worsens only marginally with the increase of average load factor, which triggers us to hypo thesize that based on the data we have observed, the effect of average load factor on OTP has not yet reached the turning point at the inverted U shaped curve. Combining this line of thinking with the findings of H2, we can OPOR Low Average Load Factor High Average Load Factor 33 speculate that the effect of loa d factor on financial performance is more pronounced that its effect on OTP. The average load factor stops at 87.7% in our dataset hence we conjecture that to observe a significant drop in OTP, probably we need higher than 87.7% average load factors. By c omparing the findings between H1 and H2, we showed that the impacts of load factor on operational performance and financial performance are different. Strategically, decision and policy makers should consider the different impacts of load factor on operati onal performance and financial performance when deciding how full they want their aircrafts to be filled. Although we did not find significant evidence to support the hypothesis that the relationship between load factor and OTP is an inverted U - shaped rela tionship, the correct sign of the coefficients indicates an accelerated trend of the relationship becoming more negative. On the other hand, the inverted U - shaped relationship between load factor and financial performance makes it apparent that too little specifically, we find the turning point to be at the load factor of 8 4.2 % where financial performance starts to decline above this point while at and above the load fac tor of 84.2 %, the OTP keeps deteriorating at an accelerated speed. Since the trend of OTP is predicted to deteriorate consistently at an accelerating speed while that of financial performance is an inverted U - shape, it is advisable for an airline to operat e below the threshold of 84.2 % load factor where the financial performance increases along with the increase of load factor. An important decision to make, however, is how much trade - off an airline is willing to make between worse on - time performance and b etter financial performance. Atkinson et al. (2013) showed that a legacy (low cost) carrier is willing to trade off of 1% increase in worse on - time performance for a 0.31% (0.38%) increase in load factor which subsequently translates to more revenue. Trade off or not, this is an important decision confronted by strategic decision makers. 34 1.5.2 Insights for Operations Managers Operations managers face numerous challenges along with the issue of load factor itself, such as collateral consequences of delays etc. Thi s section aims to provide a holistic view for operations managers regarding load factor and operations related issues based on our findings. It is easy to understand that the number of enplaned passengers positively contribute to OPOR as more passengers b ring in more revenue. It is somehow counter - intuitive to find that both the between and within measures of enplaned passengers positively contribute to OTP. The explanation for the effect of between effect (average number of enplaned passengers) is that wi th an industry wide greater number of passengers, all carriers will spend more effort to ensure faster turn - around time, such as during Christmas. While for the within effect (deviation from average enplaned passengers), the explanation might be that the n umber of passengers is reported to DOT at flight level. A carrier experiencing higher number of passengers on a specific flight is likely to feel the urgency to ensure faster turn - around time compared to a half empty flight. Since our data is at firm level , we will leave this interesting finding for future research using flight level data. The implication for operations managers when designing operations strategies is clear: both internal operational scenarios and external industry wide situations will have to be counted for. Total delay, whether measured by average total delays or by longitudinal changes of delays, both worsen OTP, as expected. However, t longitudinal changes of delays on OTP is more pronounced than its relative pos ition ( i.e., its averaged delays ) compared with other competitors. But t h e longitudinal changes of delays within a carrier does not impact its financial pe rformance, suggesting that compared with competitors, if a carrier demonstrates more delays, this carrier will subsequently suffer from its financial performance. Hence the indication is clear: 35 within carriers, increas ing load factor will inevitably hurt o n - time performance although this may help to improve their financial performance up to the load factor of 84.2 %. Here th is is another trade - off decision to make for operations managers. Operation managers can refer to the above - mentioned findings as a guideline to design their operations strategy by considering both operational measures and financial measures to find a reasonable balance between them based on corporate strategy. Since the increase of load factor will make OTP marginally worse, operational managers can also design an acceptable load factor policy given the existing operating resources in line with corporate strategy. To this end , this is another trade off to make and all st akeholders should participate in the process to come up with a reasonable trade off policy. 1.6 CONCLUSION 1.6.1 Theoretical Contributions Our research demonstrates two broad contributions to knowledge advancement and knowledge accumulation in airline research. The methodology of multilevel modeling with between and within specification helps to promote more rigorous estimation method to the widely adopted panel data set compiled from DOT. Our findings, on the other hand, reconcile the conflicting findings in extant literature, encouraging researchers to leverage on existing operations theories to re - explore the relationships found in extant literature. Even though almost all airline related research compiled panel data from DOT, the research methodologies adopted of fixed effect, further to random effect. Fixed effect, although indisputably popular, has shortcomings when it comes to estimate panel data, especially when the researchers are 36 inte rested in between group differences (Singer and Willet 2003; Hoffman 2015). Random effect, with its increasing exposure in most recent research, needs to be correctly specified in order to capture both cross - sectional differences across groups and longitud inal changes within groups (Bell and Jones 2015). To this end, we leveraged on multilevel modeling literature and introduced the specification of within and between effects to investigate the relationship between load factor and firm performance. Our findi ngs strongly support the necessity and importance of the within and between specification because they lead to different interpretations regarding the examined relationship. As such, our methodological approach contributes to future airline research that a ims to model nuanced differences both between carriers and within carriers. Our result also enhances the importance of distinguishing between and within effect in e viewed from two perspectives: cross - sectionally, the impact of load factor on financial performance demonstrates an inverted U - shaped relationship; longitudinally, this relationship on carriers who consistently operate at high load factor versus carriers who consistently operate at low load factor. We also resort to existing operations theories and concepts such as the Theory of Performance Frontier and the concept of slack to re - in vestigate the relationship between load factor and firm performance, given the conflicting findings in extant literature. Although our hypothesized relationship between load factor and operational performance does not hold strong, its trend still indicates the correct direction of the relationship. The strong support for our hypothesized relationship between load factor and financial performance also reinforces the importance of re - investigating existing phenomenon through multiple theoretical lenses. Accor dingly, our 37 research reconciles these mixed findings in the literature, contributing to knowledge advancement in this field and providing references for similar future research. 1.6.2 Limitation and future research The limitation of our research bisects into two categories: utilization of variable and unit of analysis. Utilization of variable specifically refers to the variable of load factor itself. Unit of analysis refers to the fact that we focused on firm level d ata in our current research. Load factor, when measured as revenue passenger miles divided by available seat miles, has been criticized by extant researchers from two perspectives. From capacity utilization perspective, Baltagi et al. (1998) claimed that load factor only measures occupied seats relative to total miles flown so it ignores the utilization of the aircraft itself. From input - output analysis perspective, Schefczyk (1993) argued that load factor does not adequately capture and reflect overall operational performance due to the following reasons: 1) passenger load factor ignores other non - passenger inputs and/or outputs; 2) load factor does not consider other inputs besides the measurement of capacity; 3) load factor ignores differences in factor costs. To overcome the above - mentioned disadvantages, researchers have started to use aircraft utilization as the capacity measure in their research (Lapré and Scudder 2004; Ramdas and Williams 2008). In the current research, we focus on one spec ific area of aircraft utilization which is passenger load factor. We treat this as a limitation which at the same time also opens up an avenue for operational and fin ancial performance. There are three distinct units of analysis in extant airline research, which are firm level (Spiller 1983; Ramaswamy et al. 1994; Behn and Riley 1999; Shaffer et al. 2000; Mishina et al. 2004; 38 Collins et al. 2011; Atkinson et al. 2013; ), individual flight level (Bratu and Barnhart 2006; Ramdas and Williams 2008), and airport - flight level (Rupp and Holmes 2006; Dana and Orlov 2014). We focused our current research on firm level data analysis and accordingly we find it difficult to expl ain some phenomena associated with flight level analysis. For example, we found that both between effect and within effect of enplaned passengers have a positive impact on OTP, which is counter intuitive to normal senses as well as to extant findings in literature. Since th e number of passengers are reported at flight level, which was subsequently aggregated to firm level, what we have found can only be used to interpret what is going on at firm level. To this end, we did not have the ability to analyze data at flight level to further explore and explain the potential reasons behind this. We admit this as another potential limitation but at the same time, this also lends support for potential future research to investigate this specific relationship using flight level data. We also found that relative market share negatively impacts financial performance. In the marketing literature, Anderson et al. (1994) proposed that increasing market share might result in decreased customer satisfaction which in turn will lead to lower pr ofitability. However, the majority of marketing literature revealed a positive relationship between market share and financial performance (Szymanski et al. 1993). Further, management scholars have concluded that the relationship between market share and f inancial performance is context - specific (Prescott et al. 1986). Given these different findings, we conjecture that, similar to the inverted U - shaped relationship between load factor and financial performance, there also exists an inverted U - shaped relatio This serves as another different future research avenue. 39 supporting the fact that when a company operates close to its asset frontier, it will operate under the law of trade - offs while when a company operates away from its asset frontier, it will operate under the law of cumulative capabilities. Airline operations falls under the classic categorizatio n of service operations and we believe that our findings can also be generalized to other service operations to explore similar relationships, serving as another fruitful future research path. In sum, this study grounds its hypotheses in relevant scholars hip and leverages existing concepts and theories to examine the relationships between load factor, OTP, and financial performance by utilizing a between - within specification. Our findings strongly support our theorizing regarding the relationship between l existing operations management Theory of Performance Frontier, and, accordingly, making the theory and our findings generalizable to other service operations. The between - within specification open s up a new avenue in airline research to investigate minute relationships while achieving more precise and less unbiased parameter estimations as well as providing more insightful and impactful managerial recommendations. As such, we hope that our work can provide useful guidance for future relevant research in this area. 40 CHAPTER TWO BAGGAGE FEES AND FIRM PERFORMANCE 2.1 INTRODUCTION Amid the economic recession and soaring fuel prices of the late 2000s, major US airlines started to charge baggage fees as they struggled with costs associated with baggage operations (McCartney 2008a, 2008c). These major airlines claimed that the additional revenue generated from baggage fees allowed them to alleviate the impact of the economic recession and high fuel costs (McCa rtney 2010b); in the meantime, airlines also claimed that they were able to improve baggage operations and overall operational performance (McCartney 2008b, 2008c, 2010a, 2010b). Evidence regarding the effect of baggage fees is mixed. With regard to financial performance, both airline annual reports (Department of Transportation 2019) and airline research (Garrow et al. 2012, Schumann and Singh 2014) confirmed that baggage fees inc rease revenue. However, Yazdi et al. (2017) contended that the net effect of baggage fees is unknown, as it both increases revenue directly through extra fees and decreases revenue indirectly as a result of reduced consumer demand. With regard to operation al performance, such as on - time arrivals, the literature also presents mixed findings concerning the post - policy impact: there is evidence of a positive impact (Scotti et al. 2016, Nicolae et al. 2017), a negative impact (McCartney 2008a, 2010b, 2012), and a mixed impact, where an initial deterioration was followed by an improvement (Yazdi et al. 2017). With regard to consumer response, similar mixed results appear. While there has been tremendous coverage in the media of consumer outrage around new baggage fees (McCartney 2010b, Tuttle 2014, Elliott 2015), Scotti 41 et al. (2016) found no significant relationship between the new baggage fee policy and consumer response, when measured as consumer complaints. These mixed findings draw our attention. In this es say, we investigate the relationship between baggage fees and three outcome variables: financial performance, operational performance, and consumer complaints. We ground our reasoning on appraisal theory and the relevant literature; then we test this theor izing through the use of discontinuous growth modeling (Bliese and Lang 2016), which allows us to assess the relationship both immediately upon policy implementation and in the long term. We argue that upon the implementation of a baggage fee policy, airli nes will see a decrease in financial performance, an improvement in operational performance, and an increase in consumer complaints. Further, we hypothesize different polynomial relationships regarding the long - term impact on the three outcome variables. O ur hypotheses testing results support our predictions of the immediate impact on financial performance and operational performance; we also show that over time, financial performance, operational performance, and consumer complaints all demonstrate a decel erating positive trend. Our results reconciled the mixed finding in the current literature, as such, our study contributes to airline research in operations management in the following ways. This is the first study in baggage - fee - related research to draw on appraisal theory and relevant concepts to hypothesize non - linear relationships and, accordingly, to find support for these relationships. This study also uses a unique methodological approach: we adopt discontinuous growth modeling to investigate the po licy impact at two distinct stages to reconcile current mixed findings. Discontinuous growth modeling is specifically suitable to study policy impacts and gives researchers more flexibility to examine changes over time (Bliese and Lang 2016). This method h as been extensively explored in various fields, such as new organizational policies (Canato et al. 42 2013), organizational response to recession (Kim and Ployhart 2014), staff turnover effect (Hale et al. 2016), and new production systems (Parker 2003). But such an approach has not yet been applied to airline research. To the best of our knowledge, our research is the first to investigate baggage fee policy impact utilizing discontinuous growth modeling in the operations management field. Thus, our study pave s the way for future research in the field to leverage this approach to investigate the impact of external shocks and policy changes at more nuanced levels. Furthermore, our findings provide practical, empirically based guidance for airline decision makers and operations managers to design corresponding strategies. Our study is organized as follows. In section 2, we review the relevant literature and build our corresponding hypotheses. In section 3, we explain our data collection and variable construction process. In section 4, model specification and analysis are conducted to test our hypotheses. In section 5, several robustness tests are performed to validate our findings. In sections 6 and 7, managerial insights as well as theoretical contributions are d iscussed before limitations and future research are presented. 2.2 HYPOTHESES DEVELOPMENT In this section, we build our hypotheses, drawing on the findings from previous literature as well as relevant theories and concepts to explain the hypothesized relationships. Our hypotheses development also leverages the concept of discontinuous growth mo deling (Lang and Bliese 2009), which examines impacts of change at two different stages: the transition stage and the from the pre - change period are immediately 2009, p. 415). The recovery stage refers to the process of recovering following the immediate 43 - evaluate the applicability of alrea develop our hypotheses at two different stages: the immediate impact of baggage fees at the transition stage and the long - term impact of baggage fees at the recovery stage. To bett er depict the two - stage hypothesized relationships, we present all hypotheses graphically in Appendix C . 2.2.1 New Baggage Fee Policy and Carrier Financial Performance in the form of revenue, has been widely researched in the airline literature, the vast majority of which demonstrates a positive relationship. Garrow et al. ( 2012) studied the de - bundling of baggage fees among major US airlines from 2007 to 2009 and found that baggage fees contributed directly to the increase of ancillary revenues for both legacy carriers and low - cost carriers (excluding Southwest and JetBlue). In examining the impact of baggage fees on ticket price, Henrickson and Scott (2012) also observed that baggage fees have successfully helped airlines increase their revenues. Using DOT data from 2006 to 2010, Schumann and Singh (2014) also concluded that those carriers who charge baggage fees benefit from extra revenue mean Despite these findings, which suggest that revenue increases after new baggage fee policy implementation, DOT (2019) reported that revenue collection from baggage fees has declined over the years. Moreover, Yazdi et al. (2017) also called for more refined future research to 44 investigate the imp fees, (which increases revenue directly, but also decreases revenues through lower demand) is To reconcile these mixed findings regarding the relat ionship between new baggage fee policies and revenue, we focus on cognitive theory specifically, the concept of cognitive appraisal (Lazarus 1991, Scherer et al. 2001). According to cognitive theory as it is applied in airline s arise from the cognitive appraisal of ancillary fees that lead to we predict that consumers will develop cognitive appraisal of the new policy. Given checked baggage used to be a free service, consumers may develop negative emotions toward the new policy (Lazarus 1991). Related literature has observed that consumers have strong negative emotions of betrayal regarding baggage fees (Tuzovic et al. 2014). These em otions of betrayal may trigger consumers to develop coping behaviors, such as avoiding travelling by air. The avoidance of air travel can be explained from two perspectives. First, consumers may perceive baggage fees as unfair (Yazdi et al. 2017). When con sumers perceive prices as unfair, they tend to avoid purchases (Xia et al. 2004). Second, the avoidance of air travel is exacerbated in the north - eastern US, where alternatives, such as rail travel, are readily available (Morrison and Winston 2005). Moreov er, Scotti and Dresner (2015) confirmed that a one dollar increase in baggage fees leads to a loss of 0.7 passengers. Loss of passengers may translate to loss of revenue. Combining these arguments leads us to postulate that airlines are most likely to suff er from a loss in financial performance immediately upon policy implementation as a result of 45 H1a: Financial performance will decrease immediately in the transition stage of implementing baggage fees. To assess the long - term i mpact, we continue using cognitive appraisal theory to explain the expected relationships. As discussed, immediately upon policy implementation, consumers will develop coping behaviors based on their appraisals of the policy (Lazarus and Folkman 1984, Laza rus 1991, Scherer et al. 2001). This is what we designate as the initial stage appraisal. Refusal to purchase is one of the expected results in the initial stage appraisal. However, as more airlines gradually implement baggage fee policies, the progressive nature of appraisals (Lazarus expected that, as consumers kee p evaluating the policy, they will realize that refusal to purchase is futile and may affect their own interests, such as missing important client meetings. Consumers will adjust their actions by rationalizing their travelling utilities (Ben - Akiva and Lerm an 1985, Suzuki 2004). In other words, they will make travelling choices that give them the most benefit. This is what we designate as the second stage appraisal. To rationalize their trip utilities (Suzuki 2000) and coping behaviors at this stage, consume rs are more likely to start (i) travelling by air and (ii) paying for baggage fees. Such processes are not expected to happen overnight, given the continuous nature of appraisals. Consequently, as these gradual appraisals result in consumers starting to re purchase air tickets and pay for baggage fees, we predict that after the initial financial performance plunge and continuous decrease airlines will see a gradual improvement. 46 H1b: Financial performance will demonstrate a U - shaped curve in the recovery s tage. 2.2.2 New Baggage Fee Policy and On - Time Performance Among the various service dimensions in air travelling, on - time performance (OTP) is one of the indicator of - 019). OTP is a frequently researched construct in airline research, though with varying measures. One stream of Kalemba and Campa - Planas 2017 b ), while another stream of research measures arrival delays (Forbes 2008 b , Prince and Simon 2009, Cook et al. 2012, Ramdas et al. 2013, Mellat - Parast et The relationship between baggage fee policies and OTP has been investigated in airline literature with mixed findings. One stream of literature observes that OTP deteriorates after policy implementation, while other scholars find that OTP actually improves following baggage fee im plementation. Still other research claims that OTP initially deteriorates before it improves. We briefly review these three streams of literature in the following paragraphs. McCartney (2008a, 2010a, 2012) contends that OTP suffers due to a baggage fee po licy implementation for several reasons. First, after the policy implementation, more consumers carry on their luggage rather than checking it for a fee, resulting in an increased total number of carry - ons. Consequently, consumers fight for cabin storage s (McCartney 2012), which prolongs boarding time, delays flight departures, and subsequently 47 impacts on - time arrivals. Second, consumers not only avoid checked bags, they also begin to pack more into their carry - ons. As the ca - on is too bulky to fit (McCartney 2008a). This further prolongs the boarding process and potentially affects on - time arrivals. T station airline workers to screen bags at boarding gates. These workers identify bulky bags that might not fit in the cabin, consuming human resources and adding five to six additional minutes to the boarding process (McCartney 2012). This contributes further to worsened OTP. The second stream of literature uncovers the opposite findings, revealing that OTP actually improved following the policy implementation. Nicolae et al. (2017) investigated the impact of the new baggage fee policy on departure delays at route - flight level. They found that airlines who charge baggage fees witnessed a significant improvement in their departure performance following policy implementat ion, despite the increased number of carry - on bags, which might be expected to have a detrimental effect on departures. Nicolae et al. (2017) claimed that this is because the below cabin effect (ground handling of checked bags) outweighs the above cabin ef fect (cabin handling of carry - improved departures contribute to better on - time arrivals at destinations. Scotti et al. (2016), compiling panel data from 2004 to 2012, investigated the rel ationship between baggage fee policies and on - time arrivals at the carrier level and found that increases in baggage fees lead to an improvement in on - time arrivals. Combined, Nicolae et al. (2017) and Scotti et al. (2016) found that charging baggage fees is associated with improvement in airline OTP. Other findings also exist regarding the relationship between baggage fee policies and OTP. Yazdi et al. (2017) examined the impact of the new baggage fee policy on gate arrivals at carrier - 48 route level. Yazdi et al. (2017) found that from 2003Q3 to 2014Q4, on - time arrival performance initially deteriorated, then subsequently improved. Yazdi et al. (2017) further explained that in the initial stage of baggage fee implementation, the below cabin effect, which res ulted from a reduced number of checked bags and which resulted in shorter ground handling time, was not significant enough to offset the negative above cabin effect brought by an increased number of carry - ons, leading to boarding delays. As a result, the i nitial OTP deteriorated. But as time - bags may have been large enough to lead to improvements in airport - gradual improvement in OT P follows the initial deterioration in OTP. Our first hypothesis examines the immediate impact of the policy, building on the following three steps of reasoning. First, McCartney (2008a) reported that major airlines had been struggling with their baggage operations due to the overwhelming quantity of checked bags before the implementation of the new baggage fee policy. For example, American Airlines mishandled one bag for every 141 passengers in the first four months of 2008 (McCartney 2008a). Exacerbating the issue, American Airlines was ranked worst among all US airlines in - to efficiently handle checked - in baggage may be a factor of poor OTP. Second, appraisal t heory posits that consumers are likely to accumulate negative emotions toward baggage fee policies and accordingly develop coping behaviors, such as carrying on luggage to avoid baggage fees. Airline baggage literature also observes that passengers began t carrying on their luggage after policy implementation (Higgens 2010, McCartney 2008a, 2010a). As a result, a decline in checked baggage is expected immediately after policy implementation. Third, given the fact that airlines had b een struggling with handling too many bags and given our 49 theoretical prediction that the number of checked bags will decline immediately after policy from the i nflux of bags after policy implementation. Under this situation, the same number of ground handling staff should yield a higher efficiency due to the reduced number of bags. This should result in faster ground handing time and contribute to OTP improvement . H2a: OTP will improve immediately in the transition stage of implementing baggage fees. To build our arguments regarding the long - term effect of the new baggage fee policy on OTP, we continue to apply appraisal theory. First, as discussed, upon policy i mplementation, consumers will initially attempt to avoid the fees (Lazarus and Folkman 1984, Lazarus 1991, Scherer et al. 2001). However, as most airlines gradually adopt similar policies, consumers will realize that baggage fees are standard airline pract ice during their second stage appraisal. As a result, consumers are likely to change their behaviors to maximize their utilities for each trip (Suzuki 2000, 2004) by checking in luggage instead of struggling with the unpleasant boarding stampede (McCartney 2012). Consequently, we expect the number of checked bags to increase with time. Second, if the number of bags gradually increases, while the number of ground handling staff stays fixed in the near term (Bruno et al. 2019, Zeng et al. 2019), then the grou nd operations will have to handle more bags. As a result, the pace of the improvement of OTP may slow own and eventually begin to decline when ground handling staff are unable to keep up with the increase in checked bags. Third, we also argue that the incr ease in the number of checked will gradually adjust their second stage appraisals based on various externalities, such as stringent carry - on restrictions implem ented by airlines and repeated frustrations resulting from 50 fighting the cabin battle. As a result, we expect the rate of increase in checked bags to be non - linear, which will accordingly impact ground handling of baggage as well as OTP. H2b: OTP will cont inue to improve at a diminishing rate in the recovery stage. 2.2.3 New Baggage Fee Policy and Consumer Complaints Although not widely researched, the relationship between a baggage fee policy and consumer complaints also demonstrates mixed findings in the liter ature. Scotti et al. (2016) collected data from DOT for the period of 2004 to 2012 in order to investigate how charging baggage fees impacted consumer complaints. Using the number of complaints about baggage - related issues as the outcome variable and the f ee charged as the predictor, Scotti et al. (2016) were unable to find any significant relationship. Tuzovic et al. (2014) used survey data to conduct related research examining consumer response to airline price de - bundling. They found that consumers felt the strongest sense of betrayal about the baggage fee de - bundling, which had a direct impact on consumer complaints. As with previous hypotheses, we delineate the relationship between baggage fees and consumer complaints into an immediate impact upon poli cy implementation and a long - term impact over time. To explore the immediate impact upon policy implementation, we leverage appraisal theory as well as the halo effect of consumer complaint behavior (Halstead et al. 1996) to explain the expected relationsh ip. Appraisal theory suggests that once baggage fees have been de - associated with baggage fees, because they have to pay extra. The appraisal process will most l ikely result in the following three outcomes. First, consumers will assess the new baggage fees as unfair, given that the airlines charged the fees without adding any additional value to the 51 existing baggage service (Tuzovic et al. 2014, Yazdi et al. 2017) . The perceived unfairness will then engender complaints against the service associated with fees that were appraised as unfair (Zaltman et al. 1978, Tuzovic et al. 2014). Second, when airlines start to charge extra for baggage, consumers will accordingly develop expectations of receiving higher quality in baggage - related service (Forbes 2008 b ). However, given that airline ground operating procedures are normally standardized (Bazargan 2016), even with the extra fees gleaned from consumers, airlines are unl ikely to redesign their operating procedures to proactively improve their service quality. The service quality associated with baggage service, therefore, is not likely to improve coincident with the increase of baggage fees. Thus, the expectation of highe r service quality will not be met. Forbes (2008 b ) found that consumer complaints in the airline industry were driven by the gap between expectations and experienced quality levels. As a result, consumers will complain more on baggage - hey would have expected to b , p. 191). Third, complaining behavior will increase when the social climate (i.e., a social environment shared by a group of people) is favorable for complaining (Landon 1977, Halstead et al. 1996). This is especially true with baggage fee policy implementation, which has caused millions of consumers to develop a collective negative feeling (McCartney 2008a). This collective negative feeling will in turn trigger consumers to complain more, as predicted by appraisal theory (Lazarus 1991, Scherer et al. 2001). age - fee related issues, we expect that such complaints may also trigger complaints on other airline service - related issues. 52 Moreover, we expect that the implementation of a baggage fee policy might impact not only the ground handling of baggage, which is reflected in consumer complaints regarding baggage - related issues, but also the whole travel experience. If the implementation of a baggage fee policy impacts security check in, gate check in, boarding, and deplaning, it is also likely to impact consumer c omplaints in other service - related issues. Keeping these potential outcomes in mind, we expect that the total number of consumer complaints will increase immediately after the implementation of a baggage fee policy. H3a: Consumer complaints will increase immediately in the transition stage of implementing baggage fees. We also investigate the long - term impact of the baggage fee policy on consumer complaints by referring to the previously discussed two - stage appraisals. We argue that the initial stage of a ppraisal happens immediately after policy implementation, when consumers are triggered to complain more. The perceived feelings of unfairness (Yazidi et al. 2017) and betrayal due to baggage fees (Tuzovic et al. 2014) will motivate consumers to keep compla ining. However, as more and more airlines implement a baggage fee policy, it is expected that consumers will realize that baggage fees have become a standard part of airline pricing mechanisms. Accordingly, consumers will gradually develop the second stage appraisal of baggage fees and rationalize their travelling utilities, accepting the fee. Thus, we posit that the positive complaint trend against baggage - related issues will continue, but its rate will gradually diminish as consumers adapt. We further arg ue that the collective negative emotions, developed in the initial stage of appraisal immediately after policy implementation, will also diminish gradually in line e and 53 halo effect that engendered more complaints in other service areas will also be ameliorated, leading to fewer complaints regarding other service - related issues. H3b: Consumer complaints will continue to increase at a diminishing rate in the recovery stage. 2.3 DATA To test our hypotheses, we collected data from the U.S. Department of Transportation. Airlines with at least 1% of total domestic scheduled service passenger revenues are required to report their financial and operational performance to DOT. O ther carriers may also report their financial and operational performance voluntarily to DOT. At the time of accessing the DOT database, the October 2003, as well as to avoid the external shock of 9/11, we choose 2004Q1 as our data starting point and 2019Q2 as our data ending point, resulting in 16 years of data. Dependent variables and control variables were drawn from different data sources and were handled differen tly, depending on their data formats. This will be discussed in greater detail in this section. 2.3.1 Airlines A three - step data cleaning process was conducted to finalize the airline list. First, we tracked the name changes of airlines through the years and kept the most recent brand names to identify the unique airlines in our dataset. These airlines include Envo y (American Eagle until March 2003) and Endeavor (Pinnacle until December 2012). Other airlines, such as United Airlines and United Express, who simultaneously promoted two brands were grouped together as one carrier based on DOT report records. Second, we removed some ad hoc airlines who reported to DOT for only the short period of time during which they met the threshold of 1% of domestic 54 passenger revenues, such as Aloha (2006Q2 to 2008Q1), America West (2004Q1 to 2005Q4), and Independence (2004Q1 to 200 classify airlines reporting during the grace period following airline merger and acquisition. These airlines, the acquirer and target airlines, were still treated as two separate airlines until they officially reported jointly as one carrier to DOT. This three - step data cleaning process yielded a total of 18 airlines in our data set, with reporting records between 24 quarters and 62 quarters. A summary of the airline list appears in Table 6 . Table 6 Airlines in Dataset No. Airline First quarter in the sample Last quarter in the sample Total quarters in the sample 1 AIRTRAN 2004 Q1 2012 Q1 33 2 ALASKA 2004 Q1 2019 Q2 62 3 AMERICAN 2004 Q1 2019 Q2 62 4 ATLANTIC SOUTHEAST 2004 Q1 2011 Q4 32 5 COMAIR 2004 Q1 2010 Q4 28 6 CONTINENTAL 2004 Q1 2011 Q4 32 7 DELTA 2004 Q1 2019 Q2 62 8 ENVOY 2004 Q1 2019 Q2 54 9 EXPRESSJET 2004 Q4 2019 Q1 49 10 FRONTIER 2005 Q2 2019 Q2 56 11 HAWAIIAN 2004 Q4 2019 Q2 52 12 JETBLUE 2004 Q1 2019 Q2 58 13 MESA 2006 Q1 2019 Q2 33 14 NORTHWEST 2004 Q1 2009 Q4 24 15 SKYWEST 2004 Q1 2019 Q2 62 16 SOUTHWEST 2004 Q1 2019 Q2 61 17 UNITED 2004 Q1 2019 Q2 62 18 US AIRWAYS 2004 Q1 2013 Q4 40 2.3.2 Dependent Variables The immediate impact of implementing the new baggage fee policy is reflected in the potential immediate increase in revenue and subsequently in profit. Following current airline research (Dresner and Xu 1995; Tsikriktsis 2007; Mellat - Parast et al. 2015), w e elect to use OPOR absolute size differences among carriers. The relevant measures were taken from DOT Schedule 55 P12. Revenue and profit are reported on a quarte rly basis by DOT and are log transformed following current econometric research practice (Wooldridge 2010). - time and gate arrival time. We elect to use gate arr ival time for two reasons. First, DOT collects - time performance in various DOT databases. Gate departure time is only available at the airport - route - flight level, while gate arrival time is reported at different levels, including f light level, airport level, route level, and firm level. Our current research focuses on firm - level analysis. Therefore, gate arrival time is our focus. Second, Yimga (2017) found that on - - time arrivals are also directly related to complaints (DOT 2019) and indirectly related to revenue (Fornell et al. 1996) two of our outcome variables. Therefore, gate arrival time is more approp riate for our research question. This variable was taken from the DOT Monthly Consumer Report. To match financial measures, which are only Consumer complaints, our third dependent variable, have been studied in airline research in the form of complaint per 1,000 passengers (Steven et al. 2012) and total number of complaints toward flight, baggage, and overbooking (Halstead et al.1996). Since March 2002, DOT has classified consumer complaints into the following 12 categories: flight problem, over - sales, reservation/ticketing/boarding, fares, refunds, baggage, customer service, disability, advertising, discrimination, animals, and other. We elec t to use the total number of consumer complaints to capture the direct as well as the indirect impact of a baggage fee policy on consumer complaints. 56 Consumer complaints were also taken from the DOT Monthly Consumer Report. Consumer complaints were log tra nsformed in our analysis following existing practices in airline research (Lapré and Tsikriktsis 2006, Steven et al. 2012). 2.3.3 Independent Variables and Coding of Time Our three hypotheses aim to investigate both the immediate and long - term impact of a new ba ggage fee policy. Discontinuous growth modeling can be applied to capture these two different impacts (Bliese and Lang 2016, Bliese et al. 2017). The immediate impact can be estimated using a dummy variable, Transition , while the long - term impact can be mo deled by defining a time - related variable, Recovery (Bliese and Lang 2016, Bliese et al. 2017). In our dataset, Transition occurs in the quarter when the new baggage fee policy was implemented, and Recovery refers to the subsequent quarters thereafter. As such, we collected baggage fee implementation dates from the literature (Barone et al. 2012, Scotti et al. 2016, Yazdi et al. 2017, Zou et al. 2017) for coding purposes. Time , as the main independent variable, plays a crucial role, because how time is spec ified impacts the interpretation of the variables of Transition and Recovery . Since our main research interest is to investigate both the immediate and long - term impact of policy change, we elect to follow the relevant coding practices of Bliese and Lang ( 2016) to examine the changes in the value of dependent variables upon policy implementation as well as the changes in slopes of Recovery after policy implementation. Using a basic discontinuous growth model in equation 3 (adapted from Bliese and Lang 2016) and using Alaska airline as an example in Table 7 , we illustrate how Transition and Recovery are defined relative to the specification of Time . Carriers are denoted by i and measurement occasions are denoted by t in our equations. 57 Equation 3 Basic Discontinuous Growth Curve Model Table 7 Coding Time Using Alaska Airline as an Example Year Quarter Measurement Occasion Time Transition Recovery 2004 1 1 0 0 0 2004 2 2 1 0 0 2004 3 3 2 0 0 2004 4 4 3 0 0 2008 3 18 18 0 0 2008 4 19 19 0 0 2009 1 20 20 0 0 2009 2 21 21 0 0 2009 3 22 21 1 0 2009 4 23 21 1 1 2010 1 24 21 1 2 2010 2 25 21 1 3 2010 3 26 21 1 4 2010 4 27 21 1 5 2019 1 61 21 1 38 2019 2 62 21 1 39 Table 7 is Alaska Airline data extracted from our finalized dataset. The starting point of Alaska is 2004Q1 and the ending point is 2019Q2, resulting in a total of 61 measurement occasions. Alaska implemented its new baggage fee policy in 2009Q3. Therefore, the d ummy variable Transition becomes 1 in 2009Q3 and remains constant as 1 in the quarters thereafter. The Recovery variable codes the observation in 2009Q3 (i.e., the policy implementation quarter) as 0 and codes the remaining observations in their sequential in the column of Time start with 0 but become constant in 2009Q2 (i.e., one quarter before the policy implementation) to reflect the policy change. When Time is coded in this format, the intercept captures the value of the dependent variable at Time 0, which is the first measurement occasion. represents the slope before the 58 implementation of the new baggage fee policy (i.e., the pre - change slope). reflects the absolute change in the value of the dependent variable relative to 0 upon the implementation of the baggage fee policy. , the slope estimate, also represents the absolute change in slope relative to 0 after the policy implementation. As such, Time specification captures the absolute ch anges in both Transition and Recovery . In sum, our main independent variables are Time , Transition , and Recovery . To capture the hypothesized polynomial relationships, we also included the square terms of Recovery , denoted as Recovery.SQ , in our final anal ysis. 2.3.4 Control Variables Enplaned Passengers The number of enplaned passengers is a commonly used control variable in baggage fee and airline related research. Prince and Simon (2009) used the monthly total number of passengers at route level to control for market demand. Nicolae et al. (2017) calculated the expected averaged number of passengers at flight level as a control for consumer demand. Yimga (2017) aggregated the number of passengers to itinerary level to control for market demand. Because our u nit of analysis is at carrier level, we use the total monthly enplaned passengers of each carrier, as reported by DOT. This variable is taken from the DOT Air Travel Monthly Consumer Report. We include the number of passengers as a control variable for tw o main reasons. First, the number of passengers has a direct impact on revenue, as greater numbers of passengers translate to more revenue. Second, the number of passengers is expected to have an indirect impact on OTP and consumer complaints. A greater nu mber of passengers imposes greater operational challenges, such as checking in, boarding, deplaning, and baggage handling, all of which could lead to potential delays in OTP as well as more consumer complaints. 59 Total Delay DOT reports flight delays in ter ms of departure delays. Since October 2003, DOT has classified flight departure delays into five categories: air carrier delay, extreme weather delay, National Aviation System delay, security delay, and late arriving aircraft delay. Some of the baggage lit erature focuses on weather - related delays (Anderson et al. 2009, Ramdas et al. 2013, Nicolae et al. 2017), while other literature focuses on carrier - induced delays (Scotti et al. 2016). However, research has shown that, regardless of type, any delay will h ave a direct impact on consumer complaints (Forbes 2008 b ) and financial performance (Mellat - Parast et al. 2015). As such, we elect to include the total number of departure delays to reflect the holistic impact of flight departure delays on consumer complai nts and financial performance. Total delay is compiled from the DOT Air Travel Monthly Consumer Report. Number of Employees As part of airline scheduling practices, airline staff are assigned to each airport to service ground operations, gate operations, and flight operations (Ernst et al. 2004). The number of staff scheduled directly impacts our three outcome variables. First, employees, as the human capital of how efficiently the ground operating staff handles baggage and other operations directly impacts OTP. Third, line personne l, such as check - in agents, gate agents, and flight attendants, play important roles in shaping consumer complaint behavior through their employee - customer interactions (Anderson et al. 2009). DOT only reports the total number of full - time employees, but n ot every employee has a direct impact on the three outcome variables. Those employees who have direct impacts on the three outcome variables, such as pilots, flight attendants, check - in and gate agents, and ground operating agents, account for 85% of an ai 60 (DOT 2019). Hence, we take 85% of the full - time employees reported by DOT as a control Data Report. Market Share Market share is a commonly used market power. Market share also has relationships with our three outcome variables. First, the relationship between market share and firm financial performance has long been established (Buzz ell et al. 1975). Second, Suzuki (2000) found that carriers with better on - time performance enjoy greater market share. To account for the potential reverse causality, we use market share to control for its impact on OTP. Lastly, greater market share indic ates greater number of passengers. With everything else being equal, a greater number of passengers is likely to result in more complaints. Therefore, we included market share as a control variable. Market share has been operationalized several ways in ai rline research. Rupp et al. (2006) divided by the total number of scheduled flights on that route. Prince and Simon (2009) defined number of enplanements on the route divided by the total number of enplanements. Shaffer et al. (2000) calculated individual carrier market share as the ratio of a ger miles. We elect to follow Shaffer et al. (2000) to operationalize market share, because this operationalization calculates market share at carrier level, consistent with our unit of analysis. The related variables to construct market share are taken fr om Schedule T1 in DOT Form 41 Air Carrier Summary Data. 61 Low Cost Carriers (LCCs) DOT classifies carriers into Legacy Carriers and Low Cost Carriers (LCCs) and identifies six low cost carriers: Allegiant Air, Frontier, JetBlue, Southwest, Spirit, and Virgi n America, four of which (Frontier, JetBlue, Southwest, and Spirit) are in our dataset. LCCs are characterized by their point - to - point operating model and simplified fleet operations (i.e., they normally fly a single type of aircraft) (Mellat - Parast et al. terminology by using the term LCCs (Rupp et al. 2006, Yazdi et al. 2017, Yimga 2017), while 2006, Tsikriktsis 20 07, Mellat - throughout our research. Researchers have noticed the differing roles played by LCCs in all three outcome variables. With regard to on - time arrival performance, both Baker (2013) and Rupp et al. (2006) found that legacy carriers had overall better OTP than LCCs. Rhoades and Waguespack (2008), however, argued that LCCs grab more market share from legacy carriers by providing better OTP. In investigating the relationship between consumer com plaints and carrier profitability, Mellat - Parast et al. (2015) found that LCCs were affected less than legacy carriers in their profitability (2006) observation that L CCs learn to reduce consumer complaints more quickly than legacy carriers. Accordingly, we expect that LCCs and legacy carriers would handle consumer complaints differently. The relationship between baggage fee policy and LCC revenue provides the most inte resting findings in the literature. Garrow et al. (2012) found that Southwest, who 2009 as a result of consumers shifting from other carriers to avoid baggage fee s. Conversely, 62 Southwest, actually incurred revenue loss for Southwest. Given the distinctive roles of LCCs in all three outcome variables, we hereby include LCCs as another control variable. Other Macro - Economic Control Variables One of the most challenging problems empirical studies face in airline research is endogeneity bias (Scotti and Dresner 2015, Yazdi et al. 2017). The changes in OPOR, OTP, and cons umer complaints are also likely to be driven by other macro - economic variables. As such, endogeneity bias could yield a positive correlation between the variables measuring the changes in our three outcome variables and the error term. Therefore, we also i nclude selected macro - economic variables that could affect consumer travelling behavior to further address the concern of endogeneity bias. The first variable is the Smoothed US Recession Probabilities (Piger and Chauvet 2019), retrieved from Federal Reser ve Economic Data. The second variable is percentage change of GDP (Bureau of Economic Analysis 2019), also retrieved from Federal Reserve Economic Data. Lastly, we retrieved and compiled data from DOT to calculate the quarter - over - quarter change of average d domestic fares from 2004 to 2019. 2.3.5 Summary Statistics Table 8 provides a list of variables used in our analysis, their definitions, and their data sources. Those variables that are reported at a monthly level are subsequently collapsed into a quarterly level to match financial measures that are only available at a quarterly level. For relevant control variables, we construct the within effects to investigate their impacts at more nuanced levels (Bell and Jones 2015, Bell et al. 2018). A brief discussion of between and within specification follows in the next section. 63 Appendix D shows the summary statistics and correlation matrix between the three outcome variables and other variables. Following Bliese and Lang (2016), we keep the original form for all ti me - related variables. We also take the natural logarithm for relevant control variables following current airline literature (Garrow et al. 2012, Scotti and Dresner 2015, Yazdi et al. 2017) as well as standard econometric practice (Wooldridge 2010). Table 8 Variables Used in Analysis Variable Formula or Definition Data Source OPOR Airline's operating profit divided by operating revenue as reported each quarter. DOT Schedule P1.2 in Form 41 Financial Data On - Time Performance Monthly overall percentage of fights arriving within 15 minutes of scheduled arrival time. Table 9 in DOT Air Travel Monthly Flight Delays Report Consumer Complaint Total number of consumer complaints each month. Table 3 in DOT Air Travel Monthly Consumer Complaint Report Fuel Cost Total scheduled domestic fuel cost (Dollars) each month. DOT Schedule P12A in Form 41 Financial Data Enplaned Passengers Monthly enplaned passengers for each carrier. DOT Air Travel Monthly Mishandled Baggage Report Market Share The ratio of a carrier's quarterly revenue passenger miles to the sum of revenue passenger miles of the total carriers in that quarter. DOT Schedule T1 in Form 41 Air Carrier Summary Data Number of Employees Monthly number of FTEs for each carrier. DOT Schedule P1(a) in Form 41 Financial Data Total Delay The sum of delays caused by cancelled flights, diverted flights, aircraft delay, extreme weather delay, national aviation system delay, security delay, and late arriving aircraft delay. DOT Air Travel Monthly Consumer Report LCC Low cost carrier defined by DOT. DOT Air Travel Monthly Consumer Report Recession Smoothed U.S. Recession Probabilities. Federal Reserve Economic Data GDP % Change Quarter over quarter change in GDP. Federal Reserve Economic Data Fare % Change Quarter over quarter change in average domestic fares. DOT DB1B * Monthly data was aggregated into quarterly data to match the quarterly financial measures. 64 2.4 ANALYSIS AND RESULTS 2.4.1 Between and Within Specification In our longitudinal dataset, time - varying variables can be decomposed into two parts: one part that is relatively constant compared with other carriers and one part that varies within individual carriers over time (Bell and Jones 2015). The two parts accor dingly have their own effects in modeling 2015, Bell et al. 2018). For example, in our 16 - year dataset for carrier - controllable delays, AirTran has an average delay of 3.76%, wh ile in the same period American shows an average delay 5.95%. The difference in carrier averaged delays in these 16 years is referred to as a demonstrates fluctuatio ns around its mean delay of 3.76%. This fluctuation is accordingly modeling , two new variables for a time - varying covariate can be constructed: the between variable, using its grou p mean (Mundlak 1978) 3.76% for AirTran in our case and the within variable, using the deviation from its mean (Bells and Jones 2015, Hoffman 2015) the fluctuation around 3.76% in our case. To reflect the fact that the new baggage fee policy is mostly a wi thin - carrier effect, we constructed the within variables for all time - varying control variables at carrier level. 2.4.2 Model Testing Procedures Following Bliese and Lang (2016) and Bliese et al. (2017), we build our discontinuous growth models in four differen t stages. Stage 1 builds a random intercept model to calculate the Intraclass Correlation (ICC). Stage 2 models time in discontinuous growth models, as discussed in the previous section. Stage 3 selects the appropriate random effects by comparing model fit in terms of different random effect specifications. Stage 4 is the final step, where the finalized 65 models are used to test hypothesized relationships. All analyses were conducted in Stata/SE Version 15.1. We start our analysis by building a random interc ept model to calculate the ICC. ICC in a longitudinal dataset refers to the proportion of variances that is between groups in mean differences because of the random intercepts (Hoffman 2015). An ICC of 1 indicates that the data is not longitudinal, in whic h case cross - sectional analysis will be sufficient to answer the research questions. On the other hand, an ICC of 0 means that the data is purely longitudinal. We use a random intercept model to calculate ICC for our three dependent variables. The results are reported in Table 9 . Table 9 Random Intercept Model to Calculate ICC Parameter OPOR OTP Complaint Fixed Effect Intercept 0.04** (3.23) 0.78** (79.64) 4.08** (16.39) Random Effect Level 2: Carriers 0.0028 0.0017 1.11 Level 1: Occasion 0.0079 0.0023 0.40 Measures of Fit - 2 Log Likelihood - 1697.04 - 2762.63 1777.32 ICC 0.26 0.42 0.73 - tailed). Z - tests are reported in parentheses for the fixed effects parameters. ICC for OTP as the dependent variable model is 0.42, indicating that 42% of variation is between carriers, while the remaining 58% of variation is within carriers. The between carrier variation is 26% and 73% respectively for OPOR and complaint as dependen t variables. The ICC result corroborates what we have observed from our dataset in that OPOR is mainly an airline idiosyncratic characteristic, while OTP and consumer complaints show heterogeneity due to the different firm sizes among carriers. 66 Since our main research interest is to investigate the immediate and long - term impact of a baggage fee policy on the three outcome variables, the Time and Recovery variables become our focal variables. After adding all control variables, we use the finalized model i n equation 4 to test our hypotheses. is one of our three outcome variables for carrier i in quarter t . is a vector of control variables that have been discussed in the previous section. To correctly specify the random effects of Time , Transi tion , and Recovery for our three outcome variables, we also test the corresponding random effects, the results of which are reported in Table 1 0 . The Likelihood Ratio Test (LRT) is used to compare model fit. Based on the LRT results, the random effects for the OPOR model are random intercept and random slope of Time; the random effects for the OTP model are random intercept only; and the random effects for the Complaint model are random intercept and random slope of Time. Table 10 Select Model Random Effects Random Term Model df - 2 LogLik AIC BIC Test L.Ratio p Value OPOR Intercept 1 20 - 2106.09 - 2066.09 - 1971.66 Time 2 23 - 2118.49 - 2072.49 - 1963.90 1 vs 2 12.40 0.006 Time + Tran 3 26 - 2119.59 - 2067.59 - 1944.83 2 vs 3 1.10 0.778 Time + Tran + Recov 4 - - - - - - - OTP Intercept 1 20 - 3117.43 - 3077.43 - 2982.99 Time 2 22 - 3118.16 - 3074.16 - 2970.29 1 vs 2 0.74 0.692 Time + Tran 3 25 - 3122.29 - 3072.29 - 2954.26 2 vs 3 4.13 0.248 Time + Tran + Recov 4 - - - - - - - Complaint Intercept 1 19 815.34 853.34 943.05 Time 2 21 803.19 845.19 944.35 1 vs 2 12.14 0.002 Time + Tran 3 - - - - - - - Time + Tran + Recov 4 - - - - - - - 67 Equation 4 Model Specification 2.4.3 Hypotheses Testing Results The hypotheses testing results are summarized in Table 1 1 . In the final hypothesis testing models, we also allow residuals that are one measurement occasion apart to be correlated (AR1). Our first set of hypotheses predicts that financial performance will decrease immediately upon policy implementation (H1a), but over time, financial performance will demonstrate a U - shaped curve (H1b). The immediate impact is tested by the coefficient of Transition with a parameter estimate of 0.026 ( p = 0.077 ), supportin g H1a and indicating that OPOR decreased by 0.026 units immediately upon policy implementation. The long - term impact is jointly tested by the coefficients of Recovery (0.005, p = 0.004 ) and Recovery.SQ ( 0.00009, p = 0.041 ), which are both significant but show signs opposite to what has been hypothesized, indicating an inverted U - shaped relationship. So H1b is not supported. To better visualize the hypotheses testing results, we have plotted the results in Figure 3 by taking the mean values of all significa nt control variables. To make the plot easier to read, we use American Airline as an example (the exclusion of other carriers does not impact interpretation of hypotheses testing results). The horizontal axis represents quarters in our dataset. The vertica l axis represents OPOR. 68 Table 11 Final Model to Test Hypotheses Parameter OPOR OTP Complaint Fixed Part Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Intercept 0.013 (0.87) 0.041 (1.65) 0.78** (75.87) 0.78 ** (49.20) 3.50** (13.62) 3.41** (26.18) Time 0.0003 (0.22) - 0.002 ( - 1.14) - 0.0019** ( - 2.91) - 0.016 ( - 1.47) 0.05** (6.13) 0.026** (2.81) Time.SQ 0.00005 (1.63) 0.00005 (1.34) - 0.00016 (1.05) 0.00003 (0.97) - 0.0002 ( - 1.03) - 0.0003 ( - 1.36) Transition - 0.04** ( - 3.10) - 0.026 ( - 1.77) 0.041** (6.14) 0.028** (3.49) - 0.34** ( - 4.20) - 0.03 ( - 0.31) Recovery 0.0006** (5.17) 0.005** (2.90) 0.0012 (1.78) 0.0023 * (2.15) 0.045** (5.54) 0.042** (4.40) Recovery.SQ - 0.00008 ** ( - 2.88) - 0.00009 * ( - 2.05) - 0.00003 ( - 1.59) - 0.00005 * ( - 1.96) - 0.0008** ( - 3.92) - 0.001** ( - 4.69) Fuel Cost 0.013* (2.20) 0.002 (0.06) 0.09** (3.12) No. of Passengers 0.19** (6.73) 0.019 (1.20) 1.29** (8.32) Total Employees - 0.18 ** ( - 5.56) - 0.04* ( - 2.09) - 0.84** ( - 4.95) Market Share 0.51 (1.90) - 0.03 ( - 0.19) - 1.03 ( - 0.72) Total Delay - 0.023 ( - 1.86) 0.79** (11.70) LCC 0.059* (1.99) 0.013 (0.70) 1.01** (6.62) Recession - 0.014 ( - 1.09) - 0.006 ( - 0.92) 0.002 (0.03) GDP % change - 0.004 ( - 0.83) 0.0063** (2.49) 0.019 (0.67) Fare % change - 0.061* ( - 2.55) - 0.005 ( - 0.39) 0.076 (0.58) Carrier fixed effects: YES Random Part Level 2: Carriers 0.00284 0.002052 0.00148 0.006957 1.10 1.083153 Level 1: Occasion AR1 0.437504 0.482750 0.402528 Variance 0.0692 0.006050 0.00203 0.001653 0.306 0.175371 Measures of Fit - 2 Log Likelihood - 1815.14 - 2000.26 - 2878.19 - 3065.71 1540.37 873.27 AIC - 1799.14 - 1964.26 - 2862.19 - 3031.71 1556.37 909.27 BIC - 1760.95 - 1879.27 - 2824.00 - 2951.44 1594.56 994.26 R 2 (MVP) 40.20% 45.09% 87.65% Total R 2 20.00% 48.52% 88.91% = p < 0.10; * = p < 0.05; ** = p < 0.01 (two - tailed). Z - tests are reported for fixed effects parameters. 69 Figure 3 Hypothesis 1 OPOR Graph (American Airline as an Example) Our second set of hypotheses posits that OTP will improve immediately upon policy implementation (H2a), but over time, OTP performance will decline at a diminishing rate (H2b). The coefficient of Transition is 0.028 and is significant ( p = 0.000 ), indicati ng that upon policy implementation OTP did improve by 0.028 points (i.e., 2.8%); so, H2a is supported. The long - term impact is again jointly tested by the coefficients of Recovery (0.002, p = 0.031 ) and Recovery.SQ ( - 0.00005, p = 0.050 ), both of which are significant. In a quadratic change model, instantaneous p. 226). The interpretation for the coefficients of Recovery and Recovery.SQ is as follows: after policy implementati on, the linear time slope will immediately increase by 0.002 point at post - policy measurement occasion 1 (time 0). Since the fixed quadratic time slope is negative ( - 0.00005, p = 0.050 ), it consequently creates a decelerating positive trajectory such that the 70 positive linear rate of change will become less positive per occasion, starting from post - policy measurement occasion 2 (time 1), by twice the quadratic time slope of 0.00005. In other words, the linear rate of change of 0.002 point at post - policy sess ion 1 will become less positive by 0.0001 point per each occasion thereafter. Combined, the hypothesis testing result supports H2b. The relationship is graphed in Figure 4 using Frontier Airline as an example. Figure 4 Hypothesis 2 OTP Graph (Frontier as an Example) Turning our attention to the third set of hypotheses, consumer complaints are predicted to increase immediately upon policy implementation (H3a), but over time, complaints are expected to continue increasi ng at a diminishing rate (H3b). As with our previous hypotheses, the immediate impact is tested by the coefficient of Transition and the long - term impact is jointly tested by the coefficients of Recovery and Recovery.SQ . The coefficient of Transition is 0 .03 ( p = 0.759 ). So, H3a was not supported. The coefficients of Recovery and Recovery.SQ are 0.04 ( p = 0.000 ) and 0.001 ( p = 0.000 ) respectively, implying an inverted U - shaped relationship. The 71 linear time slope will increase by a 0.04 unit immediately a t post - policy measurement occasion 1 (time 0). The negative time quadratic slope, however, creates a decelerating positive trajectory such that the positive linear rate of change will become less positive per occasion by 0.002 point (twice the quadratic ti me slope of 0.001) thereafter. So, H3b is fully supported. The relationship was also plotted in Figure 5 using Frontier Airline as an example. Figure 5 Hypothesis 3 Complaint Graph (Frontier as an Example) 2.5 ROBUSTNESS TEST Severa l tests were conducted to assess the robustness of our findings. First, following Bliese and Lang (2016), we specified time in its relative format. Then we compared the results of the relative time specification with the results from our previous hypothese s testing section. This robustness test was conducted for all three outcome variables. Comparing the two different Time specifications, we see that except for the coefficients of Transition , the two sets of models yield exactly the same parameter estimatio ns. The only difference is in the coefficients of Transition . 72 This is because relative time specification estimates Transition as the difference between expected value (derived from Time) and observed value, while our previous specification estimates Transition as the absolute change upon policy implementation. Both specifications estimate the recovery slope relative to zero; therefore, we observed identical coefficients for Recovery and Recovery.SQ . In sum, change of Time specification does not impact our hypotheses testing results. Our next robustness test excluded airlines with fewer measurement occasions from our analysis to see if this affected our results. Five airlines with fewer than 35 quarters of observations (Airtran, Atlantic Southeast, Con tinental, Mesa, Northwest) were removed from the original data. The same analysis was conducted in this new dataset as was done in Table 11 . Except for a slight change in parameter estimations, the significance levels and signs of all variables remained th e same. We also extracted a smaller sample, which consists only of five major airlines that span the entire 62 quarters (Alaska, American, Delta, SkyWest, and United), to conduct the same analysis. The Transition and Recovery parameter estimates for all th ree outcome variables still had the same signs, although the associated significance levels differed slightly. In addition, we also trimmed down our data to three years before and three years after policy implementation (from 2005 to 2011) and tested the m odels again. Both Transition and Recovery parameter estimations for all the three outcome variables, although slightly different, were still statistically significant with the same sign. From methodological perspective, we run two different robustness tes ts. First, given the limited number of our higher - level entities (carriers in our case), we also conducted an analysis allowing only the intercept to vary randomly for each carrier. The estimates of the coefficients of Recovery and Recovery.SQ remained unc hanged and that of Transition slightly differed. The 73 significance levels of the associated coefficients are consistent with our previous hypotheses testing results. Second, we removed Southwest from our dataset and rerun the analysis. In our initial analys is, Southwest, the carrier who has not implemented the baggage fee policy, was also included in our data to construct a multiple - arm design to enhance the validity of our findings (Hoffman 2015). By removing Southwest from our data, we essentially have a s ingle - arm design. The hypothesis testing results for the single - arm design are almost the same as that of the multiple - arm design. Overall, our robustness tests show that our findings are neither affected by changing data structure nor affected by changin g model specifications. 2.6 MANAGERIAL INSIGHTS 2.6.1 For Strategic Decision Makers In 2008, airlines introduced baggage fees as a means to improve their financial situation (McCartney 2010b). Did this policy effectively improve the financial situation for airlines? OPOR actually dropped a 0.026 unit upon the policy implementation. Over t ime, the long - term impact indicates an inverted U - shaped relationship, indicating a decelerated trend of performance total quarterly performance, the combined i mmediate and long - term trend provides a clear implication for airline decision makers: de - bundling ancillary fees as a means to improve financial situation does not seem to be sustainable in the long run because OPOR starts to face a diminishing return 3.5 years into the policy implementation. A potential explanation for this may be that when airlines de - bundled the ancillary fees, airlines also gradually reduced their airfares to avoid angering customers (Jenkins et al. 2011). Henrickson and Scott (2012) a nd Scotti and Dresner (2015) confirmed a small but significant negative impact of baggage fees on airfares. As 74 result of this, in the long run, the revenue loss incurred by the reduction in airfares may have exceeded the extra revenue generated from chargi ng baggage fees. Therefore, the improvement in financial performance starts to decline after 3.5 years. When considering charging other ancillary fees in the future, airline decision makers should be prepared for a potential deterioration in financial perf ormance immediately upon the policy implementation. In addition, despite that charging ancillary fees may bring a few years of improvement in financial performance, airline decision makers should also be aware that sooner or later the improvement will star t facing a diminishing return. As a result, they should look in other areas to achieve a sustainable long - term financial performance improvement. Another strategic decision - making process relates to consumer behavior consumer complaints in our case. The a irline industry is a consumer - centric industry, where competition for customers is tight, and excellent service is key for sustainable business growth (Ostrowski et al. 1993, Park et al. 2006, Hussain et al. 2015). However, a comparison before and after po licy implementation reveals that consumer complaints kept increasing both before and after the policy implementation even though the total number of enplaned passengers remained steady. For airline decision makers, the overall question to ask is how to red uce consumer complaints by enhancing overall customer service quality, given that consumers are expecting higher service levels as they have paid extra (Forbes 2008 b ). 2.6.2 For Operations Managers The implications for operations managers are clear. First, our hypotheses testing results regarding OTP indicate that the below cabin effect is impacted by the amount of baggage. Although DOT does not report the exact number of checked bags, our theoretical arguments and airline literature (McCartney 2008b, Nicolae et al. 2017) suggest that, following the baggage fee 75 implementation, consumers checked fewer bags. As a result, both departure performance and on - time arrivals improved, due to the decrease in work required when handling fewer bags. But as the amount of bagg age increases, ground operations struggle again with more bags, which leads to worsened OTP. Operations managers need to take the number of bags into serious consideration when scheduling ground operations, because the number of bags does have an impact on OTP. Second, LCCs and legacy carriers have different OTP performance trends. Although the baggage fee policy has similar short - term and long - term impacts on both groups, our findings suggest that LCCs demonstrated a consistently lower OTP relative to lega cy carriers, in line with Rupp et al. (2006) and Baker (2013). Operations managers, thus, should notice the differing impact of late arrivals on different carriers. Third, although baggage fee policies have been implemented since 2008, the boarding stamped e and cabin battle are still on - going, gate workers are still stationed at boarding gates to screen bulky bags, and flight attendants still spend time removing oversized carry - ons from the planes to be checked (McCartney 2012). All these issues hamper the boarding process and accordingly affect OTP. Operations managers still have a long way to go to optimize the whole boarding process and ultimately achieve a sustainable, improved OTP. 2.7 CONCLUSION 2.7.1 Theoretical Contributions Our research contributes to know ledge accumulation in the following ways. In regard to methodology, we leverage discontinuous growth modeling to delineate the impacts of new baggage fee policies in both immediate and long - term effects. This allows us to explore hypothesized relationships at more nuanced levels and reconcile the mixed findings in the 76 previous literature. Discontinuous growth modeling in this sense offers an alternative approach to examining policy impact in the operations management field. Our findings also demonstrate th e importance and necessity of a nuanced analysis of immediate and long - term effects when discussing the impact of baggage fee policies. First, despite overwhelming findings in the extant literature about the positive policy impact on revenue (Henrickson an d Scott 2012, Garrow et al. 2012, Schumann and Singh 2014), our findings suggest that when OPOR (reported by DOT) was used as the outcome variable, the airlines actually suffered from a loss in financial performance coincident with policy implementation. I n the long run, the revenue generated by baggage fees does not seem to be an effective means to sustain financial performance improvement. Since most of the previous literature did not distinguish between revenue generated from baggage fees, operating reve nue, and total revenue, our findings on OPOR, to this end, extend the previous literature. Second, as regards operational performance, our theory predicts that on - time arrivals improved immediately upon policy implementation and kept improving until it fa ces a diminishing return in about four years (i.e., approximately in late 2013). Our findings suggest a curvilinear relationship which has not yet being explored in the current literature. If viewed in the first few years after the policy implementation, a s with Scotti et al. (2016) and Nicolae et al. (2017), OTP indeed improved. However, Scotti et al. (2016) examined the policy impact on OTP up to 2012, while Nicolae et al. (2017) studied the same impact up to 2009 in both cases OTP still showed the trend of improving based on our findings. Our research, by utilizing time as a continuous variable and by including more years of data, shows that OTP starts to deteriorate since late 2013. We therefore contribute to current relevant airline research by revealin g a long - term effect that has not been investigated so far. 77 consumer complaints steadily increased over the years both before and after policy implementation, despite the relatively constant number of passengers year over year. To this end, complain more after feeling betrayed by airlines. Lastly, discontinuous growth modeling is discontinuous growth m odeling can be readily applied to many other research topics in the operations and supply chain management field, such as the impact of new policies, mergers and acquisitions, supply chain interruption, and new production systems. To this end, our research serves as the foundation for similar future research. 2.7.2 Limitation and Future Research Like all research, our research demonstrates a few limitations. First, in our discontinuous growth modeling, we modeled the first baggage fee policy as the Transition va riable. However, many carriers implemented multiple baggage fee policies from 2008 to 2012 (Barone et al. 2012, Yazdi et al. 2017). So, it would be interesting to see the effects of multiple transitions and multiple recoveries by modeling multiple baggage fee policies. We observe this as a limitation, but, at the same time, it also serves as a potential topic for future research. Our next limitation is that our unit of analysis is at carrier level, the financial data of which is only reported quarterly. A ccordingly, we aggregated all relevant variables at quarterly level. Quarterly level data allowed us to conduct only one baggage fee policy analysis instead of 78 conducting a multiple baggage fee policy analysis. This is because to estimate slopes associated with the pre and post event, at least three pre - event measurement occasions are required (Bliese et al. 2017). In our case, at least three quarters before the policy change were required. Our quarterly data format unfortunately could not meet this require ment for multiple baggage fee policies. We treat this as another limitation, but it also serves as another fruitful future research avenue to use flight level data to further explore multiple baggage fee policy impacts. Our last limitation lies in our the oretical reasoning regarding the amount of baggage. As with the extant literature (Nicolae et al. 2017), we argue that the amount of baggage impacts operational performance to a great degree. However, DOT does not report the exact number of bags that were checked in and carried on. If the exact data were available, the number of bags could be linked both to revenue generated from charging baggage fees and to operating revenue, allowing for a more precise investigation of the impact of baggage fees on carrie r financial performance. In conclusion, we leverage appraisal theory to reconcile mixed findings in baggage fee literature by hypothesizing a two - stage impact. We test our hypothesized two - stage relationships by adopting discontinuous growth modeling to investigate both the immedi ate and long - term impacts of the new baggage fee policy. We specified both time and the policy change as the main independent variables, in contrast to the previous literature, which relies only on dummy variables or the actual amount of baggage fees to in vestigate the impact of policy change. Our results validate the importance of modeling the policy impact at two stages, reveal nonlinear relationships for the long - term impact, and provide alternative explanations to the mixed findings in current literatur e. As such, we hope our research can provide useful guidance for future studies that investigate the impact of external shocks and interventions in the field of operations and supply chain management. 79 CHAPTER THREE MERGER AND ACQUIS I TION AND FIRM PERFORMAN CE 3.1 INTRODUCTION meta - analysis of Datta et al, 1992; King et al., 2004; Homberg et al., 2009). Since deregulation in 1978, the U.S. airline industry has experienced numerous mer gers (Singal, 1996a; 1996b; Department of Transportation, 2019). Most research on airline mergers has focused on the impact of mergers on fares (Borenstein, 1985; Werden et al., 1991; Kim and Singal, 1993; Peters, 2006; Luo, 2014; Carlton et al., 2019). So me research has investigated merger - induced cost synergies and/or revenue synergies (Iatrou and Oretti, 2007; Merkert and Morrell, 2012; Schosser and Wittmer, 2015). More recent research has examined the impact of mergers on service quality, such as on - tim e arrivals (Steven et al., 2016; Prince and Simon, 2017; Rupp and Tan, 2019) and mishandled bags (Steven et al., 2016). However, recent research has been inconsistent in specifying merger event windows and has resulted in different merger findings (Hüsche lrath and Müllera, 2014; Steven et al., 2016; Prince and Simon, 2017). Additionally, most research has largely ignored whether the impact of attributes (Hitt et a l., 1998) and prior quality of the acquirer (Banaszak - Holl et al., 2002). Moreover, the focus of airline research has largely concerned with consumer welfare (Borenstein, 1985; Knapp, 1990; Werden et al., 1991; Kim and Singal, 1993; Morrison, 1996; Peters, 2006; Gong and Firth, 2006; Luo, 2014; Hüschelrath and Müllera, 2015; Shen, 2017) while ignoring carrier welfare. The purpose of this study, thus, is to 1) address the inconsistent specification of merger event windows; and 2) examine potential idiosyncra tic carrier characteristics that may impact post - merger performance. Further, this study focuses on two 80 outcome variables: on - time performance (OTP), measured by on - time arrivals; and financial performance, measured by revenue. Focusing on OTP and financia l performance helps to achieve the third purpose of this research, i.e., to explore the impact of mergers on both consumer welfare and carrier welfare. Our study is important given recent findings. First, Steven et al. (2016) found that mergers caused OTP to deteriorate following mergers. Conversely, Prince and Simon (2017) found that OTP first remained unaffected but then improved following mergers. Given these conflicting results, further examination appears necessary. Second, airlines often promote reve nue increase as the key benefit of mergers to their stakeholders and regulators (Carey, 2005; Delta, 2008). rast makes examining revenue - merger performance has long been observed as a strong indicator of merger success in the broader merger literature (Hansen and Wernerfelt, 1989; Heron and Lie, 2002; Zott, 200 3; Haleblian et al., 2009), but has never been considered within the airline industry. Drawing on organizational learning framework, we build a discontinuous growth model (Bliese and Lang, 2016) to study the impact of mergers at two distinctive stages. St age I (Transition) is expected to demonstrate immediate performance fluctuations during the initial period of the resource combination, reallocation, and utilization between the two merging carriers. OTP is predicted to decline while revenue is predicted t o increase during this stage. Stage II (Recovery) is expected to witness gradual performance changes following mergers as the newly merged carrier continues to manage the challenges of post - merger integration. OTP is hypothesized to demonstrate a U - shaped curve while revenue is expected to experience diminishing returns 81 during this stage. Lastly, we expect performance during the transition period and the recovery - merger performance. To examine these hypothesize d relationships, we collect data from the Department of Transportation (DOT) and study eight mergers from 2004Q1 to 2019Q2. During the transition stage, although we do not find support for the immediate OTP deterioration, our results indicate that revenue increased immediately as hypothesized. More importantly, with regard to OTP, low - performing acquirers experienced an improvement in OTP at this stage while high - performing acquirers do not experience any improvement in OTP; with regard to revenue, low - perf orming acquirers witnessed a significant increase in revenue at this stage while high - performing acquirers suffered from a revenue loss. During the recovery stage, we do not find any support for the two hypothesized polynomial long - term relationships. Over all, our results reveal that in the airline industry, the effect of mergers are short - term rather than long - term. By testing these key theoretical propositions, this study contributes to knowledge accumulation in the airline merger literature in the follo wing ways. First, our modeling approach allows us to assess a two - stage impact of airline mergers by utilizing time as a continuous variable, which enables us to provide more nuanced analysis of the impact of mergers. Second, the moderating effect of acqui - merger performance during the transition stage provides practical guidance for airline managers considering mergers. Third, this research responds to calls in the airline merger literature to investigate nonlinear relationships of post - merger imp act (Steven et al. 2016), to explore more recent U.S. airline mergers (Hüschelrath and Müllera, 2014), to consider service quality issues (Vaze et al., 2017), and to examine synergy realizations (Schosser and Wittmer, 2015). Lastly, in contrast to recent research where only a single merger was 82 investigated (Bilotkach, 2011; Luo, 2014; Hüschelrath and Müllera, 2014; Manuela et al., 2016), this study includes multiple mergers, including low - cost carrier mergers. 3.2 LITERATURE REVIEW 3.2.1 Overview of Research on Merger discipline - resources, with mixed findings (Cartwright and Schoenberg, 2006, p . S1). Although a handful of studies have explored the motives for merger (Amihud and Lev, 1981; Seth et al., 2000; Nguyen et al., 2012), the majority of merger research instead focuses upon exploring the performance implications of merger and antecedents of post - merger performance (Datta et al., 1992; King et al., 2004; Homberg et al., 2009). Datta et al. (1992) reviewed 41 merger studies to explore post - merger performance implications by examining five different antecedents that were frequently used in m erger literature. Datta et al. (1992) consequently found that shareholders in the acquiring firms do not benefit from mergers while shareholders in the target firms gain significantly. King et al. (2004) also conducted a meta - analysis of 93 merger studies, focusing on four antecedents to post - merger - - al., 2004, p. 196). Homberg et al. (2009) performed a meta - analysis on 67 merger studies, focusi ng on how related acquisitions impact post - merger performance. Homberg et al. (2009) 83 found that different forms of relatedness exhibit different impacts on merger success, moderated by knowledge intensity, absolute size of acquisitions, and geographic regi on. King et al. (2004) attributed mixed financial outcomes of merger to the following three reasons. First, scholars have used different variables to explain post - merger performance, hindering knowledge accumulation. Second, researchers have largely adopt ed event study methodology where only a short event window was considered, while the actual impact of merger may be more prolonged. Third, despite meta - analysis suggesting the existence of moderating effects, the extant merger research has not identified t concerns, we do not aim to develop new variables to explain post - merger performance. Rather, we focus upon investigating the post - merger performance using alternate event windows operationalized in both short - term and long - term, in an effort to reconcile past mixed findings. Additionally, we aim to test moderators which may help explain variations in post - merger performance. 3.2.2 Merger in the U.S. Airline Industry Background Information Although airline merger predates industry deregulation in 1978 (Lichtenberg and Kim, 1989), the majority of U.S. airline merger took place after deregulation. These mergers can be broadly classified into two waves. The first wave occurred in the 1980s shortly after deregulation, with a second wave starting in the late 1990s. The first wave was marked by the following characteristics. First, the number of mergers in this period was very high, with 27 mergers between 1985 and 1988 (Singal, 1996a; 1996b). Second, there were two diffe rent types of mergers during this period. Some of the mergers were between pairs of small carriers, such as the Braniff Florida Express merger in 1988 while other mergers were between mega carriers and small carriers, such as the American Air Cal merger in 1987. Lastly, this first wave of mergers 84 witnessed repeated mergers of a single carrier within a short period of time. As an example, Piedmont merged with Empire in 1986, and again with US Air in 1988. The second merger wave began in the late 1990s and d iffered from the previous wave in a few ways. First, there were fewer mergers, with only 20 mergers in the 20 years between 1999 and 2019. Second, mega carriers started to merge with each other, such as Delta Northwest in 2009, United Continental in 2010, and US Air American in 2013. As a result of these mergers, only three legacy mega - carriers remain Delta, United, and American (DOT, 2019). In responding to the call of examining more recent U.S. airline mergers (Hüschelrath and Müllera, 2014) as well as to provide more current managerial insights, we focus on this second wave of mergers. 3.2.3 Performance Implications of Merger in the U.S. Airline Industry Generally speaking, performance implications of U.S. airline merger have been studied in terms of fare, f light frequency, stock market response, on - time performance, and market competition effect (see Appendix E for a detailed summary). In this study, we focus on both on - time performance (measured by on - time arrivals) and financial performance (measured by re venue) to simultaneously investigate the impact of mergers on consumer welfare and carrier welfare. We examine on - time arrivals for the following two reasons. First, for consumers, timeliness of service in the airline industry is one of the most important dimensions of air travel service quality (Chen and Gayle, 2019). Historical on - (McCartney (2010), p. D1). Se cond, for researchers, past evaluation of the impact of mergers on OTP presents inconsistent findings (Steven et al., 2016; Prince and Simon, 2017), making further exploring on this topic interesting. Moreover, Richard (2003) and Vaze et al. (2017) noted t hat 85 past airline merger research has overwhelmingly focused on fare changes. Therefore, Richard (2003) and Vaze et al. (2017) subsequently called for future research to explore more service quality issues, such as on - time performance. We use revenue to mea sure financial performance based on the following findings. First, airlines commonly tout revenue increase as a key benefit of merger. U.S. Airways in 2013 estimated an additional annual revenue of $150 - 200 million in their press release when merging with American West (Carey, 2005). When merging with Northwest, Delta estimated a combined annual revenue of over 2 billion (Delta, 2008), more than the sum of the two firms. Examining revenue performance, therefore, considers this practical motivation of airlin es. Second, airline from regulators (Merkert and Morrell, 2012; Ryerson and Kim, 2014; Hüschelrath and Müllera 2014; Manuela et al., 2016). Against this motivation though, Schosser and Wittmer (2015) found that North American airlines failed to benefit from significant revenue synergies. Investigating revenue performan ce, thus, may reconcile this inconsistency. Given our research question, we turn our focus to past work on the impact of merger on consumer welfare (i.e. on - time performance) and carrier welfare (i.e., financial performance). The impact of merger upon on - time performance has been investigated only recently, and with mixed results. Steven et al. (2016) studied three U.S. domestic mergers, finding that mergers initially worsened on - time arrivals, and that this deterioration commonly persisted for three years . Prince and Simon (2017) examined five U.S. domestic airline mergers, using Mayer and post - merger actual travel times (i.e., the total time from scheduled departure t ime to the actual 86 arrived time). Contrary to Steven et al. (2016), Prince and Simon (2017) found that in the short run (i.e., 1 - 2 years post - merger), mergers do not impact their measure of OTP, while in the long run (i.e., 3 - 5 years post - merger) their OTP measure improved. Rupp and Tan (2019) investigated - time arrivals, and percentage of cancelled flights, finding that OTP in all thes e areas improved immediately following mergers. With regard to financial performance, researchers have investigated the impact of merger on shareholder value (Singal, 1996b; Flouris and Swidler, 2004; Gong and Firth, 2006), revenue synergy (Schosser and W ittmer, 2015), and profitability (Jordan, 1988). Singal (1996b) examined 14 of the 27 mergers between 1985 and 1988, concluding that merger announcements increased stock prices. Gong and Firth (2006) studied 15 mergers between 1985 and 2001 and found that both acquirer and target airlines experienced positive stock market responses in the wake of merger announcements. Considering impacts to revenue synergy, Schosser and Wittmer (2015) analyzed six large international airline mergers, including two in the U. S., and concluded With regard to merger impact on profitability, Jordan (1988) examined 24 mergers between 1985 and 1987 and found that the profits of the merged a irlines declined in the years following mergers. When reviewing the airline merger literature, we observe that most studies are grounded in econometric analysis with little discussion of their theoretical foundations. In the next section, we outline the t heoretical foundation we use to develop our hypotheses in an attempt to reconcile past mixed finding by increasing scientific understanding (Hunt, 1983), by answering the 87 questions of how, when and why (Bacharach, 1989), and by explaining how and why spec ific relationships lead to specific events (Wacker, 1998). 3.3 THEORY AND HYPOTHESES 3.3.1 Theoretical Foundation In the current study, our definition of merger is consistent with current practice in the airline literature (Gong and Firth, 2006; Gudmundsson et al., 2017) in that merger refers to both merger and acquisition. Merger in the airline industry has influential and long - lasting impact on merged firms, involving IT system reconfiguration, human resource integration, and operational procedure redesign, which can take years to complete (Mouawad, 2012). Consequently, - merger performances also differ. Understanding how and why merger events Organizational learning is a use ful theoretical lens to understand the nature of merger events and their consequent impact. Drawing on the seminal work of Cyert and March (1963), organizational learning has established itself as effective in explaining merger impact (Leroy and Ramanantso a, 1997; Zollo and Singh, 2004; Collins et al., 2009; Zollo, 2009), including airline s 1985, p. 803). This definition indicates that 1) organizational learning is a process; 2) organizational learning involves knowledge management. We explain how these two perspectives relate to mergers in the airline industry. At the process level, Argote and Miron - Spektor (2011) classified organizational learning into a three - step recursive process involving knowledge search, knowledge creation and/or knowledge 88 transfer, and knowledge retention. In the airline industry, carrier - merger integration can be illustrated within Argote and Miron - - loop learning cycle starting from knowledge search, transitioning to knowledge creation and/or knowledge transfer, and ending with knowledge retention. Because carriers are motivated to increase their market power over time (Hüschelrath and Müllera, 2014), merger is a quick means to achieve this goa l (Schosser and Wittmer, 2015; Chen and Gayle, 2019). To identify potential opportunities, acquirers continually evaluate competitor carriers, assessing expected synergies that a merger might create. Within Argote and Miron - learning framework, this is the pre - merger knowledge search stage. In cases when a merger results, upon the official merger closure, acquirers and target carriers begin the integration process, where knowledge creation and knowledge transfer between the tw o organizations play an important role in determining if the merger will be successful or not (Azan and Sutter, 2010). Lastly, after the integration process is completed, knowledge retention becomes essential and serves as a key ained growth and success (Marsh and Stock, 2006). In cases where acquirers experience multiple mergers, this cycle repeats itself with each new merger. At the knowledge management level, various scholars have related the concept of knowledge to organizati onal learning (Nonaka, 1994; Wang and Ahmed, 2003). Specifically, Nonaka (1994) distinguished knowledge between explicit knowledge and tacit knowledge. Based on Polanyi (1966), Nonaka (1994) proposed that explicit knowledge can be transferred in a formal a nd systematic language while tacit knowledge, having a personal quality, is hard to formalize or 89 explain the impact of merger on operational performance and financial p erformance in the airline industry. Achieving synergies in operational performance requires the learning process to involve more explicit than tacit knowledge, particularly in the airline industry where operational procedures are meticulously documented, including ground operations (Anderson et al., 2000; Wenner and Drury, 2000), flight - deck operations (Degani and Wiener,1998), cockpit operations (Degani and Wiener,1997; Loukopoulos et al., 2003), gate assignments (Bolat, 2000), and operations control cent ers (Clarke, 1998). With nearly all operational procedures documented, the two carriers involved in a merger begin their post - merger learning process by carefully reviewing existing documentation and by making necessary adjustments to develop a shared set of new operational procedures for the newly merged carrier. To achieve synergies in financial performance, more tacit knowledge learning is anticipated as necessary since the recipe to achieve better financial performance is normally not expressly documen ted in firms. Sound financial performance is usually the result of collective wisdom developed through individual experiences, where tacit knowledge resides (Huselid, 1995; Katzenbach and Smith, 2015). Therefore, the newly merged carrier will have to effic iently extract tacit knowledge from individual experiences across the merged firms to achieve financial synergies (Crossan et al., 1999). 3.3.2 The Impact of Merger on Operational Performance Consistent with past organizational behavior works (Kim and Ployhart, 2014; Hale et al., 2016), we consider the impact of merger at two stages: the transition stage and the recovery stage. The outines and expertise from the 90 pre - reconfigure or adapt collective p. 908). As previously described, an important merger outcome is that of consumer welfare, specifically on - time performance. Once a merger is approved by the Department of Justice, merging carriers start their operations integration, entering the transition stage where knowledge transfer between the two carriers begins. Both general organizational learning processes, as well as specific explicit knowledge transfer activities, occur during th is transition stage. From the general organizational learning process perspective, a merger can be considered a rare event, as they do not occur frequently to the same carrier. In analyzing the impact of rare events on organizational learning outcome, Lam pel et al. (2009) concluded that rare events can have a short - term negative impact on company performance. Zollo (2009) further explained that this is because rare events, when viewed as a potential threat by certain people, may engender excessive caution level (Zollo, 2009). Accordingly, we expect that a merger may have a short - term negative impact on operational outcomes, such as OTP, during the process of post - merger op erations integration. From an explicit knowledge transfer perspective, post - merger operations integration relies heavily upon existing documentation to transfer explicit knowledge. These well - documented operational procedures should facilitate explicit kn owledge transfer activities. However, we argue that the volume of these well - documented operational procedures common to the airline industry could become barriers to operations integration for two reasons. First, operations nds of procedures used by pilots and flight dispatchers, gate 91 synchronize. For example, following the United - Continental integration, flight attendants from the two c arriers still worked separately for a period of time after the merger (Josephs, 2018) as (McCartney, 2015). Similarly, we expect that the transition from two sets of complex operational procedures to a single unified set of procedures may result in operational issues, negatively affecting OTP. Second, the integration of the two different sets of operational procedures is likely to create conflicts between the two m erging organizations. One of the greatest challenges faced by United Airlines after it acquired Continental was to resolve conflicting goals between the different corporate cultures (Mouawad, 2012). The quality of learning amid these types of change is clo sely tied to the degree of conflict of goals, as goal conflict results in reduced learning outcomes (Miller, 1996). In the context of explicit knowledge transfer during the integration of post - merger operations, reduced learning outcomes are expected to le ad to OTP deterioration. Combining the above arguments from both general learning perspective and specific explicit knowledge transfer perspective, we hypothesize: H1a: OTP will deteriorate immediately after mergers. As time progresses, the newly formed carrier moves to the post - merger recovery stage. In an organizational learning context, the recovery stage can be viewed as a continuous learning process where knowledge transfer and knowledge retention take place. At this stage, we argue that the operati onal performance degradation experienced during the transition stage will continue to worsen before it eventually improves. We still rely on organizational learning framework to develop our theorization in the following steps. 92 From the explicit knowledge transfer perspective, as previously explained, there are thousands of operational procedures to be integrated, and the integration of these procedures could take years to complete (Mouawad, 2012). So long as the integration has not been completed, the two concurrent operational procedures may continue to result in subpar operational performance. Further, the conflict of goals occurred during the transition stage, especially conflicts in these thousands of operational procedures, are also expected to continu e exerting their negative impact on operational performance. However, when the two carriers finally complete their operations integration by resolving all conflicts in operational procedures, the operational performance of the newly merged carrier should b e expected to start improving with the new aligned operational processes now that all the conflicts are gone. The learning by doing concept explains why this transition is expected to happen. Learning by doing concept (Levitt and March 1988) predicts that change generally causes performance to deteriorate before it ultimately improves. In a merger, two carriers can be this should arise new solutions to shared problems. Operations managers from each side come to the merger immersed in their respective operational procedures, with already developed solutions for various situations. When developing shared solutions, operations managers may tend to refer to their a lready developed solutions to address post - merger challenges. However, resorting to pre - existing solutions can result in negative learning outcomes (March et al., 1991), which will manifest itself as deteriorated operational performance in our case. March et al. (1991) observed that after repeated negative learning outcomes resulting from relying upon pre - existing solutions, organizations (operation managers in our case) begin to develop new integrative knowledge to adjust and correct their actions. Once de veloped, this new form of knowledge can be retained in 93 Argote and Miron - Spektor, 2011) to handle post - merger challenges. Operational performance thereby, should c oincidentally start to improve. As a result of these, we expect that: H1b: OTP will demonstrate a U - shaped curve with time elapsing after mergers. 3.3.3 The Impact of Merger on Financial Performance Turning our attention to carrier welfare, we examine the impac t of mergers upon financial performance. During the transition stage, the effect of financial performance improvement should be immediate as the carriers start combining resources such as networks and markets. The extant merger literature often mentions th e importance of resource relatedness to post - merger performance. The concept of resource relatedness (i.e., resource or product - market similarity), posits that when acquirers merge with targets with more closely related resources, they are more likely to achieve post - merger synergies, such as in financial outcomes (Markides and Williamson, 1994; Palich et al., 2000; Miller, 2006). The rationale of synergy creation is that when firms share related resources, firms can either reduce redundant functions or pr ocesses (Teece, 1982) or improve resource deployment efficiency (Tanriverdi and Venkatraman, 2005), both cases of which lead to higher outputs. In the airline industry, carriers share high levels of related resources, including aircraft fleets, human capit al, network operations, and the like. Therefore, financial performance can be expected to improve following mergers of these resource - similar entities. Accordingly, we hypothesize: H2a: Financial performance will improve immediately after mergers. Follow ing the initial improvement in financial performance brought by related resources, we argue that financial performance will continue to improve but eventually improvements will face 94 diminishing returns. As discussed, in an organizational learning context, scholars commonly attribute sound financial performance to tacit knowledge learning in situations where no King and Zeithaml, 2001). This type of tacit knowled experience (Huselid, 1995; Katzenbach and Smith, 2015) which evolves through operating routines that incorporate experience (Lempel et al., 2009). Over time, individual experience is expected to increase gradually while managing these operating routines (Ethiraj et al, 2005). Increased individual experience enriches tacit knowledge accumulation which in turn enables organizations to improve their performance (Dutton and Thomas, 1984; Schilling et al., 2003). However, inc reased experience contributes to organizational learning outcomes only up to a certain level (Argote and Miron - Spektor, 2011). At high levels of experience, learning outcomes face diminishing returns (Argote and Miron - Spektor, 2011). Indeed, Kim et al. (20 09) observed that there exists an inverted U - shaped relationship between recovery experience and learning outcomes in an organizational learning environment. Accordingly, we hypothesize that: H2b: Financial performance improvement will face a diminishing return with time elapsing after mergers. 3.3.4 Moderating Effect of the Immediate and Long - term Impact of Merger While performance fluctuations in the transition stage and recovery stage are expected to occur for all acquirers, these effects may differ for some acquirers (King et al., 2004; Homberg et al., 2009). We develop our argument for these moderating effects in the following section. The performance difference during the transition period can be explained from the following two arguments. First, the conc ept of X - efficiency, originally proposed by Leibenstein (1966), was 95 subsequently extended to explain airline merger performance by Gudmundsson et al. (2017). The key concept of X - efficiency is that low - performing acquirers can utilize their managerial capa bilities to achieve superior synergy when engaged with mergers, especially when the two merged firms have similar or complementary resources (Gudmundsson et al., 2017). Carriers in the airline industry indeed share highly similar or complimentary resources (Gudmundsson et al., 2017) as explained in the previous section. Second, low - performing acquirers are likely to be under distress and may have resorted to merger as a means to improve performance (Schmidt, 2016). In this case, we may expect that the distr essed low - performing acquirers would probably work harder to turn around the situation. Given these two arguments, we propose that low - performing acquirers may improve their performance to a greater extent at the transition period. H3a: Low - performing acqu irers in terms of operational performance will demonstrate less deterioration in OTP during the transition period, and vice versa. H3b: Low - performing acquirers in terms of financial performance will demonstrate greater improvements in financial performanc e during the transition period, and vice versa. In the recovery stage, we also argue that the long - term trend should also be different for low - performing and high - performing acquirers. With regard to operational performance where explicit knowledge is mos tly involved, organizational learning literature suggests that long - term sound performance often is the result of more experienced managers (Mannor et al., 2016) and better organizational capabilities (Morash, 2001; Zott, 2003; Daugherty et al., 2009). Man nor et al. (2016) found that experienced managers are able to generate higher outputs by more efficiently redeploying flexible resources. In our case, more experienced managers would be able to generate more efficient operations solutions. It has long esta blished in supply chain management field (Morash, 2001; Daugherty et al., 2009) as well as in management field (Zott, 96 2003) that better organizational performance is driven by better organizational capabilities because better organizational capabilities, i n our case, can drive better explicit knowledge learning in operations management, such as through more efficient operations procedures. Combined, it is reasonable to assume that high - performing acquirers would have more experienced managers and have devel oped better organizational capabilities. Accordingly, during the post - merger recovery period, high - performing acquirers should be able to 1) mobilize their experienced managers to more effectively integrate and recombine resources to achieve long - term supe rior performance (Mannor et al. 2016); and 2) utilize better organisational capabilities to more efficiently manage operational performance over time. Accordingly: H3c: High - performing acquirers in terms of operational performance will demonstrate less pro nounced trajectories in OTP during the recovery period, and vice versa. With regard to financial performance, we argue that high - performing acquirers will find it more difficult to improve further. The underlying logic is straightforward and can be explai ned by the law of diminishing returns (Turgot, 1767). Schmenner and Swink (1998) adapted this law in operations management field and proposed that when performance is at a higher level, further improvements will become less pronounced. Conversely, when per formance is at a lower level, the law of diminishing returns predicts that further improvements should be easier to attain (Schmenner and Swink, 1998). Therefore: 97 H3d: High - performing acquirers in terms of financial performance will demonstrate less pron ounced trajectories in financial performance during the recovery period, and vice versa. 3.4 METHOD 3.4.1 Data Source and Sample To test our hypotheses, we collect data from the Department of Transportation (DOT). DOT requires U.S. carriers to report operational and financial performance on a regular basis if they have more than 1% domestic market share, measured by total scheduled domestic passenger revenues. At the time of accessing DOT database, operational performance, reported at monthly level, is available from January 1998 to September 2019; financial performance, reported at a quarterly level, is available from the first qu arter of 1990 to the second quarter of 2019. To avoid the impact of 9/11 as well as DOT report format changes in October 2003, we elect to start our analysis from the first quarter of 2004. Since our data ends in the second quarter of 2019, our sample cons ists of 62 quarters in total. After aggregating and cleaning operational and financial raw data, operational performance data yielded 26 carriers while financial performance data yielded 124 carriers. Accordingly, the 26 carriers from operational performa nce data was used as the index to combine financial performance data. Operational performance data was aggregated to a quarterly level using with incomplete rep orting over the sample timeframe were removed from the combined data. These carriers only infrequently met DOT reporting thresholds. As a result, our final carrier list consists of 20 carriers. Nine of these carriers spanned the entire 62 quarters, the re maining carriers spanned between 18 and 61 quarters. Table 1 2 summarizes the airlines in our sample. 98 Table 12 Airlines in Dataset No. Airline First quarter in data Last quarter in data Total quarters in data 1 AIRTRAN 2004 Q1 2012 Q1 33 2 ALASKA 2004 Q1 2019 Q2 62 3 AMERICAN 2004 Q1 2019 Q2 62 4 ATLANTIC SOUTHEAST 2004 Q1 2011 Q4 32 5 COMAIR 2004 Q1 2010 Q4 28 6 CONTINENTAL 2004 Q1 2011 Q4 32 7 DELTA 2004 Q1 2019 Q2 62 8 ENVOY 2004 Q1 2019 Q2 54 9 EXPRESSJET 2004 Q1 2019 Q2 62 10 FRONTIER 2005 Q2 2019 Q2 57 11 HAWAIIAN 2004 Q1 2019 Q2 62 12 JETBLUE 2004 Q1 2019 Q2 62 13 MESA 2006 Q1 2019 Q2 38 14 NORTHWEST 2004 Q1 2009 Q4 24 15 SKYWEST 2004 Q1 2019 Q2 62 16 SOUTHWEST 2004 Q1 2019 Q2 62 17 SPIRIT 2015 Q1 2019 Q2 18 18 UNITED 2004 Q1 2019 Q2 62 19 US AIRWAYS 2004 Q1 2013 Q4 40 20 VIRGIN AMERICA 2012 Q1 2017 Q4 24 To identify mergers within these 20 carriers, we review relevant airline research (Jain, 2015; Steven et al., 2016; Prince and Simon, 2017; Vaze et al., 2017) as well as industry reports, such as Aviation Daily. Subsequently, we identify eight mergers, the greatest number of mergers analyzed in recent airline research so far, to the best of our knowledge. These mergers were: Airways/America West (2005), Delta/Northwest (2009), Frontier/Midwest (2010), United/Continental (2010), ExpressJet/Atlantic Southeast (2011), Southwest/Air Tran (2011), US American/US Airways (2013), and Alaska/Virgin America (2016). 3.4.2 Measures Operational Performance We measure the impact of mergers on operational performance by on - time arrivals. According to DOT (2019), a flight is con sidered on time if it operated within 15 minutes of the scheduled time 99 Flight Delays Report. Financial Performance As discussed in the literature review, we elect to focus on revenue as the measure for financial collected from DOT Form 41 Financial Data Schedule P1.2. Merger Event code the quarters before the merger event as 0 and the quarters after the merger event (including the quarter where the merger event took place) as 1. By coding merg er event as a dummy variable like this, we are able to estimate the immediate impact upon mergers. Merger event is denoted as Transition in our models. Current airline merger research typically utilizes the officially released merger completion date to de fine pre - and post - merger time windows (Jain, 2015; Steven et al., 2016; Prince and Simon, 2017). However, after the official merger completion date, carriers sometimes continue to report to DOT as two individual carriers, before eventually reporting to DO T as one single carrier. To address this issue of overlapped reporting after the official merger completion date, we aggregate the overlapped DOT report records based on the official merge completion dates to align with current practices in airline literat ure. 100 Recovery Rate Following Bliese and Lang (2016), the recovery rate for acquirers was coded as follows. All quarters prior to the merger event were coded as 0; the quarter when the merger event occurred was also coded as 0; all quarters after the mer ger event were then coded in sequential order (i.e., 1, 2, 3, and etc.). This variable is denoted as Recovery in our models. To test the hypothesized non - linear effects in the recovery stage, the quadratic form of Recovery was also included in our models, denoted as Recovery.SQ , following Bliese and Lang (2016). Time The interpretation of Transition and Recovery crucially depend on how time is specified (Bliese and Lang, 2016). Because our main research interest is to investigate both the immediate performance fluctuations upon mergers as well as the long - term performance changes after this format of coding, the first measurement occasion (i.e., 2004Q1) was coded as 0 and the following measurement occasions were coded in sequential order (i.e., 1, 2, 3, and etc.). However, this sequential order stops one quarter before the merger event and remains constant thereafter. By coding time in this manner, the coefficient of Transition measures the immediate absolute performance fluctuations while the coefficient of Recovery measure s the absolute recovery slope. The coding of time is denoted as Time in our models. An example of coding Transition , Recovery , and Time for the Delta/Continental merger is presented in Table 1 3 . 101 Table 13 Coding Transition, Recove ry, and Time Using Delta Airline as an Example Year Quarter Measurement Occasion Time Transition Recovery Recovery.SQ 2004 1 1 0 0 0 0 2004 2 2 1 0 0 0 2004 3 3 2 0 0 0 2004 4 4 3 0 0 0 2009 1 21 20 0 0 0 2009 2 22 21 0 0 0 2009 3 23 22 0 0 0 2009 4 24 23 0 0 0 2010 1 25 23 1 0 0 2010 2 26 23 1 1 1 2010 3 27 23 1 2 4 2010 4 28 23 1 3 9 2019 1 61 23 1 36 1296 2019 2 62 23 1 37 1369 High Performing Acquirers Our third set of hypotheses compares performance differences between high - performing acquirers and low - performing acquirers during the transition and recovery stages. To identify high - performing and low - performing acquirers, we refer to existing practices in management research, setting the industry performance as the reference point (Lang et al., 1989; Servaes, OTP and revenue, were calculated for the pre - merger p eriods for each acquirer. OTP was calculated as the averaged performance during the pre - merger period. Revenue was calculated as annual compound growth rate during the pre - merger period. The detailed calculation results appear in Table 1 4 . For OTP, if an a cquirer possessed above industry averaged performance, it was assigned as high - performing; if an acquirer exhibited below industry averaged performance, it was assigned as low - performing. The same logic applies in assigning high - performing and low - performi ng acquirers for revenue performance. Coincidently, Alaska, Frontier, and Southwest are the three high - performing carriers in both OTP and revenue during their 102 respective pre - merger periods. So, a dummy variable was created to indicate high performing acqu irers with 1 assigned to Alaska, Frontier and Southwest and 0 to the remaining five acquirers. Then, this dummy variable was interacted with Transition , Recovery , and Recovery.SQ , which were denoted as Transition Interaction , Recovery Interaction , and Recovery.SQ Interaction in our models. Table 14 Define High - performing and Low - performing Acquirers Carrier Pre - merger Period OTP (Averaged) Revenue (CAGR) Start End Acquirer Industry Acquirer Industry ALASKA 2004 2015 81.42% 78.44% 10.02% 4.68% AMERICAN 2004 2013 75.73% 78.37% 1.85% 5.04% DELTA 2004 2009 76.76% 77.00% 0.23% 4.51% EXPRESSJET 2004 2011 75.53% 78.10% - 5.55% 5.46% FRONTIER 2005 2009 79.68% 77.21% 8.85% 4.51% SOUTHWEST 2004 2013 80.53% 78.37% 12.78% 5.04% UNITED 2004 2011 77.49% 77.80% 3.53% 5.46% US AIRWAYS 2004 2005 76.84% 78.25% 5.21% 18.63% Control Variables To address potential endogeneity bias, an area of challenge in airline research (Scotti and Dresner, 2015; Yazdi et al., 2017), we control for factors that might influence the relationships e. Among carrier specific factors, we control for fuel cost, number of enplaned passengers, market share, total number of employees, total departure delays, and carrier group. We briefly discuss the reasons to include these control variables. First, incre ased fuel costs have pushed airlines to keep raising their fares, which will be directly reflected in revenue (Alexander, 2006). Higher save costs, impacting OTP. Second, the number of enplaned passengers directly impacts OTP and revenue. Greater numbers of passengers impose greater operational challenges, potentially leading to worsened OTP, although at the benefit of increased revenue. Third, market share als o 103 has a direct influence on OTP and revenue. Greater market share translates to higher revenues (Bolton, 2004) but also indicates more complicated networks to manage, potentially hindering carriers from achieving better OTP. Fourth, airline industry is a l abor - intensive industry where heavily depends on the number of em ployees, especially given that ground staff account for 85% OTP and as a result, incurs significant direct and indirect costs to carriers (Cook et al., 2012; P eterson et al., 2013), which would in turn negatively impact revenue performance. Finally, DOT (2019) classifies carriers into two groups, Low Cost Carriers (LCCs) and legacy carriers. LCCs are characterized by limited point to point operations, covering o nly specific geographic areas while legacy carriers, cover wider geographic areas through hub - and - spoke networks (Mellat - Parast et al., 2015). Luo (2014) observed that LCCs exert greater impacts on post - merger fare increases compared to legacy carriers, su ggesting that LCCs would benefit more from revenue enhancement during mergers. Regarding macro - economic factors that could affect the relationships of interest, we control for recession, GDP change, and fare change. Changes in revenue and OTP may also be driven by these unobservable macro - economic factors (Gayle and Wu, 2013; Luo, 2014). As such, we collect data from different sources to further address these endogeneity bias concerns. Recession is collected from Federal Reserve Economic Data in the form o f Smoothed U.S. Recession Probabilities (Piger and Chauvet, 2019). GDP was also collected from Federal Reserve Economic Data in the form of quarter over quarter percentage change (Bureau of Economic 104 Analysis, 2019). For quarter over quarter fare percentage changes, we compile data from DOT and manually calculate the changes across our study timeframe of 2004 to 2019. The definitions of variables and data sources are summarized in Table 1 5 . In addition, following modeling practices of Bell and Jones (2015) and Hoffman (2015) as well as to reflect our main research interest, which is to investigate the longitudinal impact of mergers within individual carriers, we construct the within effects for relevant carrier - specific variables in our models to better capt ure the impact of mergers on individual carriers. Table 15 Variables Used in Analysis Variable Formula or Definition Data Source On - Time Performance Overall percentage of fights arriving within15 minutes of scheduled arrival time. Aggregated to quarterly level. DOT Air Travel Monthly Flight Delays Report Operating Revenue Airline's operating revenue as reported each quarter by DOT. DOT Schedule P1.2 in Form 41 Financial Data Fuel Cost Total scheduled domestic fuel cost (Dollars) each month. Aggregated to quarterly level. DOT Schedule P12A in Form 41 Financial Data Enplaned Passengers Number of enplaned passengers each carrier each month. Aggregated to quarterly level. DOT Schedule T1 in Form 41 Air Carrier Summary Data Market Share The ratio of a carrier's quarterly revenue passenger miles to the sum of revenue passenger miles of the total 20 carriers in that quarter. DOT Schedule T1 in Form 41 Air Carrier Summary Data Number of Employees Monthly Number of Full - Time Equivalent Employees. DOT Schedule P1(a) in Form 41 Financial Data Total Departure Delay The monthly sum of delays caused by cancelled flights, diverted flights, aircraft delay, extreme weather delay, national aviation system delay, security delay, and late arriving aircraft delay. DOT Air Travel Monthly Flight Delays Report LCC (Low Cost Carriers) Six low cost carriers defined by DOT (Allegiant Air, Frontier, JetBlue, Southwest, Spirit, Virgin America). DOT Website Definition Recession Smoothed U.S. Recession Probabilities. Federal Reserve Economic Data GDP % Change Quarter over quarter change in GDP. Federal Reserve Economic Data Fare % Change Quarter over quarter change in average domestic fares. DOT DB1B * Monthly data was aggregated into quarterly data to match the quarterly financial measures. 105 3.4.3 Descriptive Statistics Table 1 6 presents the observed means, standard deviations, and correlations for all variables in interest. Regarding OTP, we see that both Transition and Recovery are positively correlated with OTP, suggesting that OTP improved following mergers. Regarding revenue, we see that both Transition and Recovery are also positively correlated with revenue, implying that revenue increased post - merge r. Regarding carrier - specific control variables, most of the correlations are as expected, such as number of employees being positively correlated with both OTP and revenue. For macro - economic factors, most of the correlations are also as expected, such as recession being negatively correlated with revenue. These correlation results provide preliminary evidence for our hypothesized relationships . T able 16 Summary Statistics and Correlation Matrix 3.4.4 Analytical Method Since our hypotheses involve the investigation of immediate performance fluctuations upon mergers as well as long - term performance changes following mergers, we resort to discontinuous growth modeling to specify our models (Singer and Willet, 2003; Lang an d Bliese, 2009; Hoffman, 2015; Bliese and Lang, 2016). Also known as piecewise hierarchical linear modeling 106 (Raudenbush and Bryk, 2002; Hoffman, 2015), discontinuous growth modeling allows us to examine the immediate and long - term impact of mergers while c apturing the nested observations in our dataset, i.e., time (level 1) is nested within carriers (level 2). We follow the model specification and model testing procedures proposed by Bliese and Lang (2016) to conduct analyses in Stata SE15.1 using xtmixed c ommand with maximum likelihood estimator. R 2 calculations follow Nakagawa and Schielzeth (2013) as in Equation 5 . Our full model specification is expressed in Equation 6 . Equation 5 MVP R 2 Calculation Equation 6 Full Model Specification In Equation 6 , is one of our two outcome variables for carrier i in quarter t . is a vector of control variables discussed in the previous section. represents a vector of carrier fixed effects, which is used to control any time - invariant carrier - specific factors that may affect the two outcome variables. is the idiosyncratic error term. The interpretation of other coefficients are as follows. T he intercept captures the value of dependent variables at time 0 which is the first quarter of 2004. represents the slope before the merger event, i.e., the pre - merger slope. is our coefficient in interest which reflects the absolute changes in t he value of dependent variables relative to 0, upon merger. A statistically positive (negative) means that OTP and revenue have improved (deteriorated) immediately upon merger. , the post - merger 107 slope estimate also represents the absolute change in sl ope relative to 0. A statistically positive (negative) means that OTP and revenue show an upward (downward) trend following merger. tests the non - linear growth rate for OTP and revenue following merger. A statistically positive (negative) means t hat OTP and revenue show an accelerating or decelerating trend following merger, depending on the signs of . In discontinuous growth modeling , the inclusion of both linear and quadratic growth terms for the recovery period can also help to 1) control for post - merger seasonality, addressing endogeneity concerns; 2) prevent random variations between post - merger quarters from showing up as noises in the model; and 3) use only two parameters instead of many post - merger quarter dummies to model the time effect , resulting in more parsimonious models (Hale et al., 2016). 3.5 RESULTS Table 17 presents h ypotheses testing results from a mixed effect model with autocorrelated residuals. Model 1 tests hypotheses 1 (i.e., the impact on OTP) while model 3 tests hypotheses 2 (i.e., the impact on revenue). Moderating effects in hypotheses 3 were tested respectively in Model 2 and Model 4 by the associated interaction terms. H1a proposed that OTP would deteriorate immediately after mergers. This hypothesis was tested by the c oefficient of Transition in Model 1. The coefficient of Transition is significant but had the opposite sign anticipated (0.02, p = 0.000 ). Therefore, H1a was not supported. H1b expected that OTP would demonstrate a U - shaped curve over time after merger. This hypothesis was tested jointly in Model 2 by the coefficients of Recovery ( - 0.0015, p = 0.094 ) and Recovery.SQ 108 Table 17 Hypotheses Testing Results Parameter Model 1 OTP Model 2 OTP Interaction Model 3 Revenue Model 4 Revenue Interaction Fixed Effect Intercept 0.79** (73.43) 0.78** (96.20) 13.53** (110.2) 12.65** (7.92) Time - 0.001 ( - 1.35) - 0.0018* ( - 2.94) 0.014 (1.37) 0.014 (1.34) Time.SQ 0.00001 (0.57) 0.00003 * (2.12) - 0.00003 ( - 0.11) - 0.00002 ( - 0.07) Transition 0.02** (3.53) 0.033 ** (4.93) 0.08* (1.94) 0.16** (3.22) Recovery - 0.0014 ( - 1.63) - 0.0015 ( - 1.68) 0.01 (0.97) 0.002 (0.18) Recovery.SQ 0.00002 (0.74) 0.00001 (0.64) - 0.00016 ( - 0.53) - 0.00001 ( - 0.03) Fuel Cost 0.002 (1.09) 0.0018 (1.16) - 1.07 ( - 0.25) 3.39 (0.55) Enplaned Passengers 0.10** (15.66) 0.11** (16.95) 0.36 (0.40) - 3.70 (0.52) Market Share - 0.06 ( - 0.69) - 0.15 ( - 1.71) - 5.14 ( - 0.79) - 7.62 ( - 0.30) Total Employees 0.03** (3.30) 0.32** (3.57) 1.27 (1.45) - 1.29 ( - 0.27) Total Departure Delay - 0.18** ( - 68.50) - 0.18** ( - 68.51) - 0.61 ( - 1.11) 4.45 (0.49) LCC - 0.014 ( - 0.69) 0.016 (1.20) 0.29 (0.80) 2.31 (0.65) Recession - 0.006* ( - 2.15) - 0.005 ( - 1.94) 0.009 (0.45) 0.01 (0.47) GDP % change - 0.00006 ( - 0.06) - 0.00008 ( - 0.09) 0.016* (2.41) 0.016* (2.45) Fare % change 0.008 (1.64) 0.008 (1.67) - 0.22** ( - 6.92) - 0.22** ( - 6.89) Better Performer Dummy 0.032** (2.70) - 5.23 ( - 0.57) Transition Interaction - 0.038** ( - 3.56) - 0.24** ( - 2.81) Recovery Interaction 0.00004 (0.02) 0.026 (1.01) Recovery.SQ Interaction 0.00001 (0.20) - 0.0005 ( - 0.54) Carrier Fixed Effects: Yes Random Effect Level 2: Carriers Variance (Intercept) .001558 .000000 .000000 .000000 Level 1: Residual AR1 .764501 .693162 .943472 .933287 Variance .000578 .000443 .133061 .111179 Measures of Fit - 2 Log Likelihood - 4875.59 - 4892.58 - 1247.79 - 1258.00 AIC - 4839.59 - 4848.58 - 1213.79 - 1216.00 BIC - 4753.12 - 4742.90 - 1131.51 - 1114.35 R 2 (MVP) 86.87% 87.70% 90.88% 91.21% Total R 2 45.33% 89.76% 92.24% 93.47% - tailed). Z - tests are reported in parentheses for the fixed effects parameters. 109 (0.00001, p = 0.524 ). Both coefficients are non - significant although with the expected sign. Therefore, H1b was not supported. H2a suggests that financial performance would increase immediately after mergers, which was tested by the coefficient of Transition in Model 3. The significant coefficient ( - 0.08, p = 0.048 ) supported. H2b posits that financial performance improvement will experience diminishing returns over time after mergers. H2b was tested in Model 4 jointly by the coefficients of Recovery (0.002, p = 0.856 ) and Recovery.SQ ( - 0.0001, p = 0.976 ). Both coefficients are non - significant, thus, H2b was not supported. The third set of hypotheses investigate moderating effect of ac - merger performance during initial transition and the following recovery periods. H3a predicts that low - performing acquirers will experience less deterioration in OTP during the transition period than high - performing acquirers. H3a was tested i n Model 2 by the coefficient of Transition Interaction . The associated coefficient is significant ( - 0.038 , p= 0.000 ) with the expected sign . Therefore, we find support for H3a. H3b predicts that low - performing acquirers will demonstrate greater improvement s in financial performance during the transition period. This hypothesis was tested by the coefficient of Transition Interaction in Model 4. The coefficient of Transition Interaction ( - 0.24, p= 0.005 ) was statistically significant with the expected sign. T herefore, H3b was also supported. To better illustrate the interaction effect, we plot the interaction following Dawson (2014) in Figure 6 and Figure 7 . From the plot, we see that with regard to OTP, the pre - and post - merger performance for high - performing acquirers are almost the same albeit with a slight decrease post - merger. But for low - performing acquirers, there was a significant increase. Financial performance wise, we see that low - performing acquirers were able to significantly 110 increase their perform ance during the transition period. However, high - performing acquirers actually suffered from a loss during this period. Figure 6 Moderating Effect on OTP at Transition Figure 7 Moderating Effect on Financial Performance at Transition 111 H3c proposed that high - performing acquirers would demonstrate less pronounced trajectories in OTP during the recovery period. This hypothesis was tested jointly by the coefficients of Recovery Interaction (0.00004, p = 0.949 ) and Recovery.SQ Interaction (0 .00001, p = 0.886 ) in Model 2. Neither of the coefficients were significant, failing to support H3c. H3d predicted that high - performing acquirers would demonstrate less pronounced trajectories in financial performance during the recovery period. The coeffi cients of Recovery Interaction (0.026, p = 0.335 ) and Recovery.SQ Interaction ( - 0.0005, p = 0.563 ) in Model 4 test this hypothesis. Both coefficients are non - significant, thus, H3d was not support. 3.6 DISCUSSION 3.6.1 Theoretical Contributions This study provides several broad theoretical contributions to the airline merger literature. First, the discontinuous growth modeling offers a framework to study the impact of merger over time, answers the call of King et al. (2004) to investigate the imp act of merger over longer event windows. Our specification of time differs from recent airline merger research where varying definitions of short - term and long - term post - merger event windows were used (Hüschelrath and Müllera, 2014; Prince and Simon, 2017; Yan et al., 2019). By defining time as the continuous predictor, we are able to examine both the immediate and long - term impact of merger. More importantly, with eight merges included, our results show that the effect of merger in the U.S. airline industr y is only short - term. More specifically, with regard to operational performance, in contract to the current literature (Steven et al., 2016; Prince and Simon, 2017) that shows long term positive trend, we only find an immediate OTP improvement. Regarding r evenue performance, we also find an immediate revenue increase. By defining time in its sequential 112 order, we demonstrate that using the variable Time is fundamental to capture the nature of the impact of merger. Second, the two - stage model design used her ein provides the ability to analyze the impact of merger at more nuanced levels, laying a foundation for future related research examining similar events. While using a Time variable in the recovery stage underscores the needs to model the longitudinality of the impact of merger events, the significance uncovered by the two - stage design lends credence to the utility of examining both immediate and long - term impacts present in more complicated relationships. In particular, the two - stage design enables us to make more precise theoretical predications. In addition, our two - stage design and the associated hypotheses testing also reveal that the impact of merger on both operational performance and financial performance are only immediate rather than long - term. Wi thout the two - stage design, we would not be able to capture this nuance. Third, our results also show that the pre - merger performance level impacts the post - merger performance at the transition stage. Airline merger research to date, whether focusing on f are changes (Kwoka and Shumilkina, 2010; Luo, 2014; Hüschelrath and Müllera, 2014; Carlton et al., 2019), revenue synergies (Schosser and Wittmer, 2015), or service quality impact (Steven et al., 2016; Prince and Simon, 2017), has largely ignored the poten - merger performance on post - merger outcomes. With regards to both operational performance and financial performance, our results show that at the transition stage, low - performing acquirers, rather than high - performing acquirers, benefit from mergers by immediately improving their OTP and revenue following mergers. These findings demonstrate important implications for scholars to conceptualize the relationships between other carrier idiosyncratic factors and merger outcomes. 113 3.6.2 Mana gerial Implications This study also provides important implications for managers and policy makers. First, our results show that OTP improves immediately following mergers. This corroborates with e industry helps to improve service quality, such as on - time arrivals. However, based on our findings, the improvement of on - time arrivals is only short - term following merges. To this end, both policy makers and airline decision makers should strive to fin d solutions to achieve a sustained long - term improvement in on - time arrivals following mergers. Second, despite that most airline acquirers justify mergers by expectations of higher revenue synergies (Carey, 2005; Delta, 2008), our results indicate that me rgers as a means to achieve revenue growth may not be a valid strategy because the effect is only short - term based on our findings. Managers, in this case, will need to find other means to generate additional and sustained revenue streams after the effects of merger fade. Third, our results show that post - merger performance level is influenced by the pre - merger performance level. More specifically, only low - performing acquirers benefit from mergers during the transition period. Managers, therefore, should b e conscious of this when designing corresponding coping schemes in line with their respective pre - merger performance levels to re - allocate post - merger resources in the post - merger integration stage in order to overcome the post - merger challenges. 3.6.3 Limitati on and Future Research Like all research, ours has limitations as well as associated future research avenues. First, we rely on organizational learning framework to theorize our hypothesized relationships. This framework involves three recursive stages fro m knowledge search, to knowledge creation/transfer, and to knowledge retention (Argote and Miron - Spektor, 2011). Due to data 114 limitations, however, we could not directly test the knowledge transfer and knowledge retention activities during the transition an d recovery stages. Scholars have developed variable constructs and models to investigate knowledge transfer between organizations (Mowery et al., 1996; Mesquita et al., 2008; Reus et al., 2016). Future research might build on the knowledge transfer litera ture to investigate knowledge transfer and retention activities during mergers through other research methods and examine more precise impacts of knowledge transfer on merger outcomes. Second, to test our hypotheses, we resort to discontinuous growth mode ling (Bliese and Lang, 2016). There are other modeling choices to address these research questions, such as econometric regression discontinuity design (Hahn et al., 2001; Imbens and Lemieux, 2008), difference - in - difference (Wooldridge, 2010), and event st udy methodology (Boehmer et al., 1991). These modeling approaches make different model assumptions and utilize different estimators. Future researchers are encouraged to try different methodological approaches to examine the nuanced differences achieved by using different modeling approaches. Third, to test the hypothesized moderating effect of pre - merger performance, we construct high - performing and low - performing acquirers by comparing the averaged OTP and compound annual growth rate of the acquirers wit h that of the whole industry during the pre - merger period. Such a comparison is a cross - sectional comparison which loses its ability to further examine how longitudinal changes in OTP and financial performance impact merger outcomes. We encourage future re searchers to leverage the nature of longitudinal data to investigate this relationship. Fourth, for the interaction effect, we find support that pre - transition performance. However, the statistically significant findi ngs on this moderating effect does not necessarily mean that pre - merger performance is the main, or the only, factor that might moderate merger outcomes. There are other acquirer - specific characteristics worth investigating 115 that could have significant infl uences on merger outcomes, such as method of payment, firm characteristics, and environmental factors (King et al., 2004; Haleblian et al., 2009). Future research could explore these acquirer idiosyncratic characteristics to investigate other potential mod erating relationships. Finally, this study is a single industry study in the U.S. airline industry. Single industry study has its advantages in that it provides researchers with deeper understanding of the industry and accordingly allows researchers to di rectly compare performance differences between firms where the determinants of superior performance can be precisely identified (Garvin 1988). However, a single industry study also has its limitations in that its ability to test contextual factors may be h indered (Hale et al., 2016). Thus, we also call for future study to examine other similar research questions utilizing data from various industries where the impact of different contextual factors on merger outcomes can also be modeled. 3.7 CONCLUSION In this study, we simultaneously investigate the impact of merger on consumer welfare (i.e., operational performance) and carrier welfare (i.e., financial performance). With regards to operational performance, one of the two key reasons ofte n cited by the Department of Justice (2019) to reject any merger is the potential service deterioration post - merger. Our results suggest this concern is not unfounded, as we observe that service quality, as measured by on - time arrivals, only improved immed iately following mergers. The long - term effect of merges on on - time arrival improvement is non - significant. With regard to financial performance, scholars (Ryerson and Kim, 2014; Schosser and Wittmer, 2015) as well as airlines (Carey, 2005; Delta, 2008) to that while revenue increased immediately following mergers, there is no long - term effect as well. 116 Our study, therefore, provides important practical guidance for policy and decision makers regarding the impact of merger on operational and financial performance, in addition to the various theoretical guidance discussed in the previous section s that contribute to knowledge accumulation . 117 APPENDICES 118 APPENDIX A Comparison of C urrent R esearch with S elective L iterature Table 18 Comparison of Current Research with Selective Literature Load Factor as Main Predictor Between Carrier Effect Within Carrier Effect Curvilinear Relationship Level of Analysis Atkinson et al. (2013) No No Yes No Carrier Behn and Riley (1999) Yes No Yes No Carrier Collins et al. (2011) No No Yes No Carrier Ramdas and Williams (2008) No No Yes No Flight Sim et al. (2010) Yes No Yes No Carrier Current Research Yes Yes Yes Yes Carrier 119 APPENDIX B Summary Statistics and Correlation Matrix Table 19 Summary Statistics and Correlation Matrix 120 APPENDIX C Hypothesized R elationships for H 1, H 2, and H 3 FIGURE 8 Hypothesized R elationships for H 1, H 2, and H 3 121 APPENDIX D Summary Statistics and Correlation Matrix Table 20 Summary of Statistics and Correlation Matrix 122 APPENDIX E Summary of U.S. A irline M erger and A cquisition R esearch Table 2 1 Summary of U.S. Airline Merger Literature Authors Impact of Mergers Mergers Examined Time Period Before and After Mergers Level of Analysis Method Main Findings Bilotkach 2011 Flight frequency America West - US in 2005 Two years before and two years after Carrier - airport GLS and 2SLS Flight frequency decreased at a diminishing rate following mergers. Bilotkach et al. 2013 Flight frequency Delta - Northwest in 2009 Two years before and two years after Route Difference - in - difference Flight frequency increased in some hubs but decreased in others. Borenstein 1990 Fare Northwest Republic and Trans World Ozark in 1986 One year before and one year after Route ANOVA Price increases for the Northwest Republic merger of about 10% in total. Largely insignificant results for the Trans World Ozark merger. Carlton et al. 1980 Consumer welfare North Central - Southern in 1976 1976 the merger year City - carrier Logit regression Consumer welfare gains following mergers in terms of shorter travelling time. Carlton et al. 2019 Fare Delta - Northwest in 2009; American - US in 2013; United - Continental in 2010 Two years before and two years after Route Difference - in - difference Mergers did not increase fares. Flouris and Swidler 2004 Stock market response American Trans World in 2001 10 occasions before and after the announcement date Carrier Event study Equity value declined more than 30% following the merger. Gong and Firth 2006 Stock market response 15 mergers between 1985 and 2001 5 days centered on the announcement event. Carrier Event study Marginally positive abnormal return for bidders and hi ghly positive abnormal return for target around the 1st public announcement of the merger. Hüschelrath and Müllera 2014 Fare America West - US in 2005 Two years before and two years after Route - airport and Route - carrier Difference - in - difference Average prices increased substantially following mergers. Hüschelrath and Müllera 2015 Fare Delta - Northwest in 2009 15 years before and two years after Route Panel Data Fixed effects Short - term price increases of about 11% on overlapping routes and about 10% on non - overlapping routes. 123 Jain 2015 Fare 7 recent mergers One year before and one year after Route Panel Data Fixed effects The merger wave has increased overall prices by 2.3 - 5.9% Jordan 1988 Operating expenses and profit; flight frequency 24 mergers between 1985 and 1987 Eight years before and one to two years after Carrier Descriptive statistics Operating expenses of the merged carriers increased, which had an adverse effect on profits. Flight frequency of the merged firms either declined or grew more slowly vs non - merging firms. Kim and Singal 1993 Fare 14 mergers between 1985 and 1988 One quarter before and one quarter Route Difference - in - difference Relative fares on the merging 9.4%. Knapp 1990 Market power 9 mergers in 1986 20 days prior and 10 days after announcement Carrier Event study Abnormal return movement of rival and merging firms predicts increased firm control over fares, supporting the market power hypothesis about mergers. Kwoka and Shumikina 2010 Fare US Air - Piedmont in 1987 One year before and one year after Route Difference - in - difference Prices rose by 5.0 to 6.0 per cent on routes that one carrier served and the other was a potential entrant. Price rose by 9 1 to 10.2 percent where the two carriers had been direct competitors. L ichtenberg and Kim 1989 Unit cost; Price 5 mergers between 1970 and 1984 2.5 years before and after Carrier Difference - in - difference Mergers were associated with reductions in unit cost and increase in fares. Luo 2014 Fare Delta - Northwest in 2009 2008Q1 as before and 2010Q2 as after Airport - route OLS The merger only generated a small fare increase. Morrison 1996 Fare and competition Northwest Republic and Trans World Ozark in 1986; US Air - Piedmont in 1987 Eight to nine years before and after Route OLS 2.5% increase for the Northwest - Republic merge r and 15.3% decreases for the Trans World - Ozark merger. The US - Piedmont merger had long - run fare increases averaging nearly 23%. Morrison and Winston 1989 Consumer welfare 6 mergers between1986 - 1987 1983 as the pre - merger period; fitted values for post - merger Route Logit regression Half of the mergers increases consumer welfare . Half of the mergers reduced consumer welfare (i.e., fare increase). 124 Table 21 d ) Moss and Mitchell 2012 Fare Delta - Northwest in 2009; United - Continental in 2010 One year before and 2011 as after Route Descriptive statistics Fare increases more than 10% over the pre - to post - merger period. Peters 2006 Fare 5 mergers between 1986 to 1987 One year before and one year after Route Simulation and 2SLS Fare increased following mergers. Prince and Simon 2017 On time performance 5 recent mergers Three years before and up to five years after Carrier - route Panel Data Fixed effects In the short run, very limited evidence of worsened OTP; in the long run, travel time does not worsen, and even appears to improve relative to pre - merger levels. Rupp and Tan 2019 On time performance 4 recent mergers Four quarters before and four quarter after Carrier - route Difference - in - difference Shorter travel times following mergers. Ryerson and Kim 2014 Fuel consumption Delta - Northwest in 2009; United - Continental in 2010 Feb 2004 and Feb 2012 Aircraft Hierarchical Cluster Analysis Fuel savings achieved by both merged airline networks, ranging from 25% to 28%. Schosser and Wittmer 2015 Cost synergy and revenue synergy 6 large international mergers between 2003 and 2012, including two U.S. mergers (Delta - Northwest 2009; United - Continental 2010) First five years post mergers. Carrier Case study North American airlines expect more revenue synergies than cost synergies from airline mergers. Shaw and Ivy 1994 Network structure 15 simulated mergers 1990Q4 City - carrier level Network analysis Three network patterns (single carrier dominant, overlapping, and complementary) are identified in the study. Shen 2017 Market competition and fare United - Continental in 2010 4 years before and 3 years after Route Difference - in - difference Price increased by 7.8% following the merger. Singal 1996a Interaction between multimarket contact and fare 14 mergers between 1985 and 1988. One quarter before and one quarter after Route Difference - in - difference Airfares rise in proportion to rise in multimarket contact. Changes in concentration also contributes to rise in fares. 125 Table 21 (c ont Singal 1996b Stock market response and fare 14 mergers between 1985 and 1988 Four different event periods are used for estimation: - 1 to 0, - 1 to + 1, - 3 to + 1, and - 5 to + 1, relative to the merger announcement date. Route Event study Enhancement of market power by airline mergers is supported both by stock prices and produ ct prices. Steven et al. 2016 On time performance; Lost bags; Involuntary denied boarding Northwest - Delta in 2009; United - Continental in 2010; Southwest - Air Tran in 2011 2004 - 2013: 5 - 7 years before and 2 - 4 years after Route Difference - in - difference 1. Service deterioration in the immediate years following the mergers. However, these service deterioration fades away for both flight cancellations and mishandled bags after the sixth quarter following the merger. 2. Deterioration s in both OTP and involuntary boarding denials persist well into the third year after mergers. Vaze et al. 2017 Fare; flight frequency 5 recent mergers One year before and one year after for three mergers. One year before and 2 quarters after for the other two mergers. Carrier - route Difference - in - difference Consumer welfare gains in regions dominated by the larger carrier in the merger, and welfare losses in highly concentrated markets following legacy mergers. Werden et al. 1991 Fare Northwest Republic and Trans World Ozark in 1986 One year before and one year after Route - city OLS Substantial increases in market power (i.e., fare increases) following mergers. Current study On time performance and revenue synergy 8 recent mergers 5 years before and up to 9 years after Carrier Discontinuous Growth Modeling The impact of mergers is only short term . O n - time arrivals and financial performance improved immediately following mergers. No long - term impact was found. In addition, only low - performing acquirers benefit from mergers in both operational performance and financial performance. 126 REFERENCES 127 REFERENCES perspective on operations management research. Production and Operations Ma nagement, 16(6), pp.665 - 688. Alan, Y. and Lapré, M.A., 2018. Investigating Operational Predictors of Future Financial Distress in the US Airline Industry. Production and Operations Management, 27(4), pp.734 - 755. Alexander, K. 2006. Southwest Pushes Fares Higher, Blaming Fuel Costs. Retrieved from http://www.washingtonpost.com/wp - dyn/content/article/2006/07/05/AR2006070501721.html (Accessed on Jan 17, 2019). Altay, N. and Green III, W.G., 2006. OR/MS research in disaster operations management. Europe an Journal of Operational Research, 175(1), pp.475 - 493. Amihud, Y., & Lev, B., 1981. Risk Reduction as a Managerial Motive for Conglomerate Mergers. The Bell Journal of Economics, 12(2), pp.605 - 617 . Anderson, E.W., Fornell, C. and Lehmann, D.R., 1994. Cust omer satisfaction, market share, and profitability: Findings from Sweden. Journal of Marketing, pp.53 - 66. Anderson, K., Carr, F., Feron, E. and Hall, W., 2000. Analysis and modeling of ground operations at hub airports. Third USA/Europe Air Traffic Managem ent R&D Seminar, Napoli, 13 16 June 2000. Anderson, S.W., Baggett, L.S. and Widener, S.K., 2009. The impact of service operations failures on customer satisfaction: evidence on how failures and their source affect what matters to customers. Manufacturing & Service Operations Management, 11(1), pp.52 - 69. Argote, L. and Miron - Spektor, E., 2011. Organizational learning: From experience to knowledge. Organization Science, 22(5), pp.1123 1137. Atkinson, S., Ramdas, K. and Williams, J.W., 2013. The costs of inef ficient robust scheduling practices in the US airline industry. Ayres, R.U. and Kneese, A.V., 1969. Production, consumption, and externalities. The American Economic Review, 59(3), pp.282 - 297. Azadegan, A., Patel, P.C. and Parida, V., 2013. Operational sla ck and venture survival. Production and Operations Management, 22(1), pp.1 - 18. Azan, W. and Huber S . I., 2010. Knowledge transfer in post merger integration management: Case study of a multinational healthcare company in Greece. Knowledge Management Resear ch & Practice, 8(4), pp.307 321. 128 Baker, D.M.A., 2013. Service quality and customer satisfaction in the airline industry: A comparison between legacy airlines and low - cost airlines. American Journal of Tourism Research, 2(1), pp.67 - 77. Baltagi, B.H., Griffi n, J.M. and Vadali, S.R., 1998. Excess capacity: a permanent characteristic of US airlines?. Journal of Applied Econometrics, 13(6), pp.645 - 657. Barone, G.J., Henrickson, K.E. and Voy, A., 2012, September. Baggage fees and airline stock performance: A case study of initial investor misperception. Journal of the Transportation Research Forum, 51 (1), pp.5 - 18. Behn, B.K. and Riley Jr, R.A., 1999. Using nonfinancial information to predict financial performance: The case of the US airline industry. Journal of A ccounting, Auditing and Finance, 14(1), pp.29 - 56. Bell, A. and Jones, K., 2015. Explaining fixed effects: Random effects modeling of time - series cross - sectional and panel data. Political Science Research and Methods, 3(1), pp.133 - 153. Bell, A., Fairbrother , M. and Jones, K., 2018. Fixed and random effects models: making an informed choice. Quality and Quantity, pp.1 - 24. Belobaba, P., 2005, November. Impacts of 9/11 on us airline performance. MIT Research and Development Conference. Ben - Akiva, M.E. and Lerma n, S.R., 1985. Discrete choice analysis: theory and application to travel demand (Vol. 9). MIT press. Berlin, J . A., Stephen E.K . , Thomas R. T .H. , and Mary D.S., 1999. An Empirical Comparison of Several Clustered Data Approaches under Confounding due to Cluster Effects in the Analysis of Complications of Coronary Angioplasty. Biometrics 55(2):470 6. Bhadra, D., 2009. Race to the bottom or swimming upstream: performance analysis of US airlines. Journal of Air Transport Management, 15(5), pp.227 - 235. Bilotk ach, V., 2011. Multimarket contact and intensity of competition: Evidence from an airline merger. Review of Industrial Organization, 38(1), pp.95 115. Bilotkach, V., Fageda, X. and Flores Fillol, R., 2013. Airline consolidation and the distribution of traf fic between primary and secondary hubs. Regional Science and Urban Economics, 43(6), pp. 951 963. Blettner, D.P., He, Z.L., Hu, S. and Bettis, R.A., 2015. Adaptive aspirations and performance heterogeneity: Attention allocation among multiple reference points. Strategic Management Journal, 36(7), pp.987 1005. Bliese, P.D. and Lang, J.W., 2016. Understanding relative and absolute change i n discontinuous growth models: Coding alternatives and implications for hypothesis testing. Organizational Research Methods, 19(4), pp.562 - 592. 129 Bliese, P.D. and Ployhart, R.E., 2002. Growth modeling using random coefficient models: Model building, testing, and illustrations. Organizational Research Methods, 5(4), pp.362 - 387. Bliese, P.D., Adler, A.B. and Flynn, P.J., 2017. Transition processes: A review and synthesis integrating methods and theory. Annual Review of Organizational Psychology and Organization al Behavior, 4, pp.263 - 286. Boehmer, E., Masumeci, J. and Poulsen, A.B., 1991. Event study methodology under conditions of event induced variance. Journal of Financial Economics, 30(2), pp.253 272. Bolat, A., 2000. Procedures for providing robust gate assi gnments for arriving aircrafts. European Journal of Operational Research, 120(1), pp.63 80. Bolton, R.N., 2004. Linking marketing to financial performance and firm value. Journal of Marketing, 68(4), pp.73 75. Borenstein, S., 1985. Airline mergers, airport dominance, and market power. The American Economic Review, 80(2), pp.400 404. Bourgeois III, L.J., 1981. On the measurement of organizational slack. Academy of Management Review, 6(1), pp.29 - 39. Bourland, K.E. and Yano, C.A., 1994. The strategic use of c apacity slack in the economic lot scheduling problem with random demand. Management Science, 40(12), pp.1690 - 1704. Bratu, S. and Barnhart, C., 2006. Flight operations recovery: New approaches considering passenger recovery. Journal of Scheduling, 9(3), pp. 279 - 298. Brueckner, J.K., 2004. Network structure and airline scheduling. The Journal of Industrial Economics, 52(2), pp.291 - 312. Bureau of Economic Analysis, May 18, 2019. Gross Domestic Product [GDP], retrieved from FRED, Federal Reserve Bank of St. Loui s; https://fred.stlouisfed.org/series/GDP. Accessed on May 25, 2019. Buzzell, R.D., Gale, B.T. and Sultan, R.G., 1975. Market share - a key to profitability. Harvard business review, 53(1), pp.97 - 106. America West. Retrieved from https://www.wsj.com/articles/SB111680696611140277. Accessed on May 25, 2019. Carlton, D., Israel, M., MacSwain, I. and Orlov, E., 2019. Are legacy airline mergers pro or anti competitive? Evidence from recent US airline merge rs. International Journal of Industrial Organization, 62, pp.58 95. Cartwright, S. and Schoenberg, R., 2006. Thirty years of mergers and acquisitions research: Recent advances and future opportunities. British journal of management, 17(S1), S1 - S5. 130 Chen, Y. and Gayle, P.G., 2019. Mergers and product quality: Evidence from the airline industr y. International Journal of Industrial Organization, 62, pp.96 135. Chowdhury, S., Schulz, E., Milner, M. and Van De Voort, D., 2014. Core employee based human capital and revenue productivity in small firms: An empirical investigation. Journal of Business Research, 67(11), pp.2473 2479. Clarke, M.D.D., 1998. Irregular airline operations: A review of the state of the practice in airline operations control centers. Journal of Air Transport Management, 4(2), pp.67 76. Collins, D.L., Román, F.J. and Chan, H.C. , 2011. An empirical investigation of the relationship between profitability persistence and firms' choice of business model: Evidence from the US airline industry. Journal of Management Accounting Research, 23(1), pp.37 - 70. Collins, J.D., Holcomb, T.R., C erto, S.T., Hitt, M.A. and Lester, R.H., 2009. Learning by doing: Cross - border mergers and acquisitions. Journal of business research, 62(12), pp.1329 1334. Cook, A., Tanner, G. and Lawes, A., 2012. The hidden cost of airline unpunctuality. Journal of Tran sport Economics and Policy, 46(2), pp.157 - 173. Crossan, M.M., Lane, H.W. and White, R.E., 1999. An organizational learning framework: From intuition to institution. Academy of Management Review, 24(3), pp.522 537. Cyert, R.M. and March, J.G., 1963. A behav ioral theory of the firm. Englewood Cliffs, NJ, 2(4), pp.169 187. Dana Jr, J.D. and Orlov, E., 2014. Internet penetration and capacity utilization in the US airline industry. American Economic Journal: Microeconomics, 6(4), pp.106 - 37. Datta DK, Pinches GP and Narayanan VK., 1992. Factors influencing wealth creation from mergers and acquisitions: a meta - analysis. Strategic Management Journal, 13(1), pp.67 84. Daugherty, P.J., Chen, H., Mattioda, D.D. and Grawe, S.J., 2009, Marketing/logistics relationships: Influence on capabilities and performance, Journal of Business Logistics, 30(1), pp.1 18. Dawson, J.F., 2014. Moderation in management research: What, Why, When, and How. Journal of Business and Psycho logy, 29(1), pp.1 - 19. Degani, A. and Wiener, E.L., 1997. Procedures in complex systems: The airline cockpit. IEEE Transactions on Systems, Man, and Cybernetics Part A: Systems and Humans, 27(3), pp.302 312. Degani, A. and Wiener, E.L., 1998. Design and ope rational aspects of flight deck procedures. International Air Transport Association (IATA): Human Factors. 131 Delta Airlines, 2008. Delta and Northwest merge, creating premier global airline. Retrieved from https://news.delta.com/delta and northwest merge cr eating premier global airline. Accessed on May 25, 2019. Department of Justice (DOJ), 2019. Statutory provisions and guidelines of the antitrust division. https://www.justice.gov/atr/file/761131/download. Accessed on May 25, 2019. Department of Transporta tion (DOT), 2017. The Economic Impact of Civil Aviation on the U.S. Economy, September 2017. Retrieved from https://www.faa.gov/about/plans_reports/media/2017 - economic - impact - report.pdf (Accessed on Jan 17, 2019). Department of Transportation, 2019. Data directory: list of databases. https://www.transtats.bts.gov/DataIndex.asp. Accessed on May 25, 2019. Diana, T., 2006. Benchmarking airport efficiency: an application of data envelopment analysis. Air Traffic Control Quarterly 14, pp.183 202. Dresner, M. and Xu, K., 1995. Customer service, customer satisfaction, and corporate performance. Journal of Business logistics, 16(1), pp.23 - 40. Dutton, J.M. and Thomas, A., 1984. Treating progress functions as a managerial opportunity. Academy of Managem ent Review, 9(2), pp.235 - 247. Ernst, A.T., Jiang, H., Krishnamoorthy, M. and Sier, D., 2004. Staff scheduling and rostering: A review of applications, methods and models. European journal of operational research, 153(1), pp.3 - 27. Ethiraj, S., Kale, P., Kri shnan, M. and Singh, J., 2005. Where do capabilities come from and how do they matter? A study in the software services industry. Strategic Management Journal, 26(1), pp.25 45. Fiol, C. and Lyles, M., 1985. Organizational learning. Academy of Management Re view, 10(4), pp.803 813. Fitzmaurice, G.M., Laird, N.M., and Ware, J.H., 2011. Applied Longitudinal Data Analysis. 2nd ed. Hoboken, NJ: John Wiley and Sons. Flouris, T. and post analysis of finan cial market information. Journal of Air Transport Management, 10(3), pp.173 180. Forbes, S.J., 2008a. The effect of air traffic delays on airline prices. International Journal of Industrial Organization, 26(5), pp.1218 - 1232. Forbes, S.J., 2008b. The effect of service quality and expectations on customer complaints. The Journal of industrial economics, 56(1), pp.190 - 213. 132 Garrow, L.A., Hotle, S. and Mumbower, S., 2012. Assessment of product debundling trends in the US airline industry: Customer service and pu blic policy implications. Transportation Research Part A: Policy and Practice, 46(2), pp.255 - 268. Garvin, D., 1988. Managing quality: The strategic and competitive edge. Simon and Schuster. Gayle, P. and Wu, C., 2013. A re se to the threat of entry: Evidence from the airline industry. Economics of Transportation, 2(4), pp.119 130. Gillen, D. and Lall, A., 2004. Competitive advantage of low - cost carriers: some implications for airports. Journal of Air Transport Management, 10 (1), pp.41 - 50. Gong, S. and Firth, M., 2006. The competitive effects of airline mergers and acquisitions: More capital market evidence. Journal of Air Transportation, 11(1), pp.88 105. Graham, D.R., Kaplan, D.P. and Sibley, D.S., 1983. Efficiency and compe tition in the airline industry. The Bell Journal of Economics, pp.118 - 138. Gudmundsson, S., Merkert, R. and Redondi, R., 2017. Cost functions and determinants of unit cost effects in horizontal airline M&As. Transportation Research Part A: Policy and Pract ice, 103, pp.444 454. Hahn, J., Todd, P. and Van der Klaauw, W., 2001. Identification and estimation of treatment effects with a regression oÀ discontinuity design. Econometrica, 69(1), pp. 201 p$ 209. Hale Jr, D., Ployhart, R. and Shepherd, W., 2016. A two phase longitudinal model of a turnover event: Disruption, recovery rates, and moderators of collective performance. Academy of Management Journal, 59(3), pp.906 929. Haleblian, J., Devers, C., McNamara, G., Carpenter, M.A. and Da vison, R.B., 2009. Taking stock of what we know about mergers and acquisitions: A review and research agenda. Journal of Management, 35(3), pp.469 502. Halstead, D., Morash, E.A. and Ozment, J., 1996. Comparing objective service failures and subjective com plaints: An investigation of domino and halo effects. Journal of Business Research, 36(2), pp.107 - 115. Hansen, G. and Wernerfelt, B., 1989. Determinants of firm performance: The relative importance of economic and organizational factors. Strategic Manageme nt Journal, 10(5), pp.399 411. Hansen, M.M., Gillen, D. and Djafarian - Tehrani, R., 2001. Aviation infrastructure performance and airline cost: a statistical cost estimation approach. Transportation Research Part E: Logistics and Transportation Review, 37(1 ), pp.1 - 23. Hendricks, K.B., Singhal, V.R. and Zhang, R., 2009. The effect of operational slack, diversification, and vertical relatedness on the stock market reaction to supply chain disruptions. Journal of Operations Management, 27(3), pp.233 - 246. 133 Henric kson, K.E. and Scott, J., 2012. Chapter 8 Baggage Fees and Changes in Airline Ticket Prices. In Pricing behavior and non - price characteristics in the airline industry. pp.177 - 192. Emerald Group Publishing Limited. Heron, R. and Lie, E., 2002. Operating per formance and the method of payment in takeovers. Journal of Financial and Quantitative Analysis, 37(1), pp.137 155. Higgens, M., 2010. Dodging Those Pesky Airline Fees. Retrieved from https://www.nytimes.com/2010/12/05/travel/05pracfees.html (Accessed on A pril 17, 2019). Hoffman, L. and Stawski, R.S., 2009. Persons as contexts: Evaluating between - person and within - person effects in longitudinal analysis. Research in Human Development, 6(2 - 3), pp.97 - 120. Hoffman, L., 2015. Longitudinal analysis: Modeling wit hin - person fluctuation and change. Routledge. Homberg, F., Rost, K. and Osterloh, M., 2009. Do synergies exist in related acquisitions? A meta - analysis of acquisition studies. Review of Managerial Science, 3(2), pp.75 116. Homsombat, W., Fu, X. and Sumalee , A., 2010. Policy implications of airline performance indicators: Analysis of major North American airlines. Transportation Research Record: Journal of the Transportation Research Board, (2177), pp.41 - 48. Hüschelrath, K. and Müller, K., 2014. Airline netw orks, mergers, and consumer welfare. Journal of Transport Economics and Policy, 48(3), pp.385 407. Hüschelrath, K. and Müller, K., 2015. Market power, efficiencies, and entry evidence from an airline merger. Managerial and Decision Economics, 36(4), pp.239 255. Huselid, M.A., 1995. The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38(3), pp.635 - 672. IATA, 2018. Fuel Fact Sheet. Retrieved from https://www.iata.org/ pressroom/facts_figures/fact_sheets/Documents/fact - sheet - fuel.pdf (Accessed on Jan 17, 2019). Iatrou, K. and Oretti, M., 2007. Airline choices for the future: From alliances to mergers. England: Ashgate Publishing Ltd. Imbens, G.W. and Lemieux, T., 2008. Regression discontinuity designs: A guide to practice. Journal of Econometrics, 142(2), pp.615 635. Industry High Level Group (IHLG), 2017. Retrieved from https://www.icao.int/sustainability/Documents/AVIATION - BENEFITS - 2017 - web.pdf (Acc essed on Jan 17, 2019). 134 Jain, V., 2015. What did the wave bring?: Short term price effect of the us airline merger wave (2009 2012). Journal of Economic Policy and Research, 10(2), pp.49 68. Jordan, W.A., 1988. Problems stemming from airline mergers and a cquisitions. Transportation Journal, 27(4), pp.9 30. Josephs, L., 2018. Years after airline mergers, flight attendants are finally flying together. Retrieved from https://www.cnbc.com/2018/10/01/years after airline mergers flight attendants start to fly to gether.html. Accessed on May 25, 2019. Kalemba, N. and Campa - Planas, F., 2017a. An overview of the quality concept in the air transportation business: a systematic literature review. International Journal for Quality Research, 11(1). Kalemba, N. and Campa - Planas, F., 2017b. The Quality Profitability Link in the US Airline Business: A Study Based on the Airline Quality Rating Index. Procedia Engineering, 187, pp.308 - 316. Katzenbach, J.R. and Smith, D.K., 2015. The wisdom of teams: Creating the high per formance organization. Harvard Business Review Press. Kavanagh, M.H. and Ashkanasy, N.M., 2006. The impact of leadership and change management strategy on organizational culture and individual acceptance of change during a merger. British journal of manage ment, 17(S1), S81 S103. Kelton, W.D., 2002. Simulation with ARENA. McGraw - hill. Kim, E.H. and Singal, V., 1993. Mergers and market power: Evidence from the airline industry. The American Economic Review, 83(3), pp.549 569. Kim, J.Y., Kim, J.Y. and Miner, A .S., 2009. Organizational learning from extreme performance experience: The impact of success and recovery experience. Organization Science, 20(6), pp.958 978. Kim, Y., and Ployhart, R. E., 2014. The effects of staffing and training on firm productivity an d profit growth before, during, and after the great recession. Journal of Applied Psychology, 99, pp.361 - 389. King, A.W. and Zeithaml, C.P., 2001. Competencies and firm performance: Examining the causal ambiguity paradox. Strategic Management Journal, 22(1 ), pp.75 99. King, D.R., Dalton, D.R., Daily, C.M. and Covin, J.G., 2004. Meta oÀ analyses of post oÀ acquisition performance: Indications of unidentified moderators. Strategic Management Journal, 25(2), pp. 187 p$ 200. Kleinrock, L., 1975. Queueing Theory. Vol. i, Wiley, New York. 135 Kleinrock, L., 1976. Queueing Systems, Volume 2: Computer applications (Vol. 66). New York: Wiley. Knapp, W., 1990. Event analysis of air carrier mergers and acquisitions. The Review of Economics and Statistics, 72(4), pp.703 707. Kovach, J.J., Hora, M., Manikas, A. and Patel, P.C., 2015. Firm performance in dynamic environments: The role of operational slack and operational scope. Journal of Operations Management, 37, pp.1 - 12. Kwoka, J. and Shumilkina, E., 2010. The price effect of eliminating potential competition: Evidence from an airline merger. The Journal of Industrial Economics, 58(4), pp.767 793. Lampel, J., Shamsie, J. and Shapira, Z., 2009. Experiencing the improbable: Rare events a nd organizational learning. Organization Science, 20(5), pp.835 845. Landon, E. L., 1977. A Model of Consumer Complaining Behavior. In Consumer Satisfaction, Dissatisfaction, and Complaining Behavior, ed. R.L. Day, Bloomington, Indiana: Bureau of Business Research, pp.31 - 35. Lang, J.W. and Bliese, P.D., 2009. General mental ability and two types of adaptation to unforeseen change: Applying discontinuous growth models to the task change paradigm. Journal of Applied Psychology, 94(2), pp.411 428. Lang, L.H., Stulz, R. and Walkling, R.A., 1989. Managerial performance, Tobin's Q, and the gains from successful tender offers. Journal of Financial Economics, 24(1), pp.137 154. Lapr pv , M.A. and Scudder, G.D., 2004. Performance improvement paths in the US airline ind ustry: Linking trade oÀ offs to asset frontiers. Production and Operations Management, 13(2), pp.123 - 134. Lapré, M.A. and Tsikriktsis, N., 2006. Organizational learning curves for customer dissatisfaction: Heterogeneity across airlines. Management Science, 52 (3), pp.352 - 366. Latham, S. F., and Braun, M. R., 2008. The performance implications of financial slack during economic recession and recovery: Observations from the software industry (2001 2003). Journal of Managerial Issues, 20, pp.30 50. Lazarus, R.S. a nd Folkman, S., 1984. Stress, appraisal, and coping. Springer publishing company. Lazarus, R.S., 1991. Emotion and adaptation. Oxford University Press on Demand. Leibenstein, H., 1966. Allocative efficiency vs. X - efficiency. The American Economic Review, 5 6(3), 392 415. 136 Leroy, F. and Ramanantsoa, B., 1997. The cognitive and behavioural dimensions of organizational learning in a merger: An empirical study. Journal of Management Studies, 34(6), pp.871 894. Levitt, B. and March, J.G., 1988. Organizational lear ning. Annual Review of Sociology, 14(1), pp.319 338. Levy, M., 2011. Knowledge retention: minimizing organizational business loss. Journal of Knowledge Management, 15(4), pp.582 - 600. Lichtenberg, F.R. and Kim, M., 1989. The effects of mergers on prices, costs, and capacity utilization in the US air transportation industry, pp.1970 1984 (No. w3197). National Bureau of Economic Research. Loukopoulos, L.D., Dismukes, R.K. and Barshi, I., 2003. Concurrent task demands in the cockpit: Challenges and vulnerabilities in routine flight operations. In Proceedings of the 12th International Symposium on Aviation Psychology. pp.737 742. The Wright State University Dayton, OH. Luo, D., 2014. The price effects of the Del ta/Northwest airline merger. Review of Industrial Organization, 44(1), pp.27 48. Mannor, M.J., Shamsie, J. and Conlon, D.E., 2016. Does experience help or hinder top managers? Working with different types of resources in Hollywood. Strategic Management Journal, 37(7), pp.1330 1340. Mantin, B. and Wang, J.H.E., 2012. Determinants of profitability and recovery from system - wide shocks: The case of the airline industry. Journal of Airline and Airport Management, 2(1), pp.1 - 21. Manuela Jr, W.S., Rhoades, D.L. and Curtis, T., 2016. The US Airways Group: A post merger analysis. Journal of Air Transport Management, 56, pp. 138 150. March, J.G., 1991. Exploration and exploitation in organizational learning. Organization Science, 2(1), pp.71 87. March, J.G., Sproull, L.S. and Tamuz, M., 1991. Learning from samples of one or fewer. Organization science, 2(1), pp.1 - 13. Markides, C.C. and Williamson, P.J., 1994. Related diversification, core competences and corporate performance. Strategic Management Journal, 15(S2), pp.149 165. Marsh, S.J. and Stock, G.N., 2006. Creating dynamic capability: The role of intertemporal integration, knowledge retention, and interpretation. Journal of Product Innovation Management, 23(5), pp.422 436. Mayer, C., and Sinai, T., 2003. Network effects, congestion externalities, and air traffic delays: Or why not all delays are evil. American Economic Revie w, 93(4), pp.1194 - 1215. 137 McCartney, S., 2008a. Space race: A battle looms for the overhead bins. Wall Street Journal (June 17) D1 D2. (September 2) D1 D3. McCartney, S., 2008c. What it costs an airline to fly your luggage. Wall Street Journal (November 25) D1 D8. D4. McCartney, S., 2010b. An airline report card: Fewer delays, hassl es last year, but bumpy times may be ahead. Wall Street Journal (January 7) D1 D3. McCartney, S., 2012a. The tough tactics to avoid luggage check - in fees. Wall Street Journal (February 2) D1 D3. McCartney, S., 2012b. Reality check: Why airlines are shrinki ng flight times. Wall Street Journal (June 14) D1 D2. McCartney, S., 2010. An airline report card: Fewer delays, hassles last year, but bumpy times may be ahead. Wall Street Journal (January 7) D1 D3. McCartney, S., 2015. The Best and Worst Airlines. Wall Street Journal (January 15). Retrieved from https://www.wsj.com/articles/the - best - and - worst - airlines - 1421254623. Mellat - Parast, M., Golmohammadi, D., McFadden, K.L. and Miller, J.W., 2015. Linking business strategy to service failures and financial perfor mance: Empirical evidence from the US domestic airline industry. Journal of Operations Management, 38, pp.14 - 24. Merkert, R. and Morrell, P.S., 2012. Mergers and acquisitions in aviation Management and economic perspectives on the size of airlines. Transp ortation Research Part E: Logistics and Transportation Review, 48(4), pp.853 862. Mesquita, L.F., Anand, J. and Brush, T.H., 2008. Comparing the resource oÀ based and relational views: Knowledge transfer and spillover in vertical alliances. Strategic Manageme nt Journal, 29(9), pp. 913 p$ 941. Miller, D., 1996. A preliminary typology of organizational learning: Synthesizing the literature. Journal of Management, 22(3), pp.485 505. Mishina, Y., Pollock, T.G. and Porac, J.F., 2004. Are more resources always better fo r growth? Resource stickiness in market and product expansion. Strategic Management Journal, 25(12), pp.1179 - 1197. Mitra, T.A., 2001. Measuring the causes of airline customer dissatisfaction (Doctoral dissertation, Massachusetts Institute of Technology). 138 M orash, E.A., 2001, Supply chain strategies, capabilities, and performance. Transportation Journal, 41(1), pp.37 54. Morrison, S.A. and Winston, C., 1989. Enhancing the performance of the deregulated air transportation. Brookings Papers on Economic Activity, 61. possible cures. Hearing before the Subcommittee on Aviation, Committee on Transportation and Infrastructure, United States House of Representatives. Morrison, S.A., 1996. Airline mergers: A longer view. Journal of Transport Economics and Policy, 30(3), pp.237 250. Moss, D.L. and Mitchel l, K., 2012. The proposed merger of US airways and American airlines: The rush to closed airline systems. White Paper produced jointly by the American Antitrust Institute (AAI) and Business Travel Coalition (BTC). Mouawad, J., 2012. For United, big proble ms at biggest airline. Retrieved from https://www.cnbc.com/id/50009207. Accessed on May 25, 2019. Mowery, D.C., Oxley, J.E. and Silverman, B.S., 1996. Strategic alliances and interfirm knowledge transfer. Strategic Management Journal, 17(S2), pp.77 91. Mu ndlak, Y., 1978. On the pooling of time series and cross section data. Econometrica: Journal of the Econometric Society, pp.69 - 85. Nahmias, S. and Cheng, Y., 2009. Production and Operations Analysis (Vol. 6). New York: McGraw - hill. Nakagawa, S. and Schielz eth, H., 2013. A general and simple method for obtaining R2 from generalized linear mixed oÀ effects models. Methods in Ecology and Evolution, 4(2), pp.133 - 142. Nguyen, H.T., Yung, K. and Sun, Q., 2012. Motives for mergers and acquisitions: Ex oÀ post market evi dence from the US. Journal of Business Finance & Accounting, 39(9 oÀ 10), pp. 1357 - 1375. analysis of the operational implications of airline baggage fees. Management Science, 63(10), pp.3187 - 3206. Nonaka, I., 1994. A dynamic theory of organizational knowledge creation. Organization Science, 5(1), pp.14 37. Ozment, J. and Morash, E.A., 1994. The augmented service offering for perceived and actual service quality. Journa l of the Academy of Marketing Science, 22(4), pp.352 - 363. 139 Palich, L.E., Cardinal, L.B. and Miller, C.C., 2000. Curvilinearity in the diversification performance linkage: An examination of over three decades of research. Strategic Management Journal, 21(2), pp.155 174. Papadakos, N., 2009. Integrated airline scheduling. Computers & Operations Research, 36(1), pp.176 - 195. Peters, C., 2006. Evaluating the performance of merger simulation: Evidence from the US airline industry. The Journal of Law and Economics, 49(2), pp.627 649. Peterson, E.B., Neels, K., Barczi, N. and Graham, T., 2013. The economic cost of airline flight delay. Journal of Transport Economics and Policy (JTEP), 47(1), pp.107 - 121. Piger, J . M . and Chauvet, M . , Smoothed U.S. Recession Probabilit ies [RECPROUSM156N], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/RECPROUSM156N. (Accessed on April 18, 2019). Prescott, J.E., Kohli, A.K. and Venkatraman, N., 1986. The market share oÀ profitability relationship : An empirical assessment of major assertions and contradictions. Strategic Management Journal, 7(4), pp.377 - 394. Prince, J.T. and Simon, D.H., 2017. The impact of mergers on quality provision: Evidence from the airline industry. The Journal of Industrial Economics, 65(2), pp.336 362. Prince, J.T. and Simon, D.H., 2009. Multimarket contact and service quality: Evidence from on - time performance in the US airline industry. Academy of Management Journal, 52(2), pp.336 - 354. Ramaswamy, K., Thomas, A.S. and Litschert, R.J., 1994. Organizational performance in a regulated environment: the role of strategic orientation. Strategic Management Journal, 15(1), pp.63 - 74. Ramdas, K. and Williams, J., 2008. An empirical investigation into the tradeoffs that impact on - time performance in the airline industry. Washington Post, pp.1 - 32. Ramdas, K., Williams, J. and Lipson, M., 2013. Can financial markets inform operational improvement efforts? Evidence from the airline industry. Manufacturing & Service Operations Manageme nt, 15(3), pp.405 - 422. Raudenbush, S.W. and Bryk, A.S., 2002. Hierarchical linear models: Applications and data analysis methods (Vol. 1). Sage. Raudenbush, S.W., 2001. Comparing personal trajectories and drawing causal inferences from longitudinal data. A nnual Review of Psychology, 52(1), pp.501 - 525. Reus, T.H., Lamont, B.T. and Ellis, K.M., 2016. A darker side of knowledge transfer following international acquisitions. Strategic Management Journal, 37(5), pp.932 944. 140 Rhoades, D.L. and Waguespack Jr, B., 2008. Twenty years of service quality performance in the US airline industry. Managing Service Quality: An International Journal, 18(1), pp.20 - 33. Richard, O., 2003. Flight frequency and mergers in airline markets. International Journal of Industrial O rganization, 21(6), pp.907 922. Rupp, N., Owens, D. and Plumly, L., 2006. Does competition influence airline on - time performance? Advances in Airline Economics, 1, pp.251 - 272. Rupp, N.G. and Tan, K.M., 2019. Mergers and product quality: A silver lining fro m de hubbing in the U.S airline industry. Contemporary Economic Policy. Advance online publication. https://doi.org/10.1111/coep.12419. Rupp, N.G. and Holmes, G.M., 2006. An investigation into the determinants of flight cancellations. Economica, 73(292), pp.749 - 783. Ryerson, M.S. and Kim, H., 2014. The impact of airline mergers and hub reorganization on aviation fuel consumption. Journal of Cleaner Production, 85, pp.395 407. Saranga, H. and Nagpal, R., 2016. Drivers of operational efficiency and its impac t on market performance in the Indian Airline industry. Journal of Air Transport Management, 53, pp.165 - 176. Schefczyk, M., 1993. Operational performance of airlines: an extension of traditional measurement paradigms. Strategic Management Journal, 14(4), p p.301 - 317. Scheraga, C.A., 2004. Operational efficiency versus financial mobility in the global airline industry: a data envelopment and Tobit analysis. Transportation Research Part A: Policy and Practice, 38(5), pp.383 - 404. Scherer, K.R., Schorr, A. and J ohnstone, T. eds., 2001. Appraisal processes in emotion: Theory, methods, research. Oxford University Press. Schilling, M.A., Vidal, P., Ployhart, R.E. & Marangoni, A., 2003. Learning by doing something else: Variation, relatedness, and the learning curve. Management Science, 49(1), pp.39 - 56. Schmenner, R.W. and Swink, M.L., 1998. On theory in operations management. Journal of operations management, 17(1), pp.97 - 113. Schmidt, A., 2016. How did American airlines recover from bankruptcy? Retrieved from https: //marketrealist.com/2016/06/american airlines turnaround bankruptcy/. Accessed on May 25, 2019. Schosser, M. and Wittmer, A., 2015. Cost and revenue synergies in airline mergers Examining geographical differences. Journal of Air Transport Management, 47, p p.142 153. Schumann, H.O. and Singh, H., 2014. The market structure dynamics created by de - bundling of airline bag fees. Journal of Economics and Economic Education Research, 15(3), pp.187 - 228. 141 Schurer, S. and Yong, J., 2012. Personality, well - being and the marginal utility of income: What can we learn from random coefficient models? Scotti, D. and Dresner, M., 2015. The impact of baggage fees on passenger demand on US air routes. Transport Policy, 43, pp.4 - 10. Scotti, D., Dresner, M. and Martini, G., 201 6. Baggage fees, operational performance and customer satisfaction in the US air transport industry. Journal of Air Transport Management, 55, pp.139 - 146. Servaes, H., 1991. Tobin's Q and the Gains from Takeovers. The Journal of Finance, 46(1), pp.409 419. Seth, A., Song, K. and Business Studies, 31, pp.387 405. Shaffer, B., Quasney, T.J. and Grimm, C.M., 2000 . Firm level performance implications of nonmarket actions. Business and Society, 39(2), pp.126 - 143. Shen, Y., 2017. Market competition and market price: Evidence from United/Continental airline merger. Economics of Transportation, 10, pp.1 7. Shenkar, O. and Li, J., 1999. Knowledge search in international cooperative ventures. Organization Science, 10(2), pp.134 143. Sim, K.L, Joon Song, C. J. and Killough, L.N., 2010. Service quality, service recovery, and financial performance: An analysis of the US airline industry. In Advances in Management Accounting. pp. 27 - 53. Emerald Group Publishing Limited. Singal, V., 1996a. Airline mergers and multimarket contact. Managerial and Decision Economics, 17(6), pp.559 574. Singal, V., 1996b. Airline merger s and competition: an integration of stock and product price effects. Journal of Business, 69(2), pp.233 268. Singer, J.D., Willett, J.B. and Willett, J.B., 2003. Applied longitudinal data analysis: Modeling change and event occurrence. Oxford University P ress. Spiller, P.T., 1983. The differential impact of airline regulation on individual firms and markets: An empirical analysis. The Journal of Law and Economics, 26(3), pp.655 - 689. Steele, D.C. and Papke - Shields, K.E., 1993. Capacity slack: strategic alternative to lead time. Production and Inventory Management Journal, 34(4), p.1. Steven, A.B., Dong, Y. and Dresner, M., 2012. Linkages between customer service, customer satisfaction and performance in the airline industry: Investigation of no n - linearities and 142 moderating effects. Transportation Research Part E: Logistics and Transportation Review, 48(4), pp.743 - 754. Steven, A.B., Yazdi, A.A. and Dresner, M., 2016. Mergers and service quality in the airline industry: A silver lining for air trav elers?. Transportation Research Part E: Logistics and Transportation Review, 89, pp.1 13. Surana, A., Kumara, S., Greaves, M. and Raghavan, U.N., 2005. Supply - chain networks: a complex adaptive systems perspective. International Journal of Production Resea rch, 43(20), pp.4235 - 4265. Suzuki, Y., 2000. The relationship between on - time performance and airline market share: a new approach. Transportation Research Part E: Logistics and Transportation Review, 36(2), pp.139 - 154. Suzuki, Y., 2004. The impact of airl ine service failures on travelers' carrier choice: A case study of central Iowa. Transportation Journal, pp.26 - 36. Szymanski, D.M., Bharadwaj, S.G. and Varadarajan, P.R., 1993. An analysis of the market share - profitability relationship. Journal of Marketin g, pp.1 - 18. Tan, J. and Peng, M.W., 2003. Organizational slack and firm performance during economic transitions: Two studies from an emerging economy. Strategic Management Journal, 24(13), pp.1249 - 1263. Tanriverdi, H., and Venkatraman, N. 2005. Knowledge r elatedness and the performance of multibusiness firms. Strategic management journal, 26(2), pp.97 - 119. Teece, D. J. 1982. Towards an economic theory of the multiproduct firm. Journal of Economic Behavior & Organization, 3(1), pp.39 - 63. Teece, D.J., Pisano, G. and Shuen, A., 1997. Dynamic capabilities and strategic management. Strategic Management Journal, 18(7), pp.509 533. Tsikriktsis, N., 2007. The effect of operational performance and focus on profitability: A longitudinal study of the US airline industr y. Manufacturing & Service Operations Management, 9(4), pp.506 - 517. Turgot, A.R.J., Observation sur un Mémoire de M. de Saint - Pérvay, ed., Schelle, G., CEuvres de Turgot et Documents le Concernant, 2. Paris: Librairie Félix Alcan, [1914], 644. Originally p ublished in (1767). Tuttle, B., 2014. Travelers: It's About Damn Time Airlines Toughened Up Baggage Policies. Retrieved from http://time.com/12835/travelers - its - about - damn - time - airlines - toughened - up - baggage - policies/ (Accessed on Jan 17, 2019). 143 Tuzovic, S Acceptability of airline ancillary fees and the effects on customer behavior. Journal of Retailing and Consumer Services, 21(2), pp.98 - 107. Vaze, V., Luo, T. and Harder, R., 2017. Impacts of airline mergers on passenger welfare. Transportation Research Part E: Logistics and Transportation Review, 101, pp.130 154. Voss, G.B., Sirdeshmukh, D. and Voss, Z.G., 2008. The effects of slack resources and environmental threat on pr oduct exploration and exploitation. Academy of Management Journal, 51(1), pp.147 - 164. Wang, C. L., and Ahmed, P. K., 2003. Organisational learning: a critical review. The learning organization, 10(1), pp.8 - 17. Wenner, C.A. and Drury, C.G., 2000. Analyzing human error in aircraft ground damage incidents. International Journal of Industrial Ergonomics, 26(2), pp.177 199. Werden, G.J., Joskow, A.S. and Johnson, R.L., 1991. The effects of mergers on price and output: Two case studi es from the airline industry. Managerial and Decision Economics, 12(5), pp.341 352. Wooldridge, J.M., 2010. Econometric analysis of cross section and panel data. MIT press. Wooldridge, J.M., 2015. Introductory econometrics: A modern approach. Nelson Educat ion. Wyckoff, D. D. and Maister, D. H., 1977. The Domestic Airline Industry, Lexington Books, Lexington, MA. Xia, L., Kent B. M., and Jennifer L. C., 2004. The Price Is Unfair! A Conceptual Framework of Price Fairness Perceptions. Journal of Marketing, 68, p.p.1 15. Yan, J., Fu, X., Oum, T.H. and Wang, K., 2019. Airline horizontal mergers and productivity: Empirical evidence from a quasi natural experiment in China. International Journal of Industrial Organization, 62, pp.358 376. Yazdi, A.A., Dutta, P. and Steven, A.B., 2017. Airline baggage fees and flight delays: A floor wax and dessert topping?. Transportation Research Part E: Logistics and Transportation Review, 104, pp.83 - 96. Yimga, J., 2017. Airline on - time performance and its effects on consumer choice behavior. Research in Transportation Economics, 66, pp.12 - 25. Yu, C., 2016. Airline productivity and efficiency: concept, measurement, and applications. In Airline Efficiency. pp. 11 - 53. Emerald Group Publishing Limited. Zollo, M. and Singh , H., 2004. Deliberate learning in corporate acquisitions: post oÀ acquisition strategies and integration capability in US bank mergers. Strategic management journal, 25(13), pp. 1233 p$ 1256. 144 Zollo, M., 2009. Superstitious learning with rare strategic decisions: Theory and evidence from corporate acquisitions. Organization Science, 20(5), pp.894 908. Zott, C., 2003. Dynamic capabilities and the emergence of intra industry differential firm performance: Insights from a simulation study. Strategic Management Journa l, 24(2), pp.97 125. Zou, L. and Chen, X., 2017. The effect of code - sharing alliances on airline profitability. Journal of Air Transport Management, 58, pp.50 - 57. Zou, L., Yu, C., Rhoades, D. and Waguespack, B., 2017. The pricing responses of non - bag fee a irlines to the use of bag fees in the US air travel market. Journal of Air Transport Management, 65, pp.209 - 219.