OPERATING DECISIONS IN THE AFTERGLOW OF A SPIKE IN BUSINESS ACTIVITIES: EVIDENCE FROM BANK S B y Hariharan Ramasubramanian A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Business Administration Accounting Doctor of Philosophy 2021 ABSTRACT OPERATING DECISIONS IN THE AFTERGLOW OF A SPIKE IN BUSINESS ACTIVITIES: EVIDENCE FROM BANKS By Hariharan Ramasubramanian Firms are required to make resource adjustments and product mix ch anges in response to a n unexpected change in the operating environment. Despite the ubiquitous nature of such changes, little is known about the nature of such resource adjustments and product mix decisions when firms move from one steady state to another but face an intermediate period of uncertainty . U sing shale oil and natural gas extraction as an exogenous positive economic shock to the operations of local banks, I find that banks reduce labor cost elasticity and increase labor employee elasticity in re sponse to an unexpected po sitive change to their operating environment . Further, labor employee elasticity increases during the later periods of the shale development when there is low er uncertainty regarding the persistence of the positive economic shock . B anks with high er forecasting ability undertake labor adjustments earlier than banks with low er forecasting ability , highlighting the importance of the internal information environment in resolving uncertainty in the operating environment. During the late r periods of the s hale boom, b anks reduce product diversity when their downside demand risk reduces reliably . Overall, results suggest that managers make dynamic adjustments to their operations in response to an unexpected change in the operating environme nt. iii To my family & friends iv ACKNOWLEDGMENTS My time at Michigan State University (MSU) has been a phase of learning and growing. Transitioning from 30 - degree Celsius to 10 - degree Fahrenheit has been nothing less than an adventure. Moreover, living wi th li mited resources in a foreign country and starting afresh has been a humbling experience. I would not have been able to complete this program without the contributions of fantastic people I met in my life. First, I want to thank my advisor Dr. Ranjani Kris hnan. She has not only been a fabulous mentor and guide during my PhD program but also a great support for me and my wife at the personal level. I will never forget the efforts she took in discussing the opportunities for my wife even before we came t o the US. She has not only been kind to accept me as her student but has spent numerous hours every time I have r equired her guidance. I will continue to seek her guidance and wisdom in the future. Her passion and dedication for research, teaching, and service is inspirational to say the least. I would also like to thank and express my sincere gratitude to the member s of my dissertation committee: Dr. Kyonghee Kim, Dr. Martin Holzhacker, and Dr. Jeffrey Wooldridge for their guidance and support not only during m y dissertation but throughout the program. Few important incidents in my life shaped my decision to pursue a doctoral program . In 2003, I was in the first year of my college when I read Freakonomics and was fascinated with the idea of using data to study interesting questions. Same year, Mr. Anand Subramanian took a chance with me to teach school students in hi s coaching institute. That was the first time I taught in a formal setting and I realize d that I truly enjoy teaching. In 2005, Mr. P. R. S. Mani ga ve an opportunity to teach pre - college mathematics , which allowed me to improvise my teaching v methods . After completing my Chartered Accountancy (CA) in 2008, I wanted to become a full - time instructor and Mr. B. Nandakumar provided an opportunity to teach CA students and he was brave to assign the responsibility of teaching multiple class es with class sizes rang ing from 100 - 150 students . I thoroughly enjoyed the opportunity and responsibility to work with smart CA students and mentor them in their journey. Working with students on various social projects such as plastic waste segregation, cleaning railway station s, fund raising for old age home, etc. is easily the most satisfying experience. I have been fortunate enough to be involved in a few social project s under the leadership of Mr. Prasad Chikshe. In 2012, I met Mr. Vinay Nair and a fantastic professional col laboration began in the field of math education. It is very rare to meet passionate educators like Vinay. Together with Mr. Vinay Nair, Mr. Mandar B hanushe, and my dear friend , Mr. Govinda Goyal, I cofounded the Raising a Mathematician Foundation (RAM Foun dation) with an objective of working towards long - term goals of education. Contributing to this not - for - profit venture in any capacity is always sat isfying and I am hoping that now I will be able to mentor students in more meaningful ways. I cannot thank Dr. Kumar Venkataraman enough because without him I would not have taken the plunge into the PhD program. I did not know what a career in academia lo oks like and was clueless about research in accounting. My selection in a prestigious PhD program would have been impossible without his guidance. S. K. Patil school was my primary and secondary school which provided a strong foundation. The school had li My schoolteachers were not motivated by monetary incentives and t ook teaching as a big responsibility and that has made a lasting impact on me. My education in accounting began at the M ulund College of Commerce (MCC) and its low tuition fees was instrumental in helping me finish my college . I am grateful to the Indian t axpayers , donors, and college truste e s because of vi whom I could afford the tuition fees . I want to thank all my teachers in the school and college for buildin g the necessary foundation. In the past several years, I had the good fortune of learning from extremely passionate teachers. I will be forever grateful to Mr. B. Nandakumar, Mr. Harish Menon, and Mr. Divyang Thakker, who are easily some of the best teache rs I learnt from and responsible for whatever little understanding of accounting I have. During my PhD program, I took several advanced courses in economics and accounting to build my research skills. I am thankful to all the faculty members who taught me at Michigan State University. Particularly, I want to thank Dr. Jeffrey Wooldridge for teaching four courses on econometrics , which I th oroughly enjoyed, and these courses have been a tremendous learning experience . Two faculty members, Dr. Andrew Acito a nd Dr. Dan Wangerin, of the accounting and information systems department at MSU were always welcoming to discuss research ideas and thi s made a huge differe nce during my initial years of the PhD program. My PhD mates have been nothing less than awesome an d each one of them is always available to discuss anything. Particularly, I am grateful to James Anderson, Joanna Shaw, Aaron Fritz, Ais hwarrya Deore, Sarah Stuber, Anh Persson, Luke Weiler, and Michael Shen. I will never forget the relentless support of t he accounting and information systems department Chairpersons, Dr. V. Sambamurthy and Dr. Chris Hogan . I would also like to thank Karla Bauer, Katie Trinklein, Joyce Hengesbach, Barbara Ritenburg , Jessica Harrington, Aybige Kocas for all the administrative support they provide to PhD students like me . I am grateful for the financial support provided by Broad College of Business, MSU . I am also thankful to the US taxpayers and donors who have subsidized my education at MSU. vii This PhD journey would have been dreadful wi thout the company of wonderful friends at MSU. Particularly, I want to thank Abhishek, Nishu, Ayaan, and Nysa Jindal, Pratap Bhanu Solanki, A mruta , Tanmay, and Ishayu Desai, Harshvardhan Kalbhor, Tanvi Nikhar, Charuta Parkhi, Salil Sapre, Harshad Badyani, Prithvi Godavarthi, Yashesh Dhebar, Rahul Dey, Anuj Pal, Ronak Sripal, and Hitesh Gakhar, among others for making this journey worthwhile. I am fortunate enough to have some fantastic friends who have been extremely supportive throughout my life. Particula rly, Govinda Goyal, Nishit Zaveri, Kajal Palan , Jatin Chandan, Rakhi Chavan, Niveditha Ramakrishnan, Hiral Nandu, Mamta Pinjani, Monica and Sonica Panjwani, Ranju Nair, Kalpana Karmarkar, Mitul Doshi, Abhishek P aranjape, Pratik Hingorani, Esaivanan Sudanthirakody, Saurabh Pha tak, Himanshu Patwardhan, Swapnil Panhale, Moh it Sathe , Yoges h Chitte, Saket Agte, Mandar Dixit, Pushkar Joshi, Pratima Patil, Pooja Joshi, Dhaval Kothari, Monica Kadam, Vikrant Kadam, among others have made my college days nothing short of awesome. During my childhood, I had the good fortune of interacting with some of the most talented friends and those conversations sparked my int erest in various fields. Particularly, my childhood days spent with Shrikrishna Samant, Santosh Sai Venkat, Saurabh Bhowar, Sagar Raste, Aditya Nakhare, Aniket Bodas, Pushkar Phatak, Rohit Bennur, Kejal Joshi, Vijaya Viswanathan, Rohit Malvankar, Rushikesh Bodas, Mayur Gon dhalekar, and Mrunal Karkera will always be memorable. My interactions with these fantastic people have been a tremendous source of unstructured learning which played a pivotal role in shaping my interests in varied areas ranging from sports to politics. I met some fantastic people in my professional capacity but became close friends and are a blessing in my life. Particularly, I want to mention Anand Subra manian, Swetha Anand, Jayaprakash Rajangam, Pradeep Jain, Vijay Sonaje, Anil viii Kshatriya, and P. R. S. Mani. I am also grateful for the blessings of many people who have silently wished for my success. My wife, Hetal Adhia is one of the most understanding an d loving people I have met. The difficult parts of the PhD journey were made bearable by her companionship. I believe that she is resp onsible for my sanity and her outlook towards life reminds me that a career in academia is a marathon and not a sprint. He r positivity and kindness make a big difference. The last lap of the port it would have been impossible to graduate and secure a job . As cliché as it may sound, I will say that I may be the o ne getting the degree, but it is earned by both of us. I am also thankful to my brother - in - law, Kartik Adhia, and his wife, Shraddha P andey, for always supporting us both Bhadresh and Urvashi Adhia, are some of the best people I have met in my life. They have been nothing but supportive in every decision we have taken. My relatives and Hetal s relative s have been extremely supp ortive and have always wished for our well - being and suc cess. Finally, I am continually grateful to my parents, Ramasubramanian and Rajalakshmi, for everything. Words are not enough to describe their contribution in my life. They have always supported in all my decisions and given me the freedom to make bizarre career choices. They have provided unwavering support t o me in treading unusual paths. My brother, Natarajan has been a friend, philosopher, and guide throughout my life. If it were not for him, I would never have mustered the courage to pursue a PhD. Having a responsible older brother is a great insurance pol icy and it would have been extremely difficult to walk the paths I did. I am also grateful for the support of m y sister - in - law, Cini Natarajan . M y beautiful nieces, Anoushka and Ishanvi , made this journey more enjoyable. ix TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ .......................... x L I ST OF FIGURES ................................ ................................ ................................ ....................... xi INTRODUCTION ................................ ................................ ................................ .......................... 1 RELATED LITERATURE AND HYPOTHESES ................................ ................................ ...... 11 The oretical background ................................ ................................ ................................ ............. 11 Labor cost management ................................ ................................ ................................ ............ 14 Shale boom and bank operations ................................ ................................ ............................... 15 Labor cost management in banks in response to a positive economic shock ........................... 17 Uncertainty, dynamic resource adjustments , and real options ................................ .................. 19 ................................ ........................... 22 Product diversity ................................ ................................ ................................ ....................... 23 RESEARCH DESIGN AND METHODOLOGY ................................ ................................ ........ 25 Banks exposed to the shale boom ................................ ................................ ............................. 25 Re search design ................................ ................................ ................................ ......................... 26 Hypothesis 1 - 3: Effect of change in operating environment on labor elasticity ................... 27 Hypothesis 4: Effect of change in operating environment on product diversity ................... 32 DATA AND RESULTS ................................ ................................ ................................ ............... 34 Sample ................................ ................................ ................................ ................................ ....... 34 Descriptive statistics ................................ ................................ ................................ .................. 34 Hypothesis 1 - 3: ........ 35 ........... 41 ROBUSTNESS TESTS ................................ ................................ ................................ ................ 43 Parallel trends assumption ................................ ................................ ................................ ......... 43 Compar ing short run and long run labor elasticity ................................ ................................ .... 44 Additional tests ................................ ................................ ................................ .......................... 45 CONCLUSION ................................ ................................ ................................ ............................. 47 APPENDICES ................................ ................................ ................................ .............................. 49 APPENDIX A : Variable D efinitions ................................ ................................ ........................ 50 APPENDIX B : Tables ................................ ................................ ................................ ............... 53 APPENDIX C : Figures ................................ ................................ ................................ ............. 68 REFERENCES ................................ ................................ ................................ ............................. 74 x L IST OF TABLES TABLE 1.1: SHALE - BOOM EXPOSED BANKS AND GROWTH IN DEPOSITS AND INCOME ................................ ................................ ................................ ................................ ... 54 TABLE 1.2: SAMPLE CONSTRUCTION ................................ ................................ .................. 55 TABLE 2.1: DE SCRIPTIVE STATISTICS ................................ ................................ ................. 56 TABLE 2.2: CORRELATIONS ................................ ................................ ................................ ... 58 TABLE 2.3: DIFFERENCE IN MEANS ................................ ................................ ..................... 59 TABLE 3 .1 : DIFFERENCE IN LABOR COST ELASTICITY FOR SHALE BOOM EXPOSED BANKS ................................ ................................ ................................ ................................ ..... 60 TABLE 4 . 1 : DIFFERENCE IN LABOR COST ELASTICITY PER EMPLOYEE AND NUM BER OF EMPLOYEES ELASTICITY F OR SHALE BOOM EXPOSED BANKS ...... 61 TABLE 5 .1 : SHALE BOOM EXPOSED BANKS AND DYNAMIC CHANGES IN LABOR COST AND LABOR EMPLOYEE ELASTICITY ................................ ................................ .. 6 2 TABLE 6 .1 : SHALE BOOM EXPOSED BANKS, FORECASTING QUALITY, AND DYNAMIC CHANGES IN LABOR EMPLOYEE ELASTICITY ................................ .......... 63 TABLE 7 .1 : SHALE B OOM EXPOSED BANKS AND PRODUCT DIVERSITY ................... 64 TABLE 8 .1 : SHALE BOOM EXPOSED BANKS AND DYNAMIC CHANGES IN PRODUCT DIVERSITY ................................ ................................ ................................ .............................. 65 TABLE 9 .1 : SHALE BOOM EXPOSED BANKS, GEOGRAPHICAL MARKETS, AND DIFFERENCE IN COST AND LABOR EMPLOY EE ELASTICITY ................................ .... 66 TABLE 10 .1 : SHALE BOOM E XPOSED BANKS, MARKET SHARE, AND DIFFERENCE IN COST AND LABOR EMPLOYEE ELASTICITY ................................ ............................. 67 xi LIST OF FIGURES FIGURE 1: MAJOR U.S. SHALE PLAY REGIONS ................................ ................................ .. 6 9 FIGURE 2: TREATMENT AND CONTROL COUNTIES WITHIN THE SHALE PLAY STATES ................................ ................................ ................................ ................................ .... 70 FIGURE 3: TIME SERIES ANALYSIS FOR DIFFERENCE IN LABOR COST ELASTICITY ................................ ................................ ................................ ................................ ................... 7 1 FIGURE 4: TIME SERIES ANALYSIS FOR DIFFERENCE IN LABOR EMPLOYEE ELASTICITY ................................ ................................ ................................ ............................ 7 2 FIGURE 5: TIME SERIES ANALYSIS FOR DIFFERENCE IN PRODUC T DIVERSITY ..... 7 3 1 I NTRODUCTION Accounting rese archers and practitioners alike recognize the importance of cost structure decisions and acknowledge their effects on firm performance. Decisions related to cost structure s , i.e., the mix of variable and fixed costs, are influenced by a variety of factors, including demand uncertainty, capacity utilization, congestion costs, industry factors, regulatory pressures, and managerial incentives. 1 In the short run, cost structure decisions are affected by capacity planning that occurs before actual demand is real ized. During the capacity planning phase, firms pre - commit to the extent of fixed inputs they will hold , to satisfy the demand for their products or services (Ba nker, Byzalov, and Plehn - Dujowich [2014]) . During this phase, firms estimate the tradeoff betwe en excess capacity costs resulting from low demand realizations, and the opportunity costs of inadequate capacity such as lost sales or premium prices for inputs capacity and resour ce procurement choices that influence cost structure decisions (Banker, Byzalov, and Plehn - Dujowich [2014]; Holzhacker, Krishnan, and Mahlendorf [2015]b; Kallapu r and Eldenburg [2005]) (Brüggen, Krishnan, and Sedatole [2011]; Roychowdhury [2006]) , and how operating costs behave in response to changes in cost drivers (Anderson, Banker, and Janakiraman [2003]; Noreen and Soderstrom [1994]) . 2 While e xtant literature in accounting has examined resource adjustment decisions by managers in response t o a decrease in demand , resource adjustment decisions in response to an 1 Examples of studies include Balakrishnan, Petersen, and Soderstrom [2004]; Banker, Byzalov, and Plehn - Dujowich [2014]; Dierynck, Landsman, and Renders [2012] ; Hall [2016]; Holzhacker, Krishnan, and Mahlendorf [2015b]; Kallapur and Eldenburg [2005]. 2 Exten sive research in accounting has examined the drivers of asymmetric responses of cost changes to contemporaneous changes in cost drivers (see Banker and Byzalov [2014] and Banker et al. [201 8 ] for a review). 2 increase in demand have not been well - documented. 3 A few notable exception s are Banker, Byzalov, Ciftci, et al. [2014] and Chen, Kama, and Lehavy [2 019]) , which examine resource adjus tment decisions when managers are optimistic about a future increase in activity. Managerial optimism stems from an expected increase in activity during the ordinary course of business. Both studies explore resource adjus tments undertaken by managers when they can forecast future increases in activity with reasonable accuracy. Since managers are familiar with the operating environment, they are less concerned about uncertainty while making resource adjustment decisions. In this paper, I investigate how fi rms adjust their cost structures and operations in response to an unexpected change in their operating environment resulting from an exogenous has a favorable venue functions. A sudden, unexpected change in the operating environment causes difficulty in forecasting the magnitude and persistence of the positive shock especially when managers are unfamiliar with the shock (Bloom [2014]) . D uring the intermediate period, m anagers face uncertainty over the permanence of the shock . Uncertainty adds noise to the estimates of future cash flows and optimal resource adjustments required to meet the increase in demand in the current period . The dramatic chang e in the operating environment and the attendant requirement for adjustments during the current period before the persistent effect of the shock is observed can lead to errors in managerial judgments (Tversky and Kahneman [1974]) . Manag fferentiate whether the change in operating environment is persistent or transitory increases the option value of waiting to observe if the change reverts. This option to wait allows Bayesian updating or learning over time about the nat ure of change in the operating 3 Some of these studies include Anderson et al. [2003]; Chen, Lu, and Sougiannis [2012]; Dierynck, Landsman, and Renders [2012]; Kama and Weiss [2013]; Pinnuck and Lil lis [2007]; Weiss [2010] among others. 3 environment to resolve uncertainty about the permanence of such change (Grenadier and Malenko [2010]) . resource adjustments will systematically vary during the initial p eriods versus the la ter periods of the shock. During the initial periods when future uncertainty about the effects of the shock is high, man a gers will increase capacity utilization as permitted by the relevant range (e.g., increase the hours per employee ) or invest in resourc es with relatively low adjustment costs to meet the increase in demand . Temporary labor is an example of a resource with low adjustment costs because such employees are not on the payroll of the firm, are hired on a contract or fee basi s (such as per diem ) and can be hired and terminated as required. 4 During the later periods when future uncertainty about the effects of the shock is low , managers will be better able to estimate the magnitude of the shock and therefore undertake operation al expansions. These operational expansions include hiring of more permanent labor i.e., labor with high adjustment costs . These resource adjustments have cost structure implications that vary during the initial periods versus the later periods. During the initial periods, when managers are cautious about adding resources because of the uncertainty over the persistence of the shock , the increase in revenue will outpace the increase in cost, with the result that cost elasticity decreases. 5 Subsequently durin g the period s when managers have a better understanding of the economic environment and firms undertake operational expansions (such as hiring permanent labor), there will be an increase in responsiveness of cost changes to revenue cha nges resulting in an increase in cost elasticity. The resource adjustment decisions should result either 4 the supplying establishment but is under the direct or general supervision of the business to whom the help is furnishe 5 Cost elasticity is the responsive ness of cost changes to changes in activity levels. 4 structure reverting to the level that existed prior to the arrival of positive economic shock or to reach a different level in th e new steady state , which is difficult to predict ex ante . The nature of such resource adjustment decisions hinge s on uncertainty, and thereby on a I examine whether f irms with bett er forecasting qualit y are able resolve the uncertainty earlier and undertake operational expansions and increase labor employee elasticity before other firm s . Another operating variable that managers are likely to alter in response to a change in business environment is produ ct mix . Product mix decision is an instrument of risk management (Carlton and Dana [2008]). Product diversity , which refers to the range of product variations offered by a firm, is an important product mix decision. Product diversity r educes risk when firm s face uncertainty in their operating environment (Miller and Shamsie [1999]) . Accordingly, I examine how firms change product diversity in response to an unexpected positive economic shock that also results in uncertainty. I expect firms to reduce product diversity and focus on fewer products after a positive economic shock , particu larly during the later periods when the uncertainty in the operating environment has declined . The role of product diversity as a n instrument to reduce risk is less important wh en managers have a better understanding of the operating environment. While st udying the effect of an unexpected change in the operating environment on cost structure and product mix decisions, it is important to accurately identify what constitute s an 6 A limitation of industry - specific economic shocks i s that these shocks affect the entire cross - section of firms , making it difficult to recognize whether the shocks are exogenous or whether they are an outcome of firm characteris tics and actions and hence endogenous. I address this problem by examining a c hange in the operating environment 6 If the change in operating environment is anticipated, firms would have undertaken resource adjustments in the past muting the resource adjustments after the cha nge. 5 that results from positive economic shocks caused by the actions of firms outside the focal industry. My setting uses the liquidity windfalls experienced by banks from oil and natural gas shale development , which was a positive economic shock to the banking industry in the counties where shale development occurred (Gilje [2019]; Plosser [2014]) . Technological advancemen ts that resulted in shale developments were unlikely to be anti cipated or influenced by bank managers (Gilje, Loutskina, and Strahan [2016]; Reed et al. [2019]) . Moreover, the viability of these technological advancements was uncertain across different geo graphical areas. Therefore, it was difficult for bank managers to evaluate whether the shale developments would be transitory or persistent at the onset of the shale boom . Additionally, liquidity windfalls for the banks were predominantly localized, result ing in within - state variation in exposure of banks to the posit ive economic shock (Gilje et al. [2016]; Plosser [2014]) . Consequently, an advantage of my empirical setting is that it provides a quasi - natural experiment to investigate how firms make changes to their cost structure and product mix when they experience an unexpected change in their operating environment . This change in operating environment demands an estimation of required adjustments to long - term resource choices . In a frictionless capital m arket, an increase in bank deposits should not affe because interbank borrowings or external capital markets would find it profitable to fund all positive NPV projects (Gilje [2019]). However, the shale boom not only increased bank de posits but also mortgage lending (e.g., Gilje et al . [2016]). Thus, banks exposed to the shale boom encountered a spike in their activities, which required adjustments to their operations. There are several factors that contributed to bank managers being blindsided by the shale boom. These include the un expectedness of the boom, the genesis of the boom being outside the 6 banking industry, and the contribution of engineering and technology to the boom which was outside the expertise of bank managers. Inde ed, even industry experts could not predict or proj ect the impact of the shale boom (Lake et al. [2013]; Reed et al. [2019]) . The lack of understanding of the magnitude and persistence of the boom increase d uncertainty and affect ed resource adjustme nt decisions. High uncertainty regarding the nature of the shale boom increases the value of the option to wait and learn about the permanence of the boom ( Grenadier and Malenko [2010]) . 7 U ncertainty in operating environment diminishes over time as bank ma na gers become more confident that the shale boom is not transitory. Accordingly, managers make dynamic adjustments to operations. B ank managers have three major avenues through which they can meet the increase in demand. First, they can adjust resources b y increasing the hours per employee (utilize the fixed resources i.e., tap into fixed cost ) or hiring temporary employees during the initial years of the demand increase when both the magnitude and per sistence of the shale boom are uncertain . 8 Temporary em ployees are cheaper than permanent employees because they do not get the same level of fringe benefits or raises and promotions. Hiring temporary employees should reduce the responsiveness of labor co st changes to revenue changes because revenue increases would outpace the increase in such labor cost s . Second , they can expand capacity by hiring full - time employees during the later years of the shale boom when uncertainty about the persistence of boom has declined . Hiring full - time employees should increase the responsiveness of labor cost changes to revenue changes because the magnitude of labor cost changes would keep pace with 7 This option to learn should be most valuable immediately after the arrival of the shale boom because uncertainty about the persistence of the boom is highest at that point . 8 Firms hire part - time employees during periods of high uncertainty such as recession (Valletta and Bengali [2013]). Ho wever, in prior studies, it is unclear whether this hiring behavior is a response to a reduction in activity or to an increase in uncertainty. 7 revenue. In shor t, banks are likely to use resources with low adjustment costs during the initial periods of the shale boom and use resources that are costly to adjust during the later periods . Finally , they could reduce product diversity particularly during the later per iods of the shale boom because the need to use product mix scope to deal with demand risk is less of a concern when uncertainty regarding the persistence of shale boom has declined . Using a difference - in - differences estimation on 14,929 bank - year observations from 2005 to 2014 , I examine labor adjustments and product mix changes undertaken by banks in response to an u nexpected change in operating environment. I find that banks exposed to the shale boom reduce labor cost elasticity on average . I breakdown the labor cost into labor cost per employee and number of employees to further explore whether banks change the labo r mix during the shale boom. I find an increase in labor employee elasticity and a decrease in labor cost elasticity per employee for banks exposed to the shale boom. 9 Taken together, these results indicate that the number of employees become more sensitiv e to changes in interest income for the banks exposed to the shale boom , and the banks hire a higher proportion of low - skilled employees or increase the hours per employee. 10 Resource adjustments using less costly re sources reduce labor cost elasticity for the banks exposed to the shale boom. I also find that banks reduce product diversity after the arrival of the shale boom, indicating a reduced role of product mix diversity as a n instrument to reduce demand risk. If uncertainty is the driver of operating decisions, then labor adjustments and product mix changes should systematically vary over the period of the shale boom. The earlier periods of the 9 Labor c ost elasticity (labor employee elasticity) is the responsiv e ness of labor cost (number of employees) to changes in interest income . 10 It is also possible that banks hire a higher proportion of temporary labor after the arrival of the shale boom. Although it is not possible to deterministically conclude the exact nature of labor adjustment s undertaken by banks, all three avenues can be categorized as resources that can be easily adjusted in the short run . 8 shale boom are characterized by high uncertainty because of the ina bility of bank managers to understand whet her the change in operating environment is transitory or persistent. Accordingly, I examine the dynamic labor adjustments and product mix changes made by banks in response to the shale boom. Results suggest that ba nk managers adjust resources dynamically i n response to the shale boom , and their response depends on the uncertainty they face with respect to the persistence of the boom. I follow Bloom [2009] and measure uncertainty as the cross - sectional spread of bank - level earnings growth . Based on this meas ure, I observe that uncertainty is high during the first three years of the shale boom. During this period, banks exposed to the shale boom reduce labor cost elasticity , which is a result of increasing the labor wi th low adjustment costs. This reduced labo r cost elasticity eventually increases to the pre - shale boom level. This adjustment results from hiring full - time employees during the later periods of the shale boom when managers update the likelihood of the permanence of the shale boom and reduce their assessment of uncertainty. Th is evidence is consistent with the analytical predictions of Grenadier and Malenko [2010] in which Bayesian uncertainty over past shocks causes significant to positive cash flow shocks. I also find that banks with a better concurrent quality of allowance for loan and lease losses make permanent labor adjustments earlier than their peers . The internal information quality plays an important role in resolv ing th e uncertainty arising from an unexpected chang e to the operating environment. Taken together, these results provide evidence of dynamic adjustments to resources by banks in response to the shale boom. Further, relative to non - exposed banks, I find that banks exposed to the shale boom reduce product di versity during the later periods of the shale boom when they face relatively low uncertainty. Also, banks exposed to the shale boom do not reduce product diversity in the initial 9 periods of the s hale boom . Overall, the results indicate that bank managers a djust product mix only when they are confident that the effects of the shale boom are persistent . I contribute to the accounting literature in several ways. First, I identify a setting in which an exogenous positive economic shock at the local level resul ts in an unexpected change in operations requiring adjustments to resources . Prior studies have examined resource adjustment decisions when managers can forecast the increase in demand with reaso nable accuracy. Therefore, my study contributes to a better u nderstanding of how uncertainty drive s operational decisions of managers when there is a sudden and unexpected change in the business environment. Second, I contribute to the literature on cost m anagement by providing evidence that adjustments to cost stru cture are dynamic in nature i.e., manager ial resource choices are contingent on the operating environment. Third, I provide empirical evidence that managers prefer to wait even during positive ec onomic shocks to learn about the permanence of the shock, whi ch results in operational expansions that occur only during the later periods of the boom. Thus, managers not only delay hiring during periods of negative economic shocks but also during periods of unexpected positive economic shocks because of high uncert ainty. Finally, I contribute to the literature on the effects of uncertainty on manag erial decisions. While a robust literature has examined negative economic shocks ( e.g., Bloom [2009], [2014]; Bloom, Bond, and Van Reenen [2007]; Jurado, Ludvigson, and Ng [2015]) , positive economic shocks have been spar s ely studied. The effects of positiv e and negative shocks are unlikely to be symmetrical. When firms face uncertainty from negative economic shocks, it is difficult to disentangle the portion of the response that arises from a reduction in activity with that arising from the effect of uncert ainty. Further, it is unclear whether uncertainty is the cause or effect of a negative 10 economic shock (Bloom [2014]) a positive economic shock if there is uncertainty about the permanence of such shock . Section 2 reviews the relevant literature and develops the hypotheses. Section 3 describes the research design and empirical model specification. Section 4 describes th e data and reports the findings of the study, while Section 5 discusses several robus tness tests. Section 6 provides concluding comments. 11 RELATED LITERATURE AND HYPOTHESES Theoretical b ackground A cost structure in which a higher proportion of costs arise from committed resources with high adjustment costs exposes firms to higher operating risk (e.g., Chen, Kacperczyk, and Ortiz - Molina [2011]; Holzhacker et al. [2015] b and Rhee [1984]; Noreen, Brewer, and Garr ison [2014]) . W hen faced with demand uncertainty, firms with a higher proportion of fixed costs (i.e., a less flexible cost structure) are exposed to more variability in cash flows and earnings. Earnings volatility and the attendant unpredictability in earnings impose ch allenges to managers in planning for operations (Graham, Harvey, and Rajgopal [2005]) . 11 A higher proportion of fixed cost s increases the break - even poin t, and when such a cost structure is accompanied by demand uncertainty, the number of actual demand realizations at which the firm will incur a loss increases. In short, in the presence of demand uncertainty, revenues of firms with less flexible cost struc tures can fall below the break - even point if demand realization is lower than expectation. Therefore, demand uncertainty interacts with rigid cost structures to increase overall risk. Prior research in accounting analytically models the tradeoff between i nvesting in fixed inputs ex ante and procuring variable inputs ex post when actual demand is realized. The corresponding cost tradeoff is the cost of carrying committed resources versus bearing the higher resource price and co ngestion costs (Banker and Hug hes [1994]; Göx [2002]) . An unexpectedly high demand realization relative to available capacity results in congestion costs because demand has to be met using higher priced variable inputs (Banker, Byzalov, and Plehn - Dujowich [2014]; 11 Earnings volatility can also be an outcome of poor accounting estimates or accounting choices that reduce th e accura cy of the mapping b etween income and expenses (Dechow and Dichev [2002]; Dichev and Tang [2009]; Sloan [1996]) 12 Banker, Datar, and Kek re [1988]) . As the likelihood of high demand realizations increases, ex ante resource commitments become attractive because expected congestion costs exceed the expected cost of unused capacity from low demand realizations. In short, considerations of cong estion costs and higher resource prices may result in managers committing to ex ante fixed inputs and the attendant less elastic cost structures, even if such a choice yields lower earnings when demand realizations are lower t han expectations. Therefore, d emand uncertainty can result in a higher proportion of fixed costs in the cost structure (Banker, Byzalov, and Plehn - Dujowich [2014]) . A large body of accounting literature examines the drivers of asymmetric managerial resource adjustment decisions in res ponse to demand increases versus demand decreases ( e.g., Anderson et al. [2003]; Chen et al. [2012]; Hall [2016]; Kama and Weiss [2013]; Weiss [2010]) . These studies focus on how managers adjust resources in response to a reduction in demand . However, spar se research examines managerial resource adjustment decisions in response to an increase in demand. A f ew notable exceptions include Ban ker, Byzalov, Ciftci, et al. [2014] and Chen, Kama, and Lehavy [2019] , which examine the effect of consecutive demand in creases on managerial resource adjustment decisions. In both studies a future increase in demand can be forecasted with reasonable accur acy because the managerial optimism about an increase in demand stems from an expected increase in activity during the o rdinary course of business. Hence , managerial resource adjustment decisions in these studies are not responding to an unexpected element with respect to the uncertainty in the future operating environment . Therefore, managers prefer fixed cost structures t o capitalize on favorable demand realizations from variations that arise from the normal course of business . However i n my setting , when faced with a positive economic shock, managers find it difficult to forecast the magnitude and 13 persistence of the posit ive shock because they are unfamiliar with the shock (Bloom [2014]) . Firms that experience such shock s move from one s teady state before the arrival of a shock to another but face uncertainty during the intermediate period because managers are unsure wheth er the change in the operating environment is transitory or persistent. Consequently, resource adjustments in response to positive economic shocks are lik ely to differ from resource adjustments in response to demand increases during normal business activit ies because of the uncertainty over the permanence of the shock. Empirical research also shows that when faced with increased uncertainty in the operati ng environment, firms respond by adjusting their operations, with corresponding effects on their cost structures. For example, hospitals have increased the elasticity of their cost structures in response to the risk imposed by fixed price regulation (Holzh acker, Krishnan, and Mahlendorf [2015]a; Kallapur and Eldenburg [2005]) . 12 Managers are likely to adjus t their operations ex ante in anticipation of an ex post change in activity level arising from uncertainty. A change in activity level does not result in an automatic change to resources unless managers decide to adjust the resources (Cooper and Kaplan [19 92]; Holzhacker et al. [2015]b) . For example, in response to a change in the operating environment, airline companies have used outsourcing to reduce the ir fixed costs (Sedatole, Vrettos, and Widener [2012]) . Holzhacker et al. [2015]b find that hospitals u se outsourcing, equipment leasing, and temporary labor to increase cost elasticity in the presence of demand uncertainty and increased risk. These studies highlight that cost elasticity is a choice variable influenced by managers in r esponse to the uncerta inty in 12 Fixed price regulation increases the risk of loss because the reduced contribution margin caused by a reduction in selling price increa ses the quantity required t o break - even. To manage this risk, firms explore using variable inputs to meet production requirements. Thus, contribution margin volatility arising from a change in the reimbursement method for hospitals from cost plus to fixed price increases the value o f having a flexible cost function. 14 which require strategic cost management decisions with respect to the management of labor cost. Labor c ost m anagement Labor i s one of the most impo rtant resource s in service firms therefore managers adjust labor cost s in response to changes in activity. Extant literature finds evidence that firms make adjustments to labor capacity to meet demand (Caballero, Engel, and Haltiwange r [1997]; Hamermesh [1 989]) . Numerous factors such as managerial incentives, industry factors, and firm characteristics can increase the frequency of labor adjustments (Hall [2016]) . For example, firms that incur an accounting loss or firms that barely mee t the zero earnings be nchmark frequently reduce labor investments (Dierynck, Landsman, and Renders [2012]; Pinnuck and Lillis [2007]) . Therefore, firm characteristics play an important role in labor adjustments. Furthermore, labor adjustments are likely to be nonlinear on avera ge. For example, Ilut, Kehrig, and Schneider [2018] study the manufacturing industry and find a slow hiring response to a positive output shock. The response is more pronounced if the establishment employs more skilled workers or future output is subject t o higher uncertainty (Ono and Sullivan [2013]) . A slow hiring response adjust hours per employee or employ tem porary workers, which allows the firm to be responsive to such shocks. Firms tend to hire labor with low adjustment costs to meet an unexpected increase in demand for their products or when an economic environment is uncertain (Wander [2018] ) . 13 This hiring strategy allows firms to reduce labor adjustment costs if the uncertainty su bsides or when an increase in demand is only temporary. 14 Firms adjust hiring in response to permanent 13 One type of labor adjustment undertaken by firms is to hire temporary labor (Rothschild [2012]) , which is a part of labor - related business strategy for many firms. 14 Hiring of temporary emplo yees is particularly salient in industries that are susceptible to demand shifts (Dey, Houseman, and Polivka [2017]) . 15 shocks to output but avoid adjustments to labor in response to t emporary shocks (Guiso, Pistaferri, and Schivardi [2005]) . Temporary labor provides flexib ility to firms e specially during periods of reduced economic activity or high uncertainty and when firms are unable to decide whether a change in the operating enviro nment is transitory or persistent. Temporary labor allows firms to respond to demand chang es with lower adjustment costs that arise from recruitment and termination. Katz et al. [1999] find that firms reduce adjustment costs by hiring temporary employees d uring periods of changing labor demand. Shale boom and bank operations Technological adv ancements have enabled the extraction of natural gas shale through Until the end of the 20 th century, shale gas was not considered to be economically viable and contributed to less than 1 percent of U.S. natural gas production. However, the industry changed drastically in 2003 with the development of Barnett shale in Texas. The challenge in extracting natural gas from shale areas arises from the highly nonporous nature of the rock that traps the gas in the rock. Fracking breaks apart shale and allows the collection of natural gas. This technological breakthrough combined with th ree - dimensional seismic imaging reduced the cost of fracking (Wang and Krupnick [2015]) . Additionall y , higher natural gas prices meant that shale reserves became economically profitable to extract. Shale booms have been studied to examine t heir effect s o n local econom ies in general ( e.g., Bartik et al. [2019]) and banks in particular (Gilje [2019]; Gil je et al. [2016]; Plosser [2014]; Stuber [2019]; Wu [2017]) . 15 S hale discoveries have an immediate effect on supply as well as demand for bank credit e specially in counties exposed to the shale boom because of an 15 For example, shale regions experienced a significant increase in employment in the construction industry betwe en 2007 and 2011 (Eberhar t [2014]) . 16 overall increase in economic activities . Typ ically drilling companies negotiate leases and pay a large bonu s upfront to the landowners , unconditional on the , followed by a royalty as a function of gas production. 16 Land owners , such as those in the Haynesville region of North Louis iana, turned into millionaires overnight. Between 2007 and 2008, the lease rates in this region increased from $100 an acre to $10,000 to $30,000 per acre (Times - Picayune [2008]) . These landowners deposit the money in local banks resulting in liquidity win dfalls for the banks. Thus, a n increase in wealth windfalls for the county residents from the shale boom increases bank deposits and bank lending (Gilje [2019]; Gilje et al. [2016]; Plosser [2014]) . In short, a shale boom results in a spike in operations f or the banks exposed to the shale boom. 17 Banks are in turn required to make resource adjustments to accommodate the spike in operations. 18 The change in operating environment for banks exposed to shale boom was unexpected for the following reasons. First, drilling of shale wells is viable only if the demand and price of natural gas are within specific parameters. These parameters are in turn a fu nction of national and global macroeconomic forces and independent of local economic conditions (Lake et al. [201 3]) . Second, even though technological breakthroughs made fracking feasible, the viability was uncertain across different geographical areas. T he U.S. Energy Information Administration (EIA) doubled its estimate of total recoverable U.S. shale gas resource s between 2009 and 2011 , highlighting the unpredictable nature of shale discoveries (Lake et al. [2013]) . It was challenging even for industry experts to forecast the n umber of wells required within a shale play area to develop recoverable resources (Gilje et al. [2016]) . Therefore, it is unlikely that the banks in 16 Royalty payments for 1.8 million lease contracts from six major shale plays were approximately $39 billion in 2014 (Brown, Fitzgerald, and Weber [2016]) . 17 In Table 1.1, I replicate the result that the shale boom increased the deposits for t he banks expos ed to the shale boom. Additionally, I also examine whether there was an increase in interest income for the exposed banks to ensure that there was indeed a spike in my proxy measure for activity. 18 For example, First International Bank in No rth Dakota add ed 65 employees to its workforce of nearly 300 existing employees across 21 branches in 2012 in response to increased customer demand (Eberhart [2014]) . 17 shale areas strategically adjusted their resources and product mix in ant icipation of a change in operating environment. Further , banks not only experienced an unexpected change in their operatin g environment , but also faced uncertainty regarding the persistent effects of the shale boom e specially during the initial periods of the boom . Labor c ost m anagement in b anks in response to a positive economic shock Efficient cost management is particular ly important for commercial banks because the banking industry is highly competitive and face s regulatory constraints. Banks must maintain adequate capital as a proportion of assets adjusted for risk to ensure that they do not under take excess leverage. G o vernment agencies such as Federal Deposit Insurance Corporation (FDIC) and Federal Reserve Boa rd (FRB) monitor this capital adequacy requirement. Capital requirements incentivize banks to focus their efforts on cost management because low regulatory capita l make s it difficult for banks to add labor when required (Hall [2016]) . I particularly f ocus on labor cost management decisions of banks because labor is the largest cost item relative to non - interest expense . Further, managers have discretion with respec t to la bor cost. For example, in my sample labor cost is 54 percent of non - interest expense and 37 percent of total expense on average. 19 Additionally, labor is the most important capacity resource for the service industry . Banks have discretion over their labor mix decisions. For example, banks can increase the proportion of variable - pay employees or delay hiring fixed - pay employees when faced with uncertainty . Because labor costs are committed in advance, banks maintain slack labor resources to meet unexp ected demand. In the service industry, demand typically follows a Poisson distribution. As a result, capacity is not a hard number, but exhibits some flexibility to make adjustments. B anks maintain a mix of labor during the normal course of business and ad just t heir 19 The average ratio of labor cost as a percentage of non - interest expense for all banks from 1997 to 2018 is also approximately 54 percent. 18 labor cost in response to a change in activity to maintain this mix . However, when there is a change in operating environment the demand distribution also shifts warranting labor adjustments to accommodate such a change. During such periods, the labor mix can be expected to change at least temporarily , which in turn affec t s cost elasticity. Therefore, labor cost elasticit y is likely to change particularly when operations fall outside the relevant range of planned capacity (Balakrishnan, Petersen, and Soderstrom [2004]) . Additionally, when the change in operating environm ent is sudden and unexpected , as was the case with the shale boom , banks are required to change their labor adjustment strategies in response to a change in activity because of t he following two conditions . First , bank managers experiencing a spike in activities from the shale boom face uncertainty regarding the nature of the shale boom, given the complexity of macroeconomic factors that contribute to the boom . Second, there is a high level of uncertainty regarding future revenue generation p articularly during the initial periods of the shale development . That is, banks are unsure what a steady state will look like in the future . Although there is an increase in activity for the ba nks exposed to the shale boom, uncertainty around its persistence warrants a caution ary strategy with respect to labor adjustments (Grenadier and Malenko [2010]) . H iring of employees is a real option because if the spike in operations is temporary, the ban ks must incur the c ost of layoff s . It is also possible that the increase in operations exceeds the effect of uncertainty and signal s a need to hire employees in response to the spike in demand (Grenadier and Malenko [2010]) . Accordingly , I expect banks to make labor adjustme nts differently in response to a change in activity during the shale boom compared to the pre - shale boom period . During the initial periods of the boom, b anks can benefit from untapped economies of scale and temporarily adjust operations by increasing hour s per employee, hire temporary workers , or hire employees with low adjustment 19 costs which changes the labor mix . 20 In short, demand driven increases in output during the initial periods can be accommodated by increasing capacity utilizati on as permitted by the relevant range (Balakrishnan et al. [ 2004 ] ). Consequently, t he increase in revenue for banks exposed to the shale boom will outpace the increase in labor costs and r educe labor cost elasticity as stated in my first hypothesis: H1: R elative to banks operating in counties that did not experience the shale development , b anks operating in counties where the shale development occurred will reduce the elasticity of labor cost during the post shale development period. Uncertainty, d ynamic r esource a djustments , and r eal o ptions The decision about how much c apacity to build is one of the most fundamental operating decision s made by firms. When faced with uncertain demand, the capacity decision involves a tradeoff between excess capacity cost and the opportunity cost of lost sales or premium price to be paid for on demand inputs ordered on a flexible basis arising from high demand realizations r elative to planned capacity (Banker, Byzalov, and Plehn - Dujowich [2014]; Petruzzi and Dada [1999]) . F irms operating in less competitive markets can make capacity decisions disregarding demand uncertainty because market power allows these firms to utilize the capacity (Van Mieghem and Dada [1999]) . Firms with market power can use inventories to manage s ale s fluctuations by smoothing out production. When service industries face uncertainty , capacity choice decision becomes even more important because the output cannot be inventoried. Accordingly, s ervice firms operating in competitive industries will pref er operational flexibility under uncertainty , i.e., invest in resources with low adjustment costs because they lack market power and output cannot be inventoried. 20 Another possibility is that banks can outsource some of the activities such as loan processing (Hall [2016]) . 20 During times of reduced economic activity, firms layoff temporary workers before laying off per manent employees and delay the hiring of permanent employees during the recovery (Heinrich and Houseman [2019]) . Delays in hiring can occur from a mismatch between the supply of skills available in the labor market and the demand for skills by firms. Su ch delays can also result from heightened uncertainty during periods of recovery (Baker, Bloom, and Davis [2016]; Kocherlakota [2010]) . For example, aggregate employment growth for long - term labor was weak during the recovery period after the financial cri sis of 2008 - 2009 because firms were uncertain whether the recovery would sustain (International Monetary Fund [2012]) . During periods of small positive demand shocks, firms adjust resources by hiring labor ( e.g., Bloom [2009]) . However, resource adjustments in response to a substantive positive economic shock is not well - documented. Grenadier and Malenko [2010] model the investment behavior of firms in response to past economic shocks. They argue that when firms observe a positive shock t o their cash flows b ut are unable to identify its true properties, there is an option value to wait and learn whether the cash flow shock is temporary or persistent. This option to wait allows Bayesian updating of the likelihood that the change in the oper ating environment is temporary. In a nutshell , uncertainty about the permanence of economic shocks increases the option value of waiting to observe if the change reverts. The resource adjustments are likely to be dynamic in response to an unexpected positive shock. During t he initial periods of the shock, firms have positive demand expectations but also face uncertainty regarding the persistence of the shock . In the later p eriods, there is less uncertainty surrounding positive demand expectations because managers are familia r with the new operating environment and update their perception s about the permanence of the shock. In short, the resource adjustments can be expected t o systematically vary during the initial and later 21 periods of the unexpected positive economic shock. B ased on real options theory and the analytical predictions of Grenadier and Malenko [2010] , I expect that manager ial resource adjustments will systematically vary during the initial periods versus later periods of the shock. During the initial periods when future uncertainty about the effects of the shock is high, man a gers will prefer to invest in resources with relatively low adjustment cos ts . 21 Additionally, managers can utilize the slack available to meet the increase in demand as permitted by the relevan t range (Balakrishnan et al. 2004). During the later periods when uncertainty is lower or managers are better able to estimate the magnit ude of the shock, banks will undertake operational expansions. These operational expansions are likely to include hirin g of more permanent labor. Overall, banks are likely to either adjust hours per employee or hire workers with low adjustment costs , or both , and delay hiring workers with high adjustment costs during periods of higher uncertainty . These resource adjustmen ts have cost structure implications that vary during the initial periods versus the later periods. During the initial periods, when a l arger proportion of additional resources with low adjustment costs are deployed, the increase in revenue will outpace the increase in cost, with the result that cost elasticity decreases. Subsequently during the period s when managers have a better understa nding of the economic environment, the firm will undertake operational expansions (such as hiring permanent labor). Conse quently, there will be an increase in responsiveness of cost changes to revenue changes resulting in an increase in cost elasticity. Ac cordingly, I state the following hypothes i s: H2: Banks operating in counties where the shale development occurred decreas e (increase) the labor cost elasticity during the initial periods (later periods) of the post 21 Te mporary labor is an example of a resource with low adjustment costs because they are n ot on the payroll of the firm, are hired on a contract or fee basis (such as per diem) and can be hired and terminated as required. 22 shale development period relative to banks that did not experience the shale development . Labor c ost a djustments and b f orecasting a bilities In the presenc e of uncertainty, decisions related to labor procurement and resource adjustments are influenced by the forecasting ability of the bank. Banks with better forecasting ability are likely to have greater confidence in their ability to make labor employee ela sticity decisions. The quality of forecasting is a latent variable; ho wever, it can be proxied with other variable in the banking industry is the allowance for l oan and lease losses. This allowance is based on the estimate of the a mount of loans that are projected to incur losses. The loss period extends beyond the balance sheet date. If the bank expects the loss to occur in future periods, then a loss is deemed t o have accrued in the current period and accordingly a provision is ma de. 22 Allowance for loan and lease losses provide forward - looking information about the bank, particularly the expected credit losses ( e.g., Beatty and Liao [2011]; Harris, Khan, and Niss im [2018]) . Additionally, bank characteristics and incentives affect the quality of loan loss provisioning (Nichols, Wahlen, and Wieland [2009]) . Khan and Ozel [2016] find that estimated credit losses of banks aggregated to the state level provide informat ion about local conditions that is incremental to other leading indic ators of economic activity. Therefore, the quality of the loan loss provision not only but also their knowledge of local economic c onditions. 23 22 Regulatory guidance requires banks t o forecast the loan losses that are expected to occur over the next one year from the date of balance sheet and maintain an allowance for loan losses ( OCC) [1998]) . 23 Consistent with prior literature, I use the ratio of charge offs in the following year t o allowance for loan losses in the current year to evaluate the forecasting quality (B eatty, Liao, and Zhang [2019]; Cantrell, McInnis, and Yust [2014]; Stuber [2019]) . 23 greater ability to estimate the magnitude and persistence of the shale boom. Such a better information base can reduce the error in the estimate of the uncertainty regarding the persistence of the sh ale boom an d accordingly diminish the option value of waiting (Grenadier and Malenko [2010]) . Therefore, I expect banks with high forecasting quality, proxied by the quality of their allowance for loan and lease losses , to increase their labor employee ela sticity ear lier than other banks as stated in the following hypothesis: H 3 : Among the b anks exposed to the shale development, banks with high forecasting ability increase labor employee elasticity earlier than other banks. Product d iversity D ecisions rel ated to labor procurement and resource adjustments are likely to influence a Product diversity , which refers to the range of product variations offered by a firm, is an important product mix decision. A full range of product s al lows firms to cater to different market segments and obtain economies of scope (Tallman and Li [1996]) . Additionally, product diversity is a risk management tool, especially during periods of unexpected changes in the operating environment (Miller and Sham sie [1999]) . Carlton and Dana [2008] argue that p roduct diversity reduces the risk arising from demand uncertainty impact of uncertainty on its outcomes could result in the firm choosing to narrow its focus on mor e profitable product lines (Miller and Shamsie [1999]) . Banking is a multi - product industry. Most banks offer a variety of loan products to their customers such as home mortgage loans, car loans, edu cation loans, credit cards, and so on. There is often s ome degree of overlap of borrowers across products. When faced with a change in the operating environment, banks are likely to alter their product mix to increase their focus on 24 more profitable product s. Although reducing product diversity could reduce the ir economies of scope, it allows them to benefit from economies of scale. 24 However, the timing of when banks will make changes to the product mix is an empirical question. It is unlikely that banks wil l make changes to their product mix immediately after the shale boom, but likely that they will respond later when the uncertainty surrounding the persistence of the shale boom is lower. I predict that banks will reduce their product diversity in response to the shale boom, but this reduction will only occur during the later periods of the boom as stated in the following hypotheses: H 4a : Banks operating in counties where the shale development occurred reduce product diversity during the post shale developm ent period relative to banks that did not experience t he shale development. H 4b : There is a significant difference in product diversity between the initial and later periods of the post shale development for b anks operating in counties where the shale deve lopment occurred . 24 Bernard, Redding, and Schott [2010] use census data to provide evidence that manufacturing firms exposed to positive productivity shocks add products to their portfolio. I study how banks reallocate their loan portfolio, which is different from adding or dropping a product altogether. 25 RESEARCH DESIGN AND METHODOLOGY Banks e xposed to the s hale b oom To determine that a change was unexpected , there must be adequate reason s to assume that the growth in extraction of oil and natural gas in shale counties was unanticipated. As discussed above, t he unexpected nature of development in shale counties was a surprise even for the experts in the field (Gilje et al. [2016]; Lake et al. [2013]; Reed et al. [2019]) . Thus, when even oil and gas scienti sts were taken by surprise by the extent of production possibilities, it is unlikely that bank managers could have anticipated such a change and ex ante adjust ed their resources to accommodate the increase in future ope rations. T he exposed counties ( shal e counties ) are those within a play state with shale formation based on the classification of U.S. Energy Information Administration (EIA) . 25 The counties within a play state without shale formation are non - exposed or non - shale counties. According to EIA, there are eight major U.S. shale regions. 26 Figure 1 shows a U.S. map that highlights each shale region. Significant fracking activity began in the Permian region in 2005 , and my sample period begins in 2005, this region is therefore ex cluded from my sampl e. Additionally, I follow Stuber [2019] and exclude the state of New York and Texas from my analyses. 27 Th ese exclusions result in six U.S. shale regions , which I include i n my analyses. Figure 2 depicts the exposed and non - exposed counties in the shale pla y states and the year when 25 EIA similar geologic and ge ographic properties. The data on shale counties can be obtained from https://www.eia.gov/tools/faqs/faq.php?id=807&t=8 26 T he Appalachia region consists of Marcellus and Utica. I treat them as two separate basins for the purpose of my analyses because signi ficant fracking began in Marcellus region in 2007 and in Utica region in 2011. 27 The state of New York within the Marcellu s shale play is excluded because drilling activity was halted https://www.cnn.com/2010/US/12/13/new.york.fracking.moratorium/index.htm l. The state of Texas is excluded because conventional oil drilling before shale boom makes it difficult to determine the pre and post periods. Consequently, the entire Eagle Ford region and parts of Haynesville region covering the state of Texas are exclu ded from analyses . 26 significant fracking began within each shale region. As shown in Figure 2, I study the operating decisions of banks from 11 play states: ND, MT, PA, WV, OK, LA, AR, CO, WY, NE, and OH. I identify exposed banks as those with the majority of branches in shale counties. B anks that operate within the shale play state but do not have the majority of their branches in shale counties are identified as non - exposed banks. Exposed Bank is determined based on the number of branches of a ba nk in the first year when significant fracking activity began within the shale play state . I nformation for bank branch location s is obtained from Federal Deposit Insurance Corporation (FDIC) Summary of Deposits (SOD) data. I restrict the analyses to banks operating within the shale play states , which allows me to compare the operati ng decisions of exposed banks with non - exposed banks that operate within the geographical proximity. Research d esign Examining the effect of change in operating environment on resource adjustment and product mix decisions requires an event that was un anticipated by managers. If managers adjust operations in anticipation of a change in operating environment, the changes in resource adjustments and product mix will be measured wit h error because it is difficult to determine the post - event period accurate ly. 28 The unexpected nature of the shale development alleviate concerns of ex ante the shale boom and operational changes. However, local eco nomic trends can confound the analysis of resource adjustments and product mix . I address this concern by comparing the exposed banks with banks that have little or no exposure to the shale boom operating within the same play state using a difference - in - di fference s analys i s. Difference - in - difference s estimation addresses the concerns that confounding factors affect the exposed and non - exposed banks during the post - shale 28 For example, if the shale boom occurs in 2007 and managers anticipate this development and add resou rces in 2006, the appropriate post - treatment period to be considered is 2006. 27 development period or th at time trends unrelated to the shale development could affect t he exposed banks (Imbens and Wooldridge [2009]) . Hypothes is 1 - 3: Effect of change in operating environment on labor elasticity Consistent with prior literature, I infer resource adjustments from the observed cost elasticities (Banker, Byzalov, and Plehn - D ujowich [2014]; Hall [2016]) . Similar to Hall [2016] , I also use labor cost elasticity and labor employee elasticity to make infer ences about resource adjustments . L abor cost constitutes approximately 5 4 percent of total non - interest expense on average and is therefore the largest category of resource costs for banks . L abor cost elasticity is the slope coefficient obtained by regressing the log change in labor cost on log change in revenue (Banker, Byzalov, and Plehn - Dujowich [2014]; Dierynck et al. [2012]; Hall [2016]; Holzhacker et al. [2015]b) . I choose the log - changes model over log - levels for three reasons. First, a log - level model measures elasticity in the long run (Noreen and Soderstrom [1994]) and it is likely that banks maintain a certai n level of labor cost elasticity in the long - run and adjust their elasticities accordingly to this level over the period of the shale boom. Thus, there may be no difference in long - run elasticities in labor cost s before and after the shale boom. Second, wi th bank fix ed effects, the levels model removes the average interest income and labor cost across the entire sample whereas a changes model differences out the interest income and labor cost of a bank from the previous year . A changes model is particularly relevant b ecause earlier sample years (2005 to 2007) are likely to exhibit a different trend than later sample years (2008 to 2014) because of financial crisis (Gilje et. al [2016]) . T he log - changes model with bank fixed effects accounts for bank - specific trends all eviating this concern (Wooldridge [2010]) . Finally, I am interested in examining how banks ma k e dynamic labor adjustments during the initial and later 28 periods of the shale boom , which requires an estimation of short - run elasticity. 29 I use interest income a s a proxy for revenue , which is consistent with prior literature that uses sales as a n activity - driver in elasticity studies (Banker, Byzalov, and Plehn - Dujowich [2014]; Kallapur and Eldenburg [2005]) . 30 H1 examine s the effect of the shale development on labor cost elasticity for exposed banks relative to non - exposed banks using the following model: (1) w here . Labor cost and Employees : The dependent variable is the natural log of labor costs in the current y ear divided by labor costs of the prior year. Consistent with the literature on cost behavior, t he labor costs are deflated by the average consumer price index (Banker, Byzalov, and Plehn - Dujowich [2014]; Holz hacker et al. [2015]b) . I separate the labor cost into labor cost per employee and number of employees for a nuanced understanding of labor adjustments by banks in response to the shale boom . Accordingly, t o examine labor employee elasticity, the dependent variable is 29 In additional analyses, I examine long - run elasticiti es and discuss them in Section 5. 30 An alternative cost driver to interest income is total revenue ( Hall [2016] ). However, Hall [2016] uses data at the bank holding company level, where a significant portion of revenue is non - interest revenue. My analysis uses bank - level data and interest i ncome constitutes a major portion of total revenue (85% on average) for the banks in my sample. Additionally, the economic magnitude of labor cost elasticity when non - interest income is used as an activity measure is clo se to 0 (untabulated). In additional tests (untabulated), I use the sum of interest and non - interest revenue as a cost driver and the results are qualitatively similar. 29 defined as the natural log of number of full - time equivalent employees in the current year divided by number of full - time equivalent employees in the previous year. Similarly, to examine labor cost elasticity per employee, the dependent variable is defined as the natural log of labor cost per full - time equivalent employee of current year divided by labor cost per full - time eq uivalent employee of the previous year. The labor cos t per full - time equivalent employee is also deflated by the average consumer price index. Revenue : I use interest income as the activity measure. is the natural log of interest income in the current year div ided by interest income in the prior year . Exposed Bank : is an i ndicator variable equal to 1 if majority of branches of bank are in counties identified as exposed to th e shale boom within a shale play state , and 0 otherwise . The number of branches in the year in which significant fracking activity began in each of the play state s is used to determine this variable . Shale development period : The shale development period is determined as the period from the year when signific ant fracking activity began in the shale play state (Bartik et al. [2019]; Gilje [2019]; Stuber [2019]) . is an indicator variable equal to 1 if significant fracking activi ty has commenced in a state where bank has branch locations , and 0 in prior years . Asset Intensity and Employee Intensity : I follow prior literature and include which is defined as the ratio of nonfinancial assets (property, plant, and equipment [PP&E]) to interest income and defined as the ratio of number of employees to interest income multiplie d by 1,000 . These variables proxy for capacity adjustment costs (Hall [2016]; Holzhacker et al. [2015]b) . Federal Funds Rate : The i nterest income of a bank can change due to changes in interest rate. Therefore, I follow (Hall [2016]) and include me asured as the percentage change in 30 the average federal funds rate multiplied by the total dollar amount of loans at the beginning of the year, where the average rate is calculated using the federal funds rate in January plus the federal fund s rate in December of year t divided by 2 . This variable also controls for the effects of macroecon ( e.g., Holzhacker et al. [2015]a) . Size : Bank size can affect the labor cost elasticities because of economies of scale (Balakrishnan, Labro, and Soderstrom [2014]) and accordingly I include defined as log of total assets of bank i at the beginning of the year . Capacity utilization : A b (Balakrishnan et al. [2004]; Cannon [2014]) and accordingly I include defined as l og change in total deposits of bank i from year t - 1 to year t (Banker, Byzalov, and Plehn - Dujowich [2014]) . I include year fixed effect s and . The year indicat ors are relative to the year of the shale development in a shale play state . Consequently, the main effect of is absorbed by year fixed effects. Similarly, the main effect of is absorbed by bank fixed effects. All the continuous variables are demeaned in the interactions so that the main effects can be interpreted as the slope at the mean levels of all variables. To study the effect of the , I adopt a difference - in - differences approach. The firs t difference measures the difference in labor cost elasticity for exposed and non - exposed banks prior to the shale boom . is the coefficient of primary interest , which measures the difference - in - difference s estimate of labor cost elasticity fo r the exposed banks during the post shale development period vis - à - vis non - exposed banks in the entire period and exposed banks in the pre shale development period. In short, captures the ch ange in labor cost elasticity from a change in operating en vironment resulting from the shale boom. A negative 31 coefficient on would indicate that labor cost elasticity decreased in the post shale development period for exposed banks. For a better und erstanding of labor adjustments undertaken by banks in r esponse to the shale boom, I replace the dependent variable with and and estimate equation (1). A negative coefficient on when the dependent variable is would indicate that labor cost elasticity per employee decreased in the post shale d evelopment period for exposed banks . To test H2, I replace the ind icator variable in equation (1) with three separate indicator variables , which takes the value 1 in the first two years of the post - shale development and 0 otherwise, , which takes the value 1 in the third and fourth year of the post - shale development and 0 otherwise, , which takes the value 1 for all years after four years of the post - shale development and 0 otherwise. Breaking the post shale development period allows me to examine the labor elasticity changes by exp osed banks during the initial and later periods of the shale boom. T o test H3 i.e., whether banks with better forecasting ability adjust labor employee elasticity earlier than other banks , I split the full sample used to test H2 into banks with high fore casting ability and low forecasting ability . B ank - years with high forecasting ability are defined as 1 if the ratio of charge offs in year t+1 to allowance in year t for a bank falls in the highest quartile. 31 31 A ratio of charge offs to allowance for loan and lease losses greater than 1 indicate s an inadequate reserve. Since the i deal measure for this ratio is 1, I treat bank - years with ratio greater than 1.68 to be of low forecasting quality. The results are consistent if all bank - years with ratio greater than 1 are excluded from the high foreca sting ability subsample and included in the low forecasting ability subsample (untabulated). 32 Hypothesis 4: Effect of change in operating env ironment on product diversity To test H4a and H4b, I estimate the effect of the shale development on product diversity using the following model: (2) Product diversity : is mea sured as the natural log of the sum of the squares of consumer loans, real estate loans, and commercial and industrial loans as a proportion of gross loans for bank i in year t . This measure is multiplied by - 1 so that the index is positive, and a higher n umber reflects a more diversifie d portfolio of a bank. For example, if a bank only provides real estate loans, its product diversity measure will be 0. Large Bank : Large banks have a greater ability to adjust the product mix in response to change in opera ting environment by expanding in to otherwise unexplored business or geographical segments. Therefore, I include an indicator variable beginning of the year in which significant fracking activity bega n in a shale play state is greater than 500 million USD and 0 otherwise (Gilje [2019]) . The variables in equation (2) , which are also a part of equation (1) are defined above and the discussion related to the main effect of a nd is also relevant here. The main variable of interest is the interaction between and . The coefficient - shale developm ent period. A negative coefficient on indicates that banks exposed to the shale boom 33 reduced product diversity. The coefficient on measures whether large banks responded to the shal e development and adjusted product diversity differently than small banks . The main effect of and will be absorbed by bank fixed effects. To test the dynamic effect of change in operating environment on product diversity (H4b) , I apply the procedure mentioned above to test H2 and replace with three indicator variables and interact them with to examine the difference in product diversit y changes during the initial and later periods of the shale development. 34 DATA AND RESULTS Sample Table 1 .2 provides the sample selection details. I begin with the year - end Consolidated Report of Condition and Income ( Call Reports ) filed by all banks in U .S. during 200 4 - 201 5 . The lagged variables for 2005 required in the analyses are computed using 2004 data a nd the charge offs required to construct the forecasting quality variable for 2014 are computed using 2015 data. The sample period begins in 2005 to avoid a long time - series because resource adjustments are likely to be affected by the structural changes in banking industry in the long - run . I discard banks in the top 1 percent of assets and banks outside the play state because these banks are not comp arable to the banks in the sample . Consistent with prior literature, I discard bank - yea r observations if salaries exceed interest income in the current or prior year or if salaries and interest income are in the top and bottom 1 percentile to ensure that t he estimates are not affected by outliers ( e.g., Banker et al. [2014]) . The final sampl e includes 14,929 bank - year observations. Descriptive statistics Table 2 .1 presents the descriptive statistics for the sample of bank - years used in my analyses. All co ntinuous variables except labor cost and interest income are winsorized at 1 and 99 percentiles. The mean labor cost is 4 million dollars, and an average bank employs approximately 68 full - time equivalent employees. The median number of employees is 36 , wh ich indicates that this number is right skewed due to the existence of few large banks in the sample. The mean growth in labor cost (employees) is 3 (2) percent . 32 T he median change in employees is 0 which is consistent with the evidence that firms d o not a djust labor regularly but often adjust labor cost. The mean growth in interest income (deposits) is 2.8 (5.9) percent . T he standard 32 The log changes in labor cost approximates a percentage change in labor cost. It measures a continuously compounded growth in labor cost. 35 deviation for growth in interest income is larger than the standard deviation for growth in labor cost. The distribution for product diversity is positively skewed , which indicates that some banks have a highly diversified loan portfolio. Table 2 .2 prese nts the correlation s among the variables used in my analyses. As expect ed , the growth in labor cost is highly correlated wit h the growth in interest income . O nly a moderate correlation can be observed between the growth in labor cost and growth in employees. A high positive correlation exists among the number of branches , log assets, and number of employees . A negative and sign ificant correlation exists between product diversity and log assets , which suggests that large banks in my sample have a relatively less diversified loan portfolio . Table 2 .3 shows the univariate analyses for exposed and non - exposed banks before and afte r the shale boom. The difference i n average labor cost scaled by number of branch es is positive and significant at the 5 percent level only during the later periods of the shale development which suggests that banks exposed to the shale boom hired costly l abor resources in the later periods of the shale boom. Similarly, the difference in interest income scaled by number of employees is positive and significant during the post shale development period. The difference in log deposits is positive and significa nt for exposed banks during the pre and post shale development period. Hypothesis 1 - The results of estimating labor cost elasticity changes in response to the shale boom are tabula ted in Table 3 .1 . In all the columns, year fixed effects are included, which are relative to the onset of boom and absorb the main effect of . Standard errors are clustered by bank to account for serial correlation. Columns (1) to (4) estimate t he labor cost elasticity changes without the control variables but includes different types of fixed effects. Column (2) includes 36 state fixed effects to account for time invariant state factors that can affect the labor cost el asticity of banks in the samp le. Column (3) includes county fixed effects to account for time invariant county factors that can affect the labor cost elasticity of the banks in the sample. Column (4) includes bank fixed effects to account for bank specific factors that do not change o ver time. Therefore, in Column (4) the main effect of is absorbed by bank fixed effects. The labor cost elasticities in Columns (1) to (4) are consistent and show a statistically significant decrease i n labor cost elasticity fo r the banks exposed to the shale boom relative to non - exposed banks providing support for H1. In the remainder of th is paper, I perform all my analyses by including bank fixed effects . Table 3 .1 , Column (5) provides the results of e stimating equation (1). Th e labor cost elasticity for the exposed as well as the non - exposed banks in the pre - shale development period is 0.681 (p - value<0.01). The labor cost elasticity for exposed banks in the post - shale development period falls to 0.501 whereas the labor cost ela sticity for non - exposed banks do es not change in the post - shale development period. The a change in interest income is lower after the shale boom. The revenue increases for th e banks exposed to the sha le boom outpace d the increase in labor cost with the attendant effect that labor cost elasticity reduced in the post - shale period . A lthough the exposed banks obtained an increase in revenues, they did not make corresponding changes to labor cost an d instea d benefited from economies of scale and slack resources that they had at their disposal . Thus, H1 is supported i.e., banks exposed to the shale boom reduce labor cost elasticity in the post - shale development pe riod. To examine whether exposed banks change d the composition of their labor resources after the shale boom, in Table 4 .1 , I analyze labor cost elasticity per employee and labor employee elasticity . In all the columns, year indicators relative to the onset of the boom absorb 37 the main effect of , and standard errors are clustered by bank. Columns (1) and (2) estimate the labor cost elasticity per employee whereas Columns (3) and (4) estimat e the labor employee elasticity. Columns (1) and (3) are estimated with county fixed effects wh ile Columns (2) and (4) are estimated with bank fixed effects. The negative coefficient on in Columns (1) and (2) show that labor cost elasticity per employee decreased f or the banks exposed to the shale boom. E xposed banks did not match the changes to their labor cost on a per - employee basis relative to the increase in their revenues from the boom. Thus, the total labor budget on a per - employee basis did not increase to c orrespond to their revenue increase. However, exposed banks increased the number of employees as can be observed from the positive coefficient on as seen in Columns (3) and (4 ) . The shale boom resulted in windfall gains for the banks and required labor to service the increase in operations . However, the labor that banks hired to account for the increase in operations was of lower cost, which maintained the total labor cost budge t to be below the corresponding increase in revenues (Table 4 .1 , negative coefficient . Thus , banks were cautious about resource adjustments in the boom period and did not increase the ir labor budget to even budget neutral levels , which would have maintained the level of labor employee elasticity and labor cost per employee at pre - boom levels . Taken together, the results of Table 3 .1 and 4 .1 suggest that exposed banks hired less costly labor during the post - shale development period. Thus, in response to a change in interest income, banks made adjust ments to labor with lo w average cost . Consequently, the elasticity of the number of employees increased but the total labor cost elasticity and the labor cost elasticity per employee decreased for banks exposed to the shale boom. 38 An alternative explanation for the reduction in cost elasticity observed in Table 3 .1 could be that banks exposed to the shale boom prefer rigid cost structures in the post - shale development period . A rigid cost structure , i.e., a reduction in cost elasticity , helps firms to earn higher profits from favo rable demand realizations. To gain better insights for the results from Table 3 .1 and 4 .1 and rule out this potential alternative explanation , I next examine dynamic labor adjustments undertaken by exposed banks during the post - shale developme nt period . If ban ks exposed to the shale boom prefer rigid cost structures to capitalize on favorable demand outcomes, I should observe an increase in labor cost elasticity during the initial periods when banks add resources followed by a reduction in elas ticity during the later periods when firms utilize the added resources. On the other hand, if managers were cautious during the initial periods of the shale boom, th e labor cost elasticity should decrease during the initial periods and increase during the later periods whe n the banks face less uncertainty in the operating environment. Dynamic labor adjustments help to examine which of the abovementioned two explanations is appropriate. Table 5 .1 presents the results for dynamic changes in labor cost and labor employee elastic ity during the initial and later periods of the shale boom. I estimate the elasticities using equation (1) but replace the indicator variable with 3 indicator variables to examine dynamic changes in elasticities. Column s (1) and (2) report the labor cost elasticities for exposed banks and non - exposed banks respectively while column (3) reports the difference. The results show that labor cost elasticities fall significantly (p<0.05) during t he initial periods of the shale boom for exposed banks , as observed earlier in Table 3 .1 , whereas the difference in labor cost elasticity is insignificant for non - exposed banks. T herefore, exposed banks did not increase the ir labor cost in proportion to their increase in revenues . However, the difference in labor cost elasticity for exposed and non - exposed banks during the later periods of the shale 39 boom is not significant. Additio nally, the labor cost elasticity for the exposed banks is not significantly different from its labor cost elasticity during the pre - shale boom period. In short, the ex eventually reverts to the pre - shale boom level. Column s (4) and (5) report the labor employee elasticity for exposed banks and non - exposed banks respectively while column (6) reports the difference. The results show that the difference s in labor employee elasticities are not significant during the initial per iods of the shale boom and only become positive and significant during the later periods of the shale boom. Th e results suggest that exposed banks made adjustments to employee count in response to an increase in interest income only during the later period s of the shale boom. Further, t he labor cost elasticity for exposed banks shows an increasing trend during the post - sha le boom period , which also coincides with an increasing trend in labor employee elasticity for the exposed banks . Overall , the results in dicate that exposed banks added more employees during the later periods of the shale boom , which also made labor cost more responsive to change in interest income. I next examine the reasons for the difference s in elasticities during the initial and later periods of the shale boom. As mentioned earlier, a reduction in labor cost elasticity during the initial periods of the shale boom is likely caused by an increase in interest income outpacing the increase in labor cost. A possible explanation is that bank s increased the hours per employee or hired temporary employees rather than hiring more expensive full - time employees to meet the increase in operations. This choice of resource adjustment could be driven by managers inability to determine whether the sha le boom is transitory or persistent. To examine whether this is a possible explanation, I follow Bloom [2009] and compute the cross - sectional standard deviation s 40 a measure of uncertainty . 33 I find that the first three years after the shale boom are represen ted in the top quartile of the uncertainty measure. 34 Therefore , there is suggestive evidence that exposed banks increased labor adjustments during the post - shal e boom period but only in the later periods when they were confident that the boom was persistent. Overall, results are consistent with the theoretical predictions of Grenadier and Malenko [2010] that uncertainty with respect to determining whethe r the pas t economic shock is transitory or persistent results in a delay in investment in response to a positive cash flow shock. Hence, I find support for H2. Next, I examine the effect of the forecasting ability on its labor adjustments. Banks wit h greater forecasting ability are in a better position to gauge whether the shale boom is persistent or transitory and make speedier responses . I estimate equation (1) by including three indicators which were included to test H2 and split the sample into banks with high forecasting ability and low forecasting ability. I use the ratio of charge offs in year t+1 asting ability. Table 6 .1 presents the labor employee ela sticit ies for banks with high and low forecasting ability. Columns (1) to (3) provide the labor employee elasticity for banks with high forecasting ability and Columns (4) to (6) provide the labor empl oyee elasticity for banks with low forecasting ability. Column (1) indicate s that labor adjustments for exposed banks in response to change in interest income a re significantly higher during years 3 and 4 on average after the shale boom than the pre - shale boom period. Additionally, Column (4) shows that labor adjustments for exposed banks in response to change in interest income are significantly higher only afte r 4 years after the shale boom than the pre - shale boom period. The results show that banks with high forecasting 33 Pretax profit growt h scaled by average interest income is winsorized at the 1% level. 34 I recompute this measure by excluding years 2009 and 2010 to ensure that the measure does not capture the uncertainty around the financial crisis. Results continue to indicate that uncert ainty was high during the first three years of the shale boom. 41 ability can better estimate the effect and magnitude of the shale boom and undertake labor adjustments during the post - shale development period earlier than banks with low forecasting ability. Overall, the results provide support for H3. One concern with the above results is that banks could have faced competition from other industries or skilled employees were difficult to find in the short run . Therefore, delay in labor adjustments could be unrelated with uncertainty around the persisten ce of shale boom. In additional analyses, I use the Bureau of Labor Statistics (BLS) data on labor force and unemployed at the county - year level and aggregate t hem for all counties in which a bank has branches . I control for the unemployment rate computed using the aggregated labor force and unemployed in my analyses. T he results are qualitatively similar and close to the estimates in Tables 3 to 6. Additionally, if labor supply was driving the results, then banks with better forecasting ability would not b e able to undertake labor adjustments earlier than peer banks. Overall, the results indicate that labor adjustments made by bank managers are influenced by unce rtainty in the operating environment. Hypothesis 4: Effect of change in operating environment on In this section, I examine the product diversity changes. I estimate equation (2) and tabulate the results in Table 7 .1 . The models include bank and year fixed effects and cluster standard errors by bank. Columns (1) and (2) show t hat there is a significant reduction in product diversity for the exposed banks relative to non - exposed banks. However, the coefficient on is not significant in Column (2) which means that there is no difference in product diversity changes between small and large banks exposed to shale boom. Overall, the results provide supp ort for H4a and suggests that banks became less 42 concerned about the downside risk and hence concentr ate their focus on fewer loan products after the shale boom. I examine at what period exposed banks reduce product diversity . If the banks were less conce rned about downside risk, then the product diversity reduction should be higher during the lat er periods of the shale boom when uncertainty is relatively lower. I adopt a similar approach followed to test H2 i.e., I replace the indicator variable with three indicator variables and estimate equation (2 ) with bank and year fixed effects , a nd cluster standard errors by bank. The results tabulated in Table 8 .1 show that there is no significant difference in product diversity between exposed and non - exposed banks during the initial periods of the shale boom. However, the difference in product di versity is negative and significant (p<0.01) during the later periods of the shale boom. Additionally, I find that product diversity for the exposed b anks during the later periods of the shale boom is lower than the product diversity during the initial two years of the shale boom . Thus, exposed banks reduce product diversity during the later periods of the shale boom when bank managers are more confiden t that the effects of the shale boom are likely to be persistent. 43 ROBUSTNESS TESTS Parallel tre nds assumption A necessary condition in difference - in - differences estimation is that in the absence of treatment, the difference between the treatment and contr ol group is constant over time. This assumption implies that in the absence of the shale boom, the difference in labor cost , labor employee elasticity , and product diversity between exposed and non - exposed banks is constant over time (Cerulli and Ventura [ 2019]) . This assumption cannot be directly tested and therefore I adopt multiple approach es to mitigate the concerns that the results are not confounded by differen ce in trends. First, estimating equation (1) and (2) by including bank fixed effects prevent s any time invariant factors from affecting the labor cost , labor employee elasticity , or product diversity . Second , I graph the difference in labor cost elasticity , labor employee elasticity, and product diversity for exposed and non - exposed banks. Figure 3 shows that there is no significant difference in labor cost elasticity for exposed and non - exposed banks during the pre - shale boom period except for one year when the difference is marginally significant at the 5% level . The combined effect of all years in the pre - shale boom period is insignificant. Similarly, figures 4 and 5 reveal that there is no significant difference in labor employee elasticity or product diversity for exposed and non - exposed banks during the pre - sh ale boom period. Third , I estimat e equation (1) by replacing the indicator variable with year indicators and test whether the differences in labor cost elasticity, labor employee elasticity , and product diversity for exposed and non - exposed bank s during the pre - shale boom pe riod are jointly significant (Granger [1969]) . The results (untabulated) indicate that differences in labor cost , labor employee elasticity , and product diversity are not jointly significant ( for labor cost elasticity; for labor employee 44 elasticity; and 35 Therefore, there is no indication of difference s in trends for the variables of interest between the exposed and non - exposed banks prior to the shale bo om . Comparing short run and lo ng run labor elasticity The baseline result s for labor cost elasticity and labor employee elasticity in Table 3 .1 , Column 5 and Table 4 .1 , Column 4 show that labor cost elasticity is significantly higher compared to labor employee elasticity (0.681 in the pre - shale boom period for labor cost and 0.204 for employees) , which is consistent with the results in prior literature ( e.g., Banker et al. [2014]) . 36 The labor cost and labor employee elasticity imply that labor is a quasi - fixed resource in the banking indus try in the short run. I examine the long - run elasticity using a log - level model and estimate equation (1) (Noreen and Sode rstrom [1994]) . I find that long - run elasticity for labor cost s or employees are almost identical (0.83 for labor cost and 0.86 for em ployees ) (Untabulated ). Further, the results show that labor cost elasticity does not decrease for the exposed banks relative to non - exposed banks after the shale development, but labor employee elasticity does increase. These results are consistent with t he overall hypothesis that in the post - shale development period, the exposed banks hire d more employees to accommodate the increase in operations. Increasing the hours per employee and hiring temporary labor to adjust for an increase in operations is not s ustainable and likely to be reversed in the long run . T herefore , it is important to study the short - run elasticity to gain insights on resource adjustments of firms from one steady state before the change in operating environment to another steady state af ter such a change. 35 The results are similar if the analysis is restricted to only three years before the shale boom. 36 The difference between labor cost elasticity and labor employee elasticity in my analyses i s larger compared to the results in Banker et al. [2014] . This differ ence is possibly driven by large variation in the skill of employees in a banking industry compared to a manufacturing industry. 45 Ad ditional tests The exposed and non - exposed banks were identified based on the location of majority of branches in a shale county. It is possible that some exposed banks are not significantly affected by the shale boom whereas some non - exposed banks are significantly affected by the shale boom. To address this concern, I undertake two additional analyses. First , I estimate equations (1) and (2) by replacing the exposed bank variable with a continuous measure based on the proportion of t otal bank branches that are in counties exposed to the shale boom. The exposure variable is 0 for all banks before the onset of shale boom and takes the v alue between 0 to 1 depending on the proportion of branches in shale counties. The advantage of using this measure is that shale boom. The results are qualitatively s imilar for all dependent variables (untabulated). Second, I estimate equation (1) by splitting the sam ple and analyze labor adjustments of exposed banks in single county and multiple counties. I include bank and year fixed effects and cluster the standard errors by bank. The results for labor cost and labor employee elasticity for banks operating in multip le counties and single counties are shown in Table 9 .1 . Columns (1) and (2) show the labor cost elasticity for banks operating in multiple counties and a si ngle county respectively whereas Columns (3) and (4) show the labor employee elasticity for banks oper ating in multiple counties and a single county respectively. The results indicate that single county banks exposed to the shale boom reduced labor cost el asticity but did not change labor employee elasticity compared to single county banks not exposed to t he shale boom. 37 A possible explanation is that single county banks can adjust labor by increasing the hours per employee to 37 In the analyses related to single county banks one could argue that standard errors should be clustered at the county level ( Abadie et al. [2017]) . Accordingly, I also estimate equation (1) for the sub - sample of single county banks in Table 9 .1 and 10 .1 by clustering the standard errors at the county level and find consistent results (untabulated). 46 accommodate the spike in operations. On the other hand, exposed banks operating in multiple counties do not have a reduction in labo r cost elastici ty relative to non - exposed banks but have an increase in labor employee elasticity. Exposed banks operating in multiple counties are likely to hir e more employees , which makes the labor employee elasticity more responsive to interest changes in the post - sh ale boom period . As a result, t heir labor cost elasticity keeps pace with the increase in interest income and does not show a decline in the post - shale boom period. I further analyze banks operating in a single county by splitting them in to banks wi th high market share and banks with low market share. I determine high market share based on a median split of proportion of deposits of bank in the county in which the bank is operating. The results tabulated in Table 10 .1 indicate that exposed ban ks with l ow market share reduce labor cost elasticity relative to non - exposed banks whereas banks with high market share increase labor employee elasticity relative to non - exposed banks. Taken together, the results suggest that exposed banks with high mark et share adjust labor to accommodate the increase in revenues from the shale boom. On the other hand, exposed banks located in a single county with a low market share do not adjust labor in response to the shale boom. Thus, these banks are likely adjusting hours pe r employee or hiring temporary labor to accommodate the increase in operations . 47 CONCLUSION In this study, I examine the effect of a n unexpected positive change in operating th e shale boom as a adjustment and product mix decisions in response to an unexpected change in operating environment that increases revenues . D ecisions surrounding a n unexpected economic boom are challenging because of the difficulty in forecasting whether the change in operating environment is transitory or persistent. Conversely , m anagers can forecast the demand with reasonable accuracy during the ordinary course of b usiness. Therefore, unfamiliarity with the operating environment is likely to cause managerial responses to differ from their responses to sales increases during normal business cycle s . I find that relative to the pre - shale boom period, banks exposed t o the shale boom adjust employee cost. However, these adjustments occur during the later periods of the boom when there is less uncertainty about the persistence of the boom . In the initial periods, banks make adjus tments to hours per employee or hire temp orary labor . As a result, the increase in interest income outpaces the increase in labor cost , which reduc es labor cost elasticity but does not impact labor employee elasticity. I further examine whether banks with superior forecasting ability are able to resolve the uncertainty around the persistence of the shale boom and undertake labor adjustment decisions earlier than other banks. The results provide evidence for earlier adjustment by banks with high forecasting ability. I also examine product mix chang es by exposed banks in response to the shale boom and find that banks reduce product diversity. A reduced downside risk allows banks to focus on fewer products and maintain their profitability . The results show that exposed banks reduce product diversity, and this reduction happens during 48 the later periods of the shale boom when bank managers are likely to be more confident that the change in operating environment is persistent. Overall, resource adjustment and product mix de cisions are dynamic in nature co ntingent on the operating environment. I contribute by showing that a change in operating environment has a differential response during the initial and later periods of such change. In the initial periods, firms are likely to adjust resources with low a djustment costs , whereas in the long - term firms alter resources with high er adjustment costs. These responses are driven by uncertainty arisin g from difficulty to assess whether the change is transitory or persistent . The internal information environment a lso plays an important role in an early resolution of uncertainty and enables managers to stabilize their cost structures earlier rather than later . 49 APPENDICES 50 APPENDIX A Variable Definitions 51 Variable Definitions Dependent Variables Log - change in labor cost deflated by average consumer price index of bank from year to year Log - change in labor cost per full - time equivalent employee deflated by a verage consumer price index of bank from year to year Log - change in number of full - time equivalent employees of bank from year to year Logarithm of sum of the squares of consumer loans, real estate loans, and commercial and industrial loans as a proportion of gross loans for bank in year . This measure is multiplied by - 1 to convert the index into a positive number. Explanatory Variables Log - change in the total interest income of bank from year to year Number of branches of bank as on June 30 of year (SOD) Indicator variable equal to 1 for bank with majority of branches in counties identified as exposed to the shale boom within a play state and 0 otherwise. This variable is computed based on the number of branches in the first year when significant fracking activity began i n each of the play sta te. Branch location is identified from FDIC Summary of Deposits database Indicator variable equal to 1 for all years once significant fracking activity began in state . The year when significant fracking activity b egan is established based on (Bartik et al. [2019]; Gilje et al. [2016]; Stuber [2019]) as follows: Marcellus (2007), Bakken (2007), Andarko (2008), Haynesville (2008), Niobrara (2010), and Utica (2011). Indicator variable equ al to 1 if a bank has total assets greater than 500 million and 0 otherwise. This criterion is applied for a bank one year before significant fracking activity began in a shale play state. Controls The ratio of non - financial assets (Property, equipment, furniture and fixtures) to interest income for bank in year The ratio of number of employees to interest income for bank in year Log of total assets of bank at the beginning of year Following Hall (2016), is computed as average federal funds rate multiplied by total loans at the beginning of year , where the average fede ral funds rate is computed as the federal funds rate in January and federal funds rate in December of year divided by 2. is computed as the change in FFRLNS of bank from year to year Log - change in total deposits of bank from year to year 52 Total cash and cash equivalents scaled by total assets of bank at the beginning of year High Forecasting Quality Indicator variable equal to 1 if t he ratio of charge offs in t+1 to allowance for loan and lease losses in year t for bank i is greater than or equal to 0.32 and less than 1.68 and 0 otherwise. 0.32 is the 75 th percentile for the ratio in the sample Single County Bank An indicator eq ual to 1 if the bank branches are in a single county High Market Share An indicator equal to 1 if the market share for single county bank is above the median market share. Market share is computed as the ratio of total deposits of a bank i located in c ounty j in year t to the total deposits of county j in year t 53 APPENDIX B Tables 54 TABLE 1 . 1: SHALE - BOOM EXPOSED BANKS AND GROWTH IN DEPOSITS AND INCOME This table reports the estimation from OLS regression s on the growth in deposits and interest income for the period 2005 - 2014. The dependent variable is log of deposits for bank in year in column (1) and log of interest income for bank in year in column (2) . Column (1) estimates the growth in deposits while column (2) estimates the growth is in interest income for banks exposed to the shale boom. The constant is not re ported becaus e it does not have an economic interpretation. Year and bank fixed effects are included, and standard errors are clustered by bank and reported in parenthesis. is measured at the bank level and is time invariant, i ts coefficient i s absorbed by bank fixed effects. Year indicators are relative to the onset of boom and therefore the coefficient on is absorbed by year fixed effects. All variables are defined in Appendix A. *, **, *** indicate significance at the two - tailed 1 0%, 5%, and 1% levels, respectively. (1) (2) Variables 0.0 40 *** 0.0 31 ** * (0.00 7 ) (0.00 9 ) 0. 781 *** 0. 918 *** (0.0 13 ) (0.0 13 ) 0. 131 ** * 0. 824 *** (0.0 37 ) (0.0 40) 0. 002 0. 001 (0.0 10 ) (0.0 12 ) Observations 14,929 14,929 Within R squared 0. 800 0. 649 Bank Fixed Effects Yes Yes Year Fixed Effects Yes Yes SE clustered by Bank Bank 55 TABL E 1 .2 : SAMPLE CONSTRUCTION This table presents the details of sample construction. The primary sample is const ructed by merging the bank regulatory Call Report database with Federal Deposit Insurance Corporation (FDIC) Summary of Deposits (SOD) database to identify the branch locations. The 79,072 Regulatory Filings for 2005 - 2014 represent all banks identified as main office in SOD database . Regulatory Filings 2005 - 2014 for banks identified as main office in SOD database 79,072 Less: Drop banks in top 1% of assets (791) Less: Bank - years with missing data for salaries or interest income either in current or previous year (17,797) Less: Bank - years with salaries greater than interest income either in current or previous year (2,571) Less: Eliminate banks outside the play - state (41,964) Less: Eliminate bank - years with salaries and interest income in top and bottom 1 percentile ( 430 ) Less: Eliminate observations with missing data ( 581 ) Less: Eliminate bank - years that had a mismatch between change in salaries and change in number of employees or changes affected by merger (9) Total number o f bank - year observations 14,929 Total number of unique banks 1,657 56 T ABLE 2 .1: DESCRIPTIVE STATISTICS All continuous variables except labor cost and interest income are winsorized at 1 and 99 percentiles. Lab or cost and interest income are trimmed at 1 and 99 percentiles before constructing this sample. Variables N Mean Std Dev p25 p50 p75 min max Labor cost (000s) 14,929 4,007.012 6,484.451 954.000 2,005.000 4,090.000 213.000 67,983.000 Labor cost def 14,9 29 1,841.825 2,968.441 441.239 930.781 1,867.617 90.818 34,456.730 14,929 0.030 0.156 0.024 0.017 0.065 2.517 2.621 14,929 67.770 95.447 17.000 36.000 73.000 5.000 606.000 14,929 0.020 0.130 0.030 0.000 0.053 1.591 2. 015 14,929 0.010 0.147 0.039 0.010 0.057 2.567 2.804 Interest income (000s) 14,929 13,204.940 21,808.200 3,085.000 6,417.000 13,503.000 587.000 236,489.000 14, 929 0.028 0.176 0.057 0.010 0.095 2.612 3.063 14,929 0.261 0.439 0.000 0.000 1.000 0.000 1.000 14,929 0.606 0.489 0.000 1.000 1.000 0.000 1.000 14,929 0.379 0.285 0.171 0.320 0.514 0.009 1.636 14,929 5.867 2.044 4.470 5.670 7.002 1.820 12.526 Average fed funds rate 14,929 1.620 1.899 0.110 0.140 3.285 0.090 4.770 (000s) 14,929 239,772.200 719,321.200 8,050.0 35 32,222.970 183,759.600 277.725 13,900,000.000 14,929 0.038 0.625 0.191 0.040 0.202 0.962 11.789 Assets (000s) 14,929 275,299.400 470,866.200 63,446.000 131,068.000 277,751.000 10,152.000 5,720,415.000 14,929 11.853 1.094 11.058 11.783 12.534 9.225 15.560 Deposits (000s) 14,929 215,637.900 318,452.500 53,044.000 109,455.000 229,677.000 13,240.000 2,012,555.000 57 TABLE 2.1 ) 14,929 11.659 1.072 10.879 11.603 12.344 9.491 14.515 14,929 0.059 0.123 0.000 0.044 0.096 1.363 1.921 14,929 0.753 0.599 0.328 0.593 1.018 0.002 2.666 14,929 0.108 0.310 0.000 0.000 0.000 0.000 1.000 14,929 0.067 0.062 0.028 0.045 0.082 0.007 0.373 14,929 5.470 6.941 2.000 3.000 6.000 1.000 44.000 58 T ABLE 2.2: CORRELATIONS This table presents the Pearson correla tions among the variables used in analyses. Numbers in shaded boxes represent significant correlations at 10% level. Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (1) 1.000 (2) 0.681 1.000 (3) 0.082 0.040 1.000 (4) 0.483 0.356 0.102 1.000 ( 5 ) 0.009 0.003 0.010 0.006 1.000 ( 6 ) 0.022 0.296 0.035 0.047 0.055 1.000 ( 7 ) 0.020 0.069 0.101 0.045 0.013 0.124 1.000 ( 8 ) 0.075 0.161 0.016 0.046 0.029 0.119 0.394 1.000 ( 9 ) 0.092 0.435 0.010 0.098 0.004 0.413 0.026 0.012 1.000 (1 0 ) 0.089 0.029 0.774 0.103 0.040 0.121 0.112 0.224 0.029 1.000 ( 11 ) 0.397 0.481 0.069 0.514 0.025 0.047 0.035 0.053 0.092 0.107 1.000 (1 2 ) 0.005 0.045 0.197 0.021 0.069 0.017 0.237 0.034 0.030 0.373 0.040 1.000 (1 3 ) 0.040 0.029 0.732 0.049 0.007 0.018 0.005 0.134 0.024 0.620 0.044 0.164 1.000 (1 4 ) 0.037 0.106 0.157 0.057 0.037 0.201 0.013 0.210 0.068 0.220 0.102 0.084 0.129 1.000 (1 5 ) 0.046 0.010 0.925 0.042 0.006 0.046 0.127 0.012 0.005 0.736 0.032 0.197 0.688 0.146 1.000 59 T ABLE 2.3 : DIFFERENCE IN MEANS This table presents the differences in means for banks exposed to the shale boom and banks not expos ed to the shale boom. Column (2) to (4) presents the mean and difference in mean for average deflated labor cost per branch. Column (5) to (7) computes the mean and difference in mean for interest income deflated per employee. Column (8) to (10) computes t he mean and difference in mean for log deposits. Year log deposits (1) Exposed Banks (2) Banks (3) Diff (4) Exposed Banks (5) Banks (6) Diff (7) Exposed Banks (8) Banks (9) Diff (10) Mean Mean Mean Mean Mean Mean Before the shale boom 11.880 12.724 0.843** 101.255 101.285 0.030 11.553 11.465 0.088*** First two years 11.452 12.050 0.598 95.809 92.069 3.740* 11.672 11.633 0.039 Next two years 11.829 11.896 0.067 86.240 83.314 2.927 11.815 11.733 0.083* After 4 years 11.917 11.060 0.857** 79.134 75.428 3.704*** 11.975 11.881 0.093** 60 TABL E 3 .1 : DIFFERENCE IN LABOR COST ELASTICITY FOR SHALE BOOM EXPOSED BANKS This table reports the estimation from OLS regression on the cost elasticity for the period 2005 - 2014. The dependent variable is log - change in deflated labor cost from year t - 1 to t. All the columns estimate the change in cost elasticity for all banks in t he sample. Columns (1) to (4) do not include any control variable whereas column (5) includes control v ariables. Column (2) includes state fixed effects, column (3) includes county fixed effects and columns (4) and (5) include bank fixed effects. Year fixe d effects are included in all columns, and standard errors are clustered by bank and reported in parent hesis. is measured at the bank level and its coefficient is absorbed by bank fixed effects in columns (4) and (5). Year indicat ors are relative to the onset of boom and therefore the coefficient on is absorbed by year fi xed effects. All variables are defined in Appendix A . *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respectively. Variables (1) (2) (3) (4) (5) 0. 598 *** 0.6 30 *** 0.6 20 *** 0.6 14 *** 0.6 81 *** (0.046) (0.04 6 ) (0.05 1 ) (0.054) (0.048) 0.01 1 ** 0.01 0 * 0.0 04 (0.006) (0.005) (0.0 10 ) 0.0 37 0.0 29 0.03 7 0.0 55 0.0 07 (0.0 74 ) (0.0 68 ) (0.0 76 ) (0.087) (0.06 4 ) 0.1 62 *** 0.1 31 ** 0.1 39 ** 0.1 4 4 ** 0. 062 (0.05 4 ) (0.05 4 ) (0.05 8 ) (0.06 3 ) (0.0 44 ) 0.00 3 0.00 3 0.00 4 0.00 8 0.00 9 (0.007) (0.00 6 ) (0.00 7 ) (0.00 8 ) (0.00 7 ) 0. 189 ** 0. 19 4 * * 0. 191 * 0. 200 * 0. 180 ** (0. 095 ) (0. 091 ) (0. 099 ) (0.1 08 ) (0.0 84 ) 0.01 4 (0.01 1 ) 0.0 4 6 (0.0 36 ) 0.02 3 *** (0.00 2 ) 0.0 13 *** (0.00 4 ) 0.0 56 *** (0.0 05 ) 0.05 3 *** (0.0 09 ) 0.0 42 ** (0.02 0 ) Observations 14,929 14,929 14,929 14,929 14,929 0.515 0.522 0.510 0. 490 0.547 Fixed Effects Year Year, State Year, County Year, Bank Year, Bank SE clustered by Bank Bank Bank Bank Bank 61 TABLE 4 .1 : DIFFERENCE IN LABOR COST ELASTICITY PER EMPLOYEE AND NUMBER OF EMPLOYEES ELASTICITY FOR SHALE BOOM EXPOSED BANKS This table reports the estimation from OLS regression on the labor cost per employee elasticity and labor employee elasticity for the pe riod 2005 - 2014. The dependent variable in column (1) and (2) is log change in labor cost per employee from year t - 1 to t and the dependent vari able in column (3) and (4) is log - change in number of employees from year t - 1 to t. Column (1) and (2) estimate t he change in labor cost per employee elasticity whereas c olumn ( 3 ) and ( 4 ) estimate the change in labor employee elasticity after the shale boo m for all banks in the sample. County and year fixed effects are included in column (1) and (3) whereas b ank and year fixed effects are included in column (2) and (4) . is measured at the bank level and its coefficient absorbed by bank fixed effects in column (2) and (4). Year indicators are relative to the onset of boom and therefore the coefficient on is absorbed by year fixed effects. All variables are defi ned in Appendix A . *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respectively. Variables (1) (2) (3) (4) 0.505 *** 0. 479 *** 0.142 *** 0. 204 *** (0.054) (0.0 55 ) (0.026) (0.030) 0.011 0.010 (0.009) (0.010) 0.045 0.031 0.046 0.039 (0.070) (0.074) (0.046) (0.050) 0.078 0.081 0.001 (0.055) (0.059) (0.032) (0.039) 0.004 0.010 0.003 0.002 (0.007) (0.007) (0.005) (0.009) ** 0. 326 *** 0.120 ** 0. 146 ** (0.083) (0.087) (0.061) (0.065) 0.008 * 0.037 *** 0.016 *** 0.052 *** (0.005) (0.011) (0.005) (0.012) 0.033 0.0 40 0.017 0.005 (0.023) (0.026) (0.018) (0.027) 0.007 *** 0.021 *** 0.014 *** 0.045 *** (0.001) (0.002) (0.001) (0.003) 0.005 0.001 0.001 0.012 *** (0.003) (0.003) (0.002) (0.003) 0.038 *** 0.029 *** 0.008 ** 0.027 *** (0.006) (0.006) (0.003) (0.003) ** 0.009 *** 0.076 *** (0.001) (0.011) (0.001) (0.011) *** *** 0.429 *** 0.378 *** (0.032) (0.030) (0.026) (0.024) Observations 14,929 14,929 14,929 14,929 0.336 0.350 0.288 0.331 Fixed Effects County, Y ear Bank, Year County, Year Bank, Year SE clustered by Bank Bank Bank Bank 62 TABLE 5 .1 : SHALE BOOM EXPOSED BANKS AND DYNAMIC CHANGES IN LABOR COST AND LABOR EMPLOYEE ELASTICITY This table reports the marginal effects of the dynamic changes in labor cost elasticity and labor employee elasticity estimated from OLS regression for the period 2005 - 2014. The dependent variable for columns (1) to (3) is log ch ange in deflated labor cost from year t - 1 to t. Column (1) and (2) estimates the labor cost elasticity before and after the shale boom for banks exposed to the shale boom and banks not exposed to the shale boom respectively whereas column (3) measures the difference in columns (1) a nd (2). The dependent variable for columns (4) to (6) is log - change in number of employees from year t - 1 to t. Column (4) and (5) estimates the labor employee elasticity before and after the shale boom for banks exposed to the sh ale boom and banks not expo sed to the shale boom respectively whereas column (6) measures the difference in columns (4) and (5). Bank and year fixed effects are included. All control variables from Table 3 .1 are included but not reported for brevity. All var iables are defined in Appen dix A . *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respectively. VARIABLES Exposed Banks ed Banks Difference Exposed Banks Non Expos ed Banks Difference (1) (2) (3) (4) (5) (6) Before the shale boom 0.658 *** 0.670 *** 0.159 *** 0.197 *** (0.063) (0.043) (0.065) (0.046) (0.029) (0.050) First two years in the shale boom 0.441 *** 0.707 *** *** 0.233 *** 0.170 *** 0.063 (0.058) (0.060) (0.075) (0.043) (0.029) (0.049) Next two years in the shale boom 0.523 *** 0.730 *** ** 0.321 *** 0.229 *** 0.092 (0.079) (0.054) (0.093) (0.060) (0.049) (0.076) After four years (Long run effect of the shale boom) 0.688 *** 0.797 *** 0.109 0.330 *** 0.178 *** 0.152 ** (0.104) (0.033) (0.107) (0.054) (0.053) (0.071) Difference in elasticity 4 years after and before the shale boom 0.029 0.126 ** 0.097 0.171 ** 0.018 0.190 ** (0.118) (0.050) (0.122) (0.071) (0.059) (0.087 ) Controls Yes Yes Observations 14,929 14,929 0.550 0.332 Bank Fixed Effects Yes Yes Year Fixed Effects Yes Yes SE clustered by Bank Bank 63 TABLE 6 .1 : SHALE BOOM EXPOSED BANKS, FORECASTING QUALITY, AND DYNAMIC CHANGES IN LABOR EMPLOYEE ELASTICITY This table reports the marginal effects of the dynamic changes in labor employee elasticity estimated for banks wit h high forecastin g quality and low forecasting quality from OLS regression for the period 2005 - 2014. The dependent variable for columns (1) to (6) is log change in number of employees from year t - 1 to t. Column (1) and (2) estimates the number of employees elasticity before and after the shale boom fo r banks exposed to the shale boom and banks not exposed to the shale boom respectively for banks with high forecasting quality whereas column (3) measures the difference in columns ( 1) and (2). Column (4) and ( 5) estimates the labor employee elasticity bef ore and after the shale boom for banks exposed to the shale boom and banks not exposed to the shale boom respectively for banks with low forecasting quality whereas column (6) measures the difference in columns (4) and (5). Bank and year fixed effects are included. All control variables from Table 4 .1 are included but not reported for brevity. All variables are defined in Appendix A . *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respec tively. VARIABLES Exposed Banks Non Expos ed Banks Difference Exposed Banks Non Expos ed Banks Difference (1) (2) (3) (4) (5) (6) Before the shale boom 0.225 *** 0.296 *** 0.071 0.151 ** 0.192 *** 0.041 (0.062) (0.058) (0.076) (0.062) (0.035) (0.065) First two years in the shale boom 0.306 ** 0.193 *** 0.113 0.208 *** 0.138 *** 0.070 (0.122) (0.072) (0.137) (0.047) (0. 039) (0.055) Next two years in the shale boom 0.460 *** 0.282 *** 0.178 0.287 *** 0.197 *** 0.090 (0.123) (0.067) (0.131) (0.071) (0.054) (0.076) After four years (Long run effect of the shale boom) 0.342 *** 0.196 * 0.146 0.328 *** 0.179 ** * 0.149 * (0.087) (0.106) (0.121) (0.062) (0.067) (0.084) Difference in elasticity 4 years after and before the shale boom 0.117 0.100 0.217 0.177 ** 0.014 0.191 * (0.095) (0.108) (0.138) (0.089) (0.072) (0.106) Difference in elasticity next two years in the shale boom and before the shale boom 0.235 * 0.014 0.249 * 0.136 0.004 0.132 (0.131) (0.073) (0.146) ( 0.092) (0.062) (0.110) Difference in elasticity first two years in the shale boom and before the shale boom 0.081 0.103 0.184 0.057 0.054 0.111 (0.129) (0.083) (0.152) (0.074) (0.046) (0.086) Controls Yes Yes Observations 3,420 11, 509 Within R squared 0.352 0.322 Bank Fixed Effects Yes Yes Year Fixed Effects Yes Yes SE clustered by Bank Bank 64 TABLE 7 .1 : SHALE BOOM EXPOSED BANKS AND PRODUCT DIVERSITY This ta ble reports the estimation results on product diversity for the period 2005 - 2014. T he estimation for product diversity is obtained from OLS regression. The dependent variable in column ( 1 ) and ( 2 ) is the product diversity of bank in year t. Column ( 1 ) and ( 2 ) estimate the difference in product diversity for exposed and non - exposed banks. Column ( 2 ) estimates the heterogenous effect of the shale boom on product diversity for large and small banks. Year and bank fixed effects are included, and standard errors are clustered by bank and reported in parenthesis. , , and are measured at the bank level and are time invariant, their coefficient is absorbed by ba nk fixed effects. All variables are def ined in Appendix A . *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respectively . ( 1 ) (2) VARIABLES 0.023 0.023 (0.0 31 ) (0.031) 0.023 ** 0.025 ** (0.011) (0.012) 0.029 ** (0.011) 0.022 (0.023) 0. 055 *** 0.056 *** (0.0 17 ) (0.017) 0.001 0.001 (0.004) (0.004) 0.0 09 *** 0.00 9 *** (0.00 3 ) (0.003) 0. 058 ** 0.059 ** (0.0 26 ) (0.026) 0. 013 0.011 (0.0 25 ) (0.025) 0. 025 0.029 (0.0 76 ) (0.076) Observations 14,929 14,929 0.090 0.091 Bank Fixed Effects Yes Yes Year Fixed Effects Yes Yes SE clustered by Bank Bank 65 TABLE 8 .1 : SHALE BOOM EXPOSED BANKS AND DYNAMIC CHANGES IN PRODUCT DIVERSITY This table reports the marginal effects of the dynamic changes in product diversity estimated from OLS regression for the period 2005 - 2014. The dependent variable for columns (1) to (3) is Product diversity for bank in year t. Colu mn (1) a nd (2) estimates the product diversity before and after the shale boom for banks exposed to the shale boom and banks not exposed to the shale boom respectively whereas column (3) measures the difference in columns (1) and (2). Bank and year fixed e ffects a re included. The marginal effect for product diversity before the shale boom is absorbed by bank fixed effects. All control variables from Table 5 .1 are included but not reported for brevity. All variables are defined in Appendix A . *, **, *** indica te signi ficance at the two - tailed 10%, 5%, and 1% levels, respectively. VARIABLES Exposed Banks Banks Difference (1) (2) (3) Before the shale boom Absorbed by Bank fixed effects First two years in the sha le boom (a) 0.752 *** 0.755 *** (0.008) (0.002) (0.010) Next two years in the shale boom 0.738 *** 0.751 *** (0.011) (0.004) (0.013) After four years (b) (Long run effect of the shale boom) 0.681 *** 0.738 *** *** (0.01 4) (0.008) (0.015) (b) (a) 0.071 *** 0.017 0.054 *** (0.015) (0.010) (0.015) Controls Yes Observations 14,929 0.093 Bank Fixed Effects Yes Year Fixed Effects Yes SE clustered by Bank 66 TABLE 9 .1 : SHALE BOOM EXPOS ED BANKS, GEOGRAPHICAL MARKETS, AND DIFFERENCE IN COST AND LABOR EMPLOYEE ELASTICITY This table reports the estimation from OLS regressio n on the cost elasticity and labor employee elasticity for the period 2005 - 2014. For Columns (1) and (2), the dependent variable is log - change in deflated labor cost from year t - 1 to t. Columns (3) and (4 ), the dependent variable is log - change in employees from year t - 1 to t. Column (1) and (3) estimate the change in cost elasticity and labor employee elasticity for banks operating in multiple counties . Column ( 2 ) and (4) estimate the change in cost elasti city and labor employee elasticity for bank s operating in single county . Bank and year fixed effects are included, and standard errors are clustered by bank and reported in parenthesis. is measured at the bank level and its coeffi cient is absorbed by bank fixed effects. Year indicators are relative to the onset of the boom and therefore the coefficient on is absorbed by year fixed effects. All va riables are defined in Appendix A . *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respectively. Variables (1) (2) (3) (4) Multiple Count ies Singl e County Multiple Count ies Single County 0.683 *** 0.662 *** 0.243 *** 0.222 *** (0.053) (0.073) (0.046) (0.034) 0.040 0.057 0. 076 0.017 (0.058) (0.114) (0.065) (0.058) 0.045 0.077 0.003 0.080 * (0.056) (0.092) (0.060) (0.048) 0.012 0.003 0.001 0.008 (0.008) (0.010) (0.008) (0.009) 0.032 0.374 ** 0.214 ** 0.112 (0.084) (0.146) (0.089) (0.075) 0.049 *** 0.002 0.094 *** 0.010 (0.016) (0.012) (0.019) (0.015) 0.051 0.054 0.034 0.052 * (0.054) (0.045) (0.052) (0.032) 0.029 *** 0.019 *** 0.050 *** 0.045 *** (0.002) (0.003) (0.004) (0.004) 0.012 0.011 ** 0.014 ** 0.012 *** (0.008) (0.004) (0.006) (0.004) 0.059 *** 0.051 *** 0.035 *** 0.025 *** (0.006) (0.007) (0.005) (0.004) 0.043 *** 0.077 *** 0.070 *** 0.147 *** (0.012) (0.020) (0.014) (0.020) 0.074 *** 0.006 0.423 *** 0.293 *** (0.028) (0.020) (0.033) (0.035) Observations 7,707 7,222 7,707 7,222 Within R squared 0.536 0.548 0.389 0.293 Fixed Effects Year, Bank Year, Bank Year, Bank Year, Bank SE clustered by Bank Bank Bank Bank 67 TABLE 10 .1 : SHA LE BOOM EXPOSED BANKS, MARKET SHARE, AND DIFFERENCE IN COST AND LABOR EMPLOYEE ELASTICITY This table reports the estimation from OLS regression on the cost elasticity and labor employee elasticity for the period 2005 - 2014 for the sub - sample of banks operating in a single county. For Columns (1) and (2), the dependent variable is log - change in d eflated labor cost from year t - 1 to t. Columns (3) and (4), the dependent variable is log - change in employees from year t - 1 to t. Column (1) and (3) estimate the change in labor cost elasticity and labor employee elasticity for single county banks with low market share. Column (2) and (4) estimate the change in labor cost elasticity and labor employee elasticity for single county banks with high market share. Bank and year fixed effects are included, and standard errors are clustered by bank and reported in parenthesis. is measured at the bank level and its coefficient is absorbed by bank fixed effects. Year indicators are relative to the onset of the boom and therefore the coefficient on is absorbed by year fixed effects . All va riables are defined in Appendix A. *, **, *** indicate significance at the two - tailed 10%, 5%, and 1% levels, respectively. Variables (1) (2) (3) (4) Low market share High m arket share Low market share High market share 0. 509 *** 0. 791 *** 0.205 *** 0.226 *** (0.0 57 ) (0.0 73 ) (0.046) (0.047) 0.261 ** 0. 188 0.056 0.162 * (0. 113 ) (0. 194 ) (0.065) (0.094) 0. 278 *** 0. 112 0.038 0.122 * (0.0 84 ) (0. 108 ) (0.075) (0.066) 0.023 * 0.0 37 ** 0. 020 0.006 (0.0 13 ) (0.0 16 ) (0.014) (0.011) 0. 521 *** 0. 219 0.058 0.270 ** (0. 149 ) (0. 218 ) (0.099) (0.123) 0. 009 0.0 27 0.023 0.024 (0.01 6 ) (0.0 22 ) (0.021) (0.023) 0.0 24 0.0 91 0.050 0.081 ** (0.0 52 ) (0.0 56 ) (0.043) (0.040) 0.02 1 *** 0.0 19 *** 0.052 *** 0.039 *** (0.00 3 ) (0.00 5 ) (0.007) (0.003) 0.0 11 ** 0.0 13 * 0.013 *** 0.008 ** (0.00 5 ) (0.00 7 ) (0.005) (0.004) 0.0 49 *** 0.0 55 *** 0.023 *** 0.026 *** (0.0 07 ) (0.0 12 ) (0.006) (0.005) 0.0 62 *** 0. 115 *** 0.139 *** 0.153 *** (0.0 24 ) (0.0 25 ) (0.024) (0.023) 0 .0 11 0.0 02 0.287 *** 0.306 *** (0.0 33 ) (0.0 42 ) (0.051) (0.039) Observations 3,611 3,611 3,611 3,611 Within R squared 0.544 0.594 0.274 0.355 Fixed Effects Year, Bank Year, Bank Year, Bank Year, Bank SE clustered by Bank Bank Bank Bank 68 APPENDI X C Figures 69 FIGURE 1: MAJOR U.S. SHALE PLAY REGIONS This figure shows the seven major U.S. Shale play regions with fracking activities during 2005 - 2014. Appalachia basin is subdivided into two regions - Marc ellus and Utica because significant fracking in Utica region (Ohio) began only in 2011 whereas signifi cant fracking activity in Marcellus region (New York, Pennsylvania, and West Virginia) began in 2007. Counties exposed to the shale boom are identified by Energy of Information Administration (EIA). Commencement of significant fracking activity is recogniz ed based on prior research (Bartik et al. [201 9 ]; Gilje et al. [2016]) . Following Stuber (2019), Texas, New Mexico and New York states are excluded from t he analyses. Therefore, Permian (New Mexico and Texas) and Eagle Ford region (Texas) are completely ex cluded whereas only a part of Haynesville and Anadarko region is excluded from analyses to the extent the counties are in Texas. Similarly, a part of Marc ellus region is excluded from analyses to the extent the counties are in New Y ork. 70 FIGURE 2 : TREATMENT AND CONTROL COUNTIES WITHIN THE SHALE PLAY STATES This figure shows th e treatment and control counties used to identify the banks exposed to the shale boom within a shale play state. Counties shaded in green are identified by U.S. Energy Information Administration (EIA) as counties with shale formation. Banks with more than 50 per cent of branches in counties shaded in green are identified as expos ed to the shale boom while other banks within the play state are considered as control banks. Therefore, control banks may also have a minority share of branches in the counties sha ded in green. 71 FIGURE 3: TIME SERIES ANALYSIS FOR DIFFERENCE IN L ABOR COST ELASTICITY The figure shows the coefficients on obtained by regressing on year indicator variables relative to the onset of the shale b oom and the interaction of , year indicators, and . All control variables used in estimating equation (1) are also included. The points on the bars represent the difference in labor cost elasticity for exposed and non - exposed banks during the sample period while the bars show the 95% confidenc e intervals. Year on the horizontal axis is relative to the onset of the shale boom. The marginal effects 4 years after and before the onset of the shale boom are combined and shown under year 5 and - 5 respectively. 72 FIGURE 4: TIME SERIES ANALYSIS FOR DIFFERENCE IN LABOR EMPLOYEE ELASTICITY The figure shows the coefficients o n obtained by regressing on year indicator variables relative to the onset of the shale boom an d the interaction of , year indicators, and . All control variables used in estimating equation (1) are also included. The points on the bars represent the difference in labor cost elasticity for e xposed and non - exposed banks during the sample period while the bars show the 95% confidence intervals. Year on the horizontal axis is relative to the onset of the shale boom. The marginal effects 4 years after and before the onset of the shale boom are co mbined and shown under year 5 and - 5 respectively. 73 FIGURE 5: TIME SERIES ANALYSIS FOR DIFFERENCE IN PRODUCT DIVERSITY The figure shows the coefficients on obtained by regressing on year indicator variables relative to the onset of the shale boom and the interaction of year indicators, and . All control variables used in estimating equation ( 2 ) are also incl uded. The points on the bars represent the difference in labor cost elasticity for exposed and non - exposed banks during the sample period while the bars show the 95% confidence intervals. Year on the horizontal axis is relative to the onset of the shale bo om. 74 REFERENCES 75 REFERENCES Abadie, A., S. Athey, G. Imbens, and J. Wooldridge When Should You Adjust Standard Errors for Clustering? (No. w24003). Cambridge, MA: National Bureau of Economic Research. 2017. https://doi.org/10.3386/w24003 Anderson, M. C., R. D. Banker, and S. N. Janakiraman "Are selling, general, and administrative Journal of Accounting Research 41 (2003 ): 47 63. Baker, S. R., N. Bloom, and S. J. Davis "Measur ing Economic Policy Uncertainty." The Quarterly Journal of Economics 131 (2016): 1593 1636. Balakrishnan, R., E. Labro, and N. S. Soderstrom "Cost structure and sticky costs." Journal of Management A ccounting Research 26 (2014): 91 116. Balakrishnan, R., M . J. Petersen, and N. S. Soderstrom "Does capacity utilization affect the Journal of Accounting, Auditing & Finance 19 (2004): 283 300. Banker, R. D., and D. Byzalov "Asymmetri c cost behavior." Journal of Management Accounting Resear ch 26 (2014): 43 79. Banker, R. D., D. Byzalov, M. Ciftci, and R. Mashruwala "The Moderating Effect of Prior Sales Changes on Asymmetric Cost Behavior." Journal of Management Accounting Research 26 ( 2014): 221 242. Banker, R. D., D. Byzalov, S. Fang, and Y . Liang "Cost Management Research." Journal of Management Accounting Research 30 (2018): 187 209. Banker, R. D., D. Byzalov, and J. M. Plehn - Dujowich "Demand Uncertainty and Cost Behavior." The Accou nting Review 89 (2014): 839 865. Banker, R. D., S. M. Dat ar, and S. Kekre "Relevant costs, congestion and stochasticity in production environments." Journal of Accounting and Economics 10 (1988): 171 197. Banker, R. D., and J. S. Hughes "Product costing an d pricing." The Accounting Review (1994): 479 494. Bartik , A. W., J. Currie, M. Greenstone, and C. R. Knittel "The Local Economic and Welfare Consequences of Hydraulic Fracturing." American Economic Journal: Applied Economics 11 (2019): 105 155. Beatty, A. , and S. Liao "Do delays in expected loss recognition aff lend?" Journal of Accounting and Economics 52 (2011): 1 20. - loan structures." Jo urnal of Accounting and Economics 67 (2019): 496 520. 76 Ber nard, A. B., S. J. Redding, and P. K. Schott "Multiple - Product Firms and Product Switching." American Economic Review 100 (2010): 70 97. Bloom, Nicholas "The impact of uncertainty shocks." Econometri ca 77 (2009): 623 685. Bloom, Nicholas "Fluctuations in u ncertainty." Journal of Economic Perspectives 28 (2014): 153 176. Bloom, Nick, S. Bond, and J. Van Reenen "Uncertainty and investment dynamics." The Review of Economic Studies 74 (2007): 391 415. Bro wn, J. P., T. Fitzgerald, and J. G. Weber "Capturing rent s from natural resource abundance: Private royalties from US onshore oil & gas production." Resource and Energy Economics 46 (2016): 23 38. Brüggen, A., R. Krishnan, and K. L. Sedatole "Drivers and c production decisions: Evidence from the auto industry." Contemporary Accounting Research 28 (2011): 83 123. Caballero, R. J., E. M. R. A. Engel, and J. Haltiwanger "Aggregate Employment Dynamics: Building from Microeconomic Eviden ce." The American Economic Review 87 (1997): 115 137. Can air transportation industry data." The Accounting Review 89 (2014): 1645 1672. Cantrell, B. W., J. M. McIn nis, and C. G. Yust "Predicting credit losses: Loan fair values versus historical costs." The Accounting Review 89 (2014): 147 176. Carlton, D. W., and J. D. Dana "Product Variety and Demand Uncertainty: Why Markups Vary with Quality ? " The Journal of Industrial Economics 56 (2008): 535 552. Cedeño,Wander "How di Review: US Bureau of Labor Statistics.". 2018, October. https://www.bls.gov/opub/mlr/2018/article/how - did - employment - fare.htm (Acc essed 2020). Cerulli , G., and M. Ventura "Estimation of pre - a nd posttreatment average treatment effects with binary time - varying treatment using Stata." The Stata Journal 19 (2019): 551 565. Chen, C. X., H. Lu, and T. Sougiannis "The agency problem, corporate governance, and the asymmetrical behavior of selling, gen eral, and administrative costs." Contemporary Accounting Research 29 (2012): 252 282. Chen, H. J., M. Kacperczyk, and H. Ortiz - Molina "Labor unions, operating flexibility, and the cost of equity." Journal of Financ ial and Quantitative Analysis 46 (2011): 2 5 58. 77 Chen, J. V., I. Kama, and R. Lehavy "A contextual analysis of the impact of managerial expectations on asymmetric cost behavior." Review of Accounting Studies 24 (2019): 665 693. Cooper, R., and R. S. Kaplan "Activity - based systems: Measuring the cos ts of resource usage." Accounting Horizons 6 (1992): 1 13. Craig Nichols, D., J. M. Wahlen, and M. M. Wieland "Publicly traded versus privately held: implications for conditional conservatism in bank accounting." R eview of Accounting Studies 14 (2009): 88 122. Dechow, P. M., and I. D. Dichev "The quality of accruals and earnings: The role of accrual estimation errors." The Accounting Review 77 (2002): 35 59. outsourcing to temporary help services: A research update." US Bureau of Labor Statistics, Working Paper 493 (2017). Dichev, I. D., and V. W. Tang "Earnings volatility and earnings predictability." Journal of Accounting and Economics 47 (2009): 160 181. D ierynck, B., W. R. Landsman, and A. Render s "Do Managerial Incentives Drive Cost Behavior? Evidence about the Role of the Zero Earnings Benchmark for Labor Cost Behavior in Private Belgian Firms." The Accounting Review 87 (2012): 1219 1246. Eberhart, D. " T he Economic Impact o f U S Shale. " Canary LL c (2014): 30. Gilje, E. P. "Does Local Access to Finance Matter? Evidence from US Oil and Natural Gas Shale Booms." Management Science 65 (2019): 1 18. Gilje, E. P., E. Loutskina, and P. E. Strahan "Exporting Liqui dity: Branch Banking and Financial Integra tion." The Journal of Finance 71 (2016): 1159 1184. Göx, R. F. "Capacity planning and pricing under uncertainty." Journal of Management Accounting Research 14 (2002): 59 78. Graham, J. R., C. R. Harvey, and S. Rajg opal "The economic implications of corpora te financial reporting." Journal of Accounting and Economics 40 (2005): 3 73. Granger, C. W. "Investigating causal relations by econometric models and cross - spectral methods." Econometrica: Journal of the Econometr ic Society (1969): 424 438. Grenadier, S. R., and A. Malenko "A Bayesian Approach to Real Options: The Case of Distinguishing between Temporary and Permanent Shocks." The Journal of Finance 65 (2010): 1949 1986. Guiso, L., L. Pistaferri, and F. Schivardi " Insurance within the Firm." Journal of Pol itical Economy 113 (2005): 1054 1087. 78 Hall, C. M. "Does Ownership Structure Affect Labor Decisions?" The Accounting Review 91 (2016): 1671 1696. Hamermesh, D. S. "Labor Demand and the Structure of Adjustment Costs. " The American Economic Review 79 (1989): 674 689. Portfolios." The Accounting Review 93 (2018): 245 271. Heinrich, C. J., and S. N. Houseman "Worker Hard and Soft Ski lls and Labor Market Outcomes: A Lens thro ugh the Temporary Help Industry over the Business Cycle." (2019): 47. Holzhacker, M., R. Krishnan, and M. D. Mahlendorf "The impact of changes in regulation on cost behavior." Contemporary Accounting Research 32 (2 015): 534 566. Holzhacker, M., R. Krishnan , and M. D. Mahlendorf "Unraveling the Black Box of Cost Behavior: An Empirical Investigation of Risk Drivers, Managerial Resource Procurement, and Cost Elasticity." The Accounting Review 90 (2015): 2305 2335. Ilut , C., M. Kehrig, and M. Schneider "Slow to hire, quick to fire: Employment dynamics with asymmetric responses to news." Journal of Political Economy 126 (2018): 2011 2071. Imbens, G. W., and J. M. Wooldridge "Recent Developments in the Econometrics of Prog ram Evaluation." Journal of Economic Liter ature 47 (2009): 5 86. International Monetary Fund "World Economic Outlook: IMF Sees Heightened Risks Sapping Slower Global Recovery.". 2012, October 9. https://www.imf.org/en/News/Articles/2015/09/28/04/53/sores10 0812a (Accessed 2020). Irvine, P. J., S. S - Base Concentration, Profitability, and the Relationship Life Cycle." The Accounting Review 91 (2016): 883 906. Jurado, K., S. C. Ludvigson , and S. Ng "Measuring uncertainty." American Economic Review 105 (2015): 1177 1216 . Kallapur, S., and L. Eldenburg "Uncertainty, Real Options, and Cost Behavior: Evidence from Washington State Hospitals." Journal of Accounting Research 43 (2005): 735 752. Kama, I., and D. Weiss "Do earnings targets and managerial incentives affect stick y costs?" Journal of Accounting Research 51 (2013): 201 224. Katz, L. F., A. B. Krueger, G. Burtless, and W. T. Dickens "The High - Pressure US Labor Market of the 1990s." Bro okings Papers on Economic Activity 1999 (1999): 1 87. Khan, U., and N. B. Ozel "Rea l Activity Forecasts Using Loan Portfolio Information." Journal of Accounting Research 54 (2016): 895 937. 79 Kocherlakota, N. "Inside the FOMC | Federal Reserve Bank of Minnea polis.". 2010, August 17. https://www.minneapolisfed.org:443/speeches/2010/inside - t he - fomc (Accessed 2020). Lake, L. W., J. Martin, J. D. Ramsey, and S. Titman "A Primer on the Economics of Shale Gas Production Just How Cheap is Shale Gas?" Journal of Appl ied Corporate Finance 25 (2013): 87 96. Mandelker, G. N., and S. G. Rhee "The impac t of the degrees of operating and financial leverage on systematic risk of common stock." Journal of Financial and Quantitative Analysis 19 (1984): 45 57. Miller, D., and J. Shamsie "Strategic Responses to Three Kinds of Uncertainty: Product Line Simplicit y at the Hollywood Film Studios." Journal of Management 25 (1999): 20. Noreen, E., and N. Soderstrom "Are overhead costs strictly proportional to activity?: Evidence from ho spital departments." Journal of Accounting and Economics 17 (1994): 255 278. Noreen , E. W., P. C. Brewer, and R. H. Garrison Managerial accounting for managers . McGraw - Hill/Irwin New York. 2014. Al lowance for Loan and Lease 1998. Ono, Y., and D. Su Census Microdata." Industrial Relations: A Journal of Economy and Society 52 (2013): 419 443. Petru zzi, N. C., and M. Dada "Pricing and the newsvendor problem: A review with extensio ns." Operations Research 47 (1999): 183 194. Pinnuck, M., and A. M. Lillis "Profits versus Losses: Does Reporting an Accounting Loss Act as a Heuristic Trigger to Exercise t he Abandonment Option and Divest Employees?" The Accounting Review 82 (2007): 1031 1053. Staff Reports (2014): 60. Reed, A., S. Ericson, M. Bazilian, J. Logan, K. Doran, and C. Nelder "Interrogating uncertainty in energy forecasts: the case of th e shale gas boom." Energy Transitions 3 (2019): 1 11. https://www.workforce.com/news/k elly - girl - turns - 66 - an - interview - with - carl - camden (Accessed 2020). Roychowdhury, S. "Earnings management through real activities manipulation." Journal of Accounting and Economics 42 (2006): 335 370. 80 Sedatole, K. L., D. Vrettos, and S. K. Widener "The use o f management control mechanisms to mitigate moral hazard in the decision to outsour ce." Journal of Accounting Research 50 (2012): 553 592. Sloan, A. "Do Stock Prices Fully Reflect Information in Accruals and Cash Flows About Future Earnings?" The Accountin g Review 71 (1996): 289 315. Stuber, S. B. The Effect of Growth on Financial Report ing and Audit Quality: Evidence from Economic Shocks to Banks. Michigan State University. Business Administration. 2019. Tallman, S., and J. Li "Effects of international div ersity and product diversity on the performance of multinational firms." Academy of Management Journal 39 (1996): 179 196. Times - Picayune, R. T. S., The "Drilling rush in North Louisiana creates new millionaires.". 2008. https://www.nola.com/news/article_c 3742712 - a908 - 587e - ae16 - 5ab729bba789.html (Accessed 2020). Tversky, A., and D. Kahne man "Judgment under Uncertainty: Heuristics and Biases." Science 185 (1974): 1124 - Time Work?" San Franc isco Federal Reserve Bank Economic Letter 24 (2013): 1 5. Van Mieghem, J. A., and M . Dada "Price versus production postponement: Capacity and competition." Management Science 45 (1999): 1639 1649. Wang, Z., and A. Krupnick "A retrospective review of shale gas development in the United States: What led to the boom?." Economics of Energy & Environmental Policy 4 (2015): 5 18. The Accounting Review 85 (2010): 1441 1471. Wooldridge, J. M. Econometric analysis of cross section and panel data. MIT press. 2010. Wu, X. "Deposit Windfalls and Bank Reporting Quality: Evidence from Shale Booms.". Presented at the Paris December 2018 Finance Meeting EUROFIDAI - AFFI. 2017 .