Session 397

Beyond the Longitudinal--Panel Analysis

Track R

Date: Tuesday, September 23, 2014

 

Time: 11:00 – 12:15

Paper

Room: Estancia 311


Session Chair:

  • Aaron Hill, Oklahoma State University

Title: A General Approach to Panel Data Set-Theoretic Research

Authors

  • Roberto Garcia-Castro, IESE Business School
  • Miguel A. Ariño, IESE Business School

Abstract: Management research based on general linear statistical models has been rapidly moving toward a greater and richer use of longitudinal (panel data) econometric methods able to cope with critical issues such as endogeneity and reverse causality. By contrast, set-theoretic empirical research in management, despite its growing diffusion, has been solely focused on cross-sectional analysis to date. This article covers this void in longitudinal set-theoretic research. We provide a general framework in which consistency and coverage can be assessed both cross-sectionally and across time. The suggested approach is based on the distinction between pooled, between and within consistency and coverage, which can be computed using panel data. We use KLD’s panel (1991–2005) to illustrate how this approach can be applied in the context of longitudinal research.

Title: Causal Inferences in Small Samples using Synthetic Control Methodology: Did Chrysler Benefit from Government Assistance?

Authors

  • Guy Holburn, University of Western Ontario
  • Adam Fremeth, University of Western Ontario

Abstract: We introduce synthetic control analysis to management research. This recently developed statistical methodology overcomes challenges to causal inference in contexts constrained by small samples or few occurrences of the phenomenon of interest. Synthetic control constructs a replica of a focal firm based on a weighted combination of untreated firms with similar attributes within the sample population. The method quantifies the magnitude and direction of a treatment effect by comparing the actual performance of a focal unit to its counterfactual replica without treatment. As an illustration, we assess the impact of government intervention in the auto sector on the performance of Chrysler which, following the financial crisis, accepted government support in return for Treasury oversight. The synthetic Chrysler we construct sold 29% more vehicles in the U.S. than did the actual firm during the intervention period.

Title: Co-location and Performance: Learning and Resource Sharing in US hotels 1977-2007

Authors

  • Robert Seamans, New York University
  • Evan Rawley, Columbia University

Abstract: Within-firm effects of co-location are important, but understudied. We know little about why co-location matters—whether the benefits are primarily derived from one-time learning (stock) effects or ongoing resource sharing (flows). We study co-location in the context of the U.S. hotel industry 1977-2007 and show how positive and negative treatment effects can be used to disentangle learning from resource sharing effects. We find that co-location improves productivity in existing hotels by approximately 3.5%, with two-thirds of the effect due to one-time learning effects. The effect is particularly strong when firms add hotels in markets where their competitors are making productivity advances. The results suggest that new establishments drive positive within-firm co-location effects by teaching existing establishments new approaches and techniques that they would otherwise miss.

Title: Examining the Influence of Endogeneity when Testing Interactions

Authors

  • Michael Withers, Texas A&M University
  • Trevis Certo, Arizona State University
  • Matthew Semadeni, Arizona State University

Abstract: We use simulations to examine how endogeneity influences the testing of hypothesized interactions. Our results indicate that ordinary least squares (OLS) regression reports unbiased coefficient estimates for interactions, even when the independent and/or moderator variable is endogenous. In contrast, our simulations indicate that two-stage least squares (2SLS) approaches rarely achieve the statistical power necessary to detect statistically significant relationships when testing interactions. Although OLS regression produces unbiased coefficient estimates for interaction terms, endogeneity biases the coefficient estimates for the main effects of the components of the interactions (i.e., independent and/or moderator variables). Based on our results we provide a series of recommendations for researchers investigating interactions.

All Sessions in Track R...

Sun: 08:00 – 09:15
Session 253: Research Methods and Publishing, and Publishing Research Method Advances
Sun: 11:15 – 12:30
Session 310: Assessing the Broad Impact of Research
Mon: 08:00 – 09:15
Session 448: Looking to the Past--Going toward the Future
Mon: 14:45 – 16:00
Session 395: Beyond the Known Methods--Introduction to Emerging Techniques
Mon: 16:30 – 17:45
Session 369: The Future of Research Methods in Strategic Management Research
Tue: 11:00 – 12:15
Session 397: Beyond the Longitudinal--Panel Analysis
Tue: 17:15 – 18:30
Session 450: Looking for Truth--Curves in Research


Strategic Management Society

Madrid