Advertisement

RBT: Vote Counting and Meta-Analysis

Abstract

Following Godfrey and Hill (1995), RBT must survive rigorous empirical testing before one could refer to it as a sufficiently supported theory.422 After having analyzed the operationalizations of the theory’s central constructs and propositions within the 192 empirical papers, the following chapter thus emphasizes the empirical corroboration of RBT in terms of its overall statistical significance regarding the empirical results. Accordingly, chapter 4 further attends to the second deficit identified within this dissertation: the lack of understanding towards the empirical validation of RBT is addressed by integrating the empirical results.

Keywords

Firm Performance Intangible Asset Financial Capital Organizational Capability Marketing Capability 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 422.
    Cf. Godfrey/ Hill (1995), p. 527.Google Scholar
  2. 423.
    Cf. Fricke/ Treinies (1985), p. 12ff.Google Scholar
  3. 424.
    Cf. Hunter/ Schmidt (1990), p. 468; Fricke/Treinies (1985), p. 12; Stamm/Schwab (1995), p. 7.Google Scholar
  4. 425.
    Cf. Fricke/ Treinies (1985), p. 66; Hedges/Olkin (1980), p. 367; Hunter/Schmidt (1990), p. 469ff.Google Scholar
  5. 426.
    Glass (1976), p. 3.Google Scholar
  6. 427.
    Cf. Schmidt (1992), p. 1174ff; Rustenbach (2003), p. 8f.Google Scholar
  7. 428.
    Cf. Knott (2003).Google Scholar
  8. 429.
    Cf. Henderson/ Cockburn (1994).Google Scholar
  9. 430.
    Cf. Hall (1992).Google Scholar
  10. 431.
    Cf. DeCarolis/ Deeds (1999).Google Scholar
  11. 432.
    Cf. Schoenecker/ Cooper (1998). Their work is thus within the “...stream of research that considers how firm resources affect strategic decisions.” Schoenecker/Cooper (1998), p. 1128.Google Scholar
  12. 433.
    Cf. Gulati (1999).Google Scholar
  13. 434.
    Cf. Hitt et al. (2000).Google Scholar
  14. 435.
    Cf. Robins/ Wiersema (1995).Google Scholar
  15. 436.
    Cf. Daily et al. (2000).Google Scholar
  16. 437.
    Cf. Markides/ Williamson (1996).Google Scholar
  17. 438.
    Cf. Bergh (1995).Google Scholar
  18. 439.
    Mosakowski (1993), p. 822.Google Scholar
  19. 440.
    Cf. Zander/ Kogut (1995).Google Scholar
  20. 441.
    Cf. Helfat (1997).Google Scholar
  21. 445.
    Cf. Markman et al. (2004).Google Scholar
  22. 446.
    Cf. DeCarolis/ Deeds (1999). Besides these results, the authors also found supportive significant relationships for stocks; those were operationalized using the number of products in development (total number of products in each of the significant stages of the pharmaceutical testing process) and citation data (the Science Citation Index according to the names of full-time top scientists employees).Google Scholar
  23. 447.
    Cf. Harrison et al. (1991), p. 175.Google Scholar
  24. 454.
    Cf. Delios/ Beamish (1999); Dhanaraj/Beamish (2003); Yeoh/Roth (1999); DeCarolis (2003).Google Scholar
  25. 455.
    Cf. DeCarolis/ Deeds (1999); Yeoh/Roth (1999); DeCarolis (2003).Google Scholar
  26. 456.
    Hoopes et al. (2003) acknowledge this problem, pointing out that “...disconfirming the RBT is difficult. The theory is supported by any evidence that interfirm variation in resources and capabilities creates sustainable performance differences. Yet, evidence to the contrary indicates only that the resources or capabilities examined lack value.” Hoopes et al. (2003), p. 891.Google Scholar
  27. 457.
    E.g., Knott (2003); DeCarolis/Deeds (1999); McEvily/Chakravarthy (2002); Miller/Shamsie (1996).Google Scholar
  28. 458.
    Cf. Fricke/ Treinies (1985), p. 12ff. Glass, who was the first scholar to refer to the concept of meta-analysis, asserts that “meta-analysis refers to the analysis of analyses. I use it to refer to the statistical analysis results of a large collection of analysis results from individual studies for the purpose of integrating the findings.” Glass (1976), p. 3.Google Scholar
  29. 459.
    Cf. Barrilleaux-Pizzolatto/ Chhokar (1985), p. 15; Eden (2002), p. 841; Schmidt (1992), p. 1174ff; Rustenbach (2003), p. 8f; Lam/Kennedy (2005), p. 168.Google Scholar
  30. 460.
    Cf. Fricke/ Treinies (1985), p. 18ff; Hedges/Olkin (1992); Hunter et al. (1982); Hunter/Schmidt (1990). For further details on different descriptive meta-analyses methods see Hunter/Schmidt (1990), p. 478ff.Google Scholar
  31. 461.
    Cf. Rosenthal (1991).Google Scholar
  32. 462.
    Cf. Cooper (1982).Google Scholar
  33. 463.
    Cf. Cooper (1982), p. 291ff; and similar Stamm/Schwarb (1995), p. 13ff.Google Scholar
  34. 464.
    Cf. Lipsey/ Wilson (2001), p. 12.Google Scholar
  35. 465.
    Cf. Cohen (1977), p. 9f; Breaugh (2003), p. 80.Google Scholar
  36. 466.
    Cf. Whitener (1990), p. 315; Stamm/Schwab (1995), p. 12.Google Scholar
  37. 467.
    Cf. Rasmussen/ Loher (1988), p. 685.Google Scholar
  38. 470.
    Cf. Lipsey/ Wilson (2001), p. 16.Google Scholar
  39. 472.
    Cf. Lipsey/ Wilson (2001), p. 73.Google Scholar
  40. 473.
    Cf. Hunter/ Schmidt (2004), p. 472.Google Scholar
  41. 474.
    Cf. Hunter/ Schmidt (2004), p. 470.Google Scholar
  42. 475.
    Cf. Fern/ Monroe (1996), p. 90; Breaugh (2003), p. 80ff.Google Scholar
  43. 476.
    Cf. Lipsey/ Wilson (2001), p. 37ff. For a review on the history of effect sizes, see Huberty (2002).Google Scholar
  44. 477.
    See Rosnow et al. (2000), p. 446ff and Fern/Monroe (1996), p. 94 for more details.Google Scholar
  45. 478.
    Cf. Rosenthal (1991), p. 14f; Rosenthal/DiMatteo (2001), p. 72.Google Scholar
  46. 479.
    See Lipsey/ Wilson (2001), p. 189ff and Rosenthal/DiMatteo (2001), p. 72 for more details on formulas for calculating effect sizes from a range of statistical data.Google Scholar
  47. 480.
    Cf. Lipsey/ Wilson (2001), p. 70.Google Scholar
  48. 481.
    Lipsey and Wilson (2001), for instance, suggest two different approaches: (1) conduct a combined significance test analysis to determine of whether the aggregate data showed a significant effect (rejection of the null hypothesis); or (2) conduct a ‘limited’ analysis of the standardized regression coefficients, calculating the median regression coefficient for each independent variable of interest. Cf. Lipsey/Wilson (2001), p. 67f. Peterson and Brown (2005) conduct a meta-analysis using standardized regression coefficients, i.e., more than 1,700 corresponding beta coefficients and correlation coefficients from published studies. Their results indicate that under certain conditions, using knowledge of corresponding beta coefficients to impute missing correlations (effect sizes) generally produces relatively accurate and precise population effect-size estimates. Based on the relationship between beta and r, the authors imputed corresponding correlation coefficients when only beta coefficients were reported. Cf. Peterson/Brown (2005), p. 175ff.Google Scholar
  49. 482.
    Cf. Hunter/ Schmidt (2004), p. 476f.Google Scholar
  50. 483.
    Cf. Cheung/ Chan (2004), p. 780; Barrilleaux-Pizzolatto/Chhokar (1985), p. 17. For further discussions on the degree of interdependence and possible solutions in this respect, see Cheung/Chan (2004).Google Scholar
  51. 484.
    Cf. Hunter/ Schmidt (1990), p. 451ff.Google Scholar
  52. 485.
    Cf. Hunter/ Schmidt (1990), p. 463ff.Google Scholar
  53. 487.
    Cf. Cheung/ Chan (2004), p. 782.Google Scholar
  54. 488.
    Cf. Lipsey/ Wilson (2001), p. 73ff; Rustenbach (2003), p. 41ff.Google Scholar
  55. 490.
    Cf. Cortina (2003), p. 416.Google Scholar
  56. 491.
    Cf. Hunter/ Schmidt (2004), p. 76.Google Scholar
  57. 492.
    Cf. Schmidt et al. (1985), p. 697ff.Google Scholar
  58. 493.
    Cf. Hunter/ Schmidt (2004), p. 76.Google Scholar
  59. 494.
    Cf. Hunter/ Schmidt (2004), p. 81. Hunter et al. (1982) assert that the sample size weighting impedes distortions. The authors decline the request for transforming z-values for mean correlation coefficients that result in negative distortions, because there would be bigger positive distortions when using z-values. Cf. Hunter et al. (1982), p. 40; Hunter/Schmidt (2004), p. 82. For an empirical study on the qualification of the population coefficient estimation formula see Cornwell/Ladd (1993).Google Scholar
  60. 495.
    Cf. Hunter/ Schmidt (2004), p. 81.Google Scholar
  61. 496.
    Cf. Hunter/ Schmidt (2004), p. 84.Google Scholar
  62. 497.
    I.e., the sampling error variance across studies equals the average of the sampling error variance within each individual study. Cf. Hunter/ Schmidt (2004), p. 85ff.Google Scholar
  63. 498.
    Cf. Hunter/ Schmidt (2004), p. 87.Google Scholar
  64. 499.
    Cf. Viswesvaran/ Sanchez (1998), p. 78f.Google Scholar
  65. 500.
    Hunter and Schmidt object to use χ2-Tests; they argue that with large sample-sizes, those tests produce significant results even with trivial variances, which then leads to rejecting H0 of a homogenous population. Cf. Hunter/ Schmidt (2004), p. 401. See also Viswesvaran/Sanchez (1998) and Cortina (2003) for more details on moderator search within meta-analyses.Google Scholar
  66. 501.
    Cf. Whitener (1990), p. 317. Koslowsky and Sagie’s (1993) empirical study on credibility intervals revealed that the distinguishing value between small and large intervals is about 0.11. Cf. Koslowsky/Sagie (1993), p. 698.Google Scholar
  67. 502.
    Cf. Hunter/ Schmidt (2004), p. 401; Koslowsky/Sagie (1993), p. 695; Koslowsky/Sagie (1994), p. 562; and Schmidt et al. (1988) as well as Rasmussen/Loher (1988) for more details.Google Scholar
  68. 503.
    Cf. Hunter/ Schmidt (2004), p. 90; Stamm/Schwarb (1995), p. 14; Viswesvaran/Sanchez (1998), p. 79f.Google Scholar
  69. 504.
    Cf. Hunter/ Schmidt (1990), p. 437f.Google Scholar
  70. 505.
    Cf. Lipsey/ Wilson (2001), p. 115; Whitener (1990), p. 317; Hunter/Schmidt (1990), p. 426ff.Google Scholar
  71. 506.
    Cf. Whitener (1990), p. 317.Google Scholar
  72. 507.
    Cf. Rosenthal (1979), p. 638ff; Lipsey/Wilson (2001), p. 165f.Google Scholar
  73. 508.
    Cf. Roth et al. (1996), p. 552f; Stamm/Schwarb (1995), p. 16ff.Google Scholar
  74. 509.
    Cf. Huffcutt/ Arthur (1995), p. 327. Similar Cohen (1990), p. 1305; Roth et al. (1996), p. 552.Google Scholar
  75. 510.
    Cf. Lipsey/ Wilson (2001), p. 108.Google Scholar
  76. 511.
    Cf. Rustenbach (2003), p. 251ff for a detailed description of presenting integrative results.Google Scholar
  77. 512.
    Cf. Lipsey/ Wilson (2001), p. 146; Cohen (1988), p. 25ff, p. 77ff and p. 284ff.Google Scholar
  78. 513.
    Cf. Eden (2002), p. 841; Barrilleaux-Pizzolatto/Chhokar (1985), p. 16.Google Scholar
  79. 514.
    Cf. Chambers (2004), p. 35.Google Scholar
  80. 515.
    Cf. Leviton/ Cook (1981), p. 233.Google Scholar
  81. 516.
    Cf. Lipsey/ Wilson (2001), p. 7.Google Scholar
  82. 517.
    Cf. Green/ Hall (1984), p. 40.Google Scholar
  83. 518.
    Cf. Eden (2002), p. 841. Eden suggests, for example, the variability of the relationship between two variables or moderator effects once the variance owing to artifacts has been removed.Google Scholar
  84. 519.
    Cf. Lipsey/ Wilson (2001), p. 6; Eden (2002), p. 841.Google Scholar
  85. 520.
    Cf. Schmidt (1992), p. 1178.Google Scholar
  86. 521.
    Cf. Hunter et al. (1982), p. 19ff.Google Scholar
  87. 522.
    Cf. Cohen (1990), p. 1308; Lipsey/Wilson (2001), p. 6.Google Scholar
  88. 523.
    Cf. Cohen (1990), p. 1310.Google Scholar
  89. 524.
    Cf. Rosenthal/ DiMatto (2001), p. 64.Google Scholar
  90. 525.
    Cf. Eden (2002), p. 843f.Google Scholar
  91. 526.
    Cf. Green/ Hall (1984), p. 40f.Google Scholar
  92. 527.
    Cf. Lipsey/ Wilson (2001), p. 9; Stamm/Schwarb (1995), p. 21; Barrilleaux-Pizzolatto/Chhokar (1985), p. 18. The present meta-analysis primarily relies on studies from high quality journals that have undergone an intensive screening process. Nonetheless, a moderator analysis will be provided.Google Scholar
  93. 528.
    Cf. Lipsey/ Wilson (2001), p. 71; Barrilleaux-Pizzolatto/Chhokar (1985), p. 16; Stamm/Schwarb (1995), p. 21; Lam/Kennedy (2005), p. 169; Chambers (2004), p. 36.Google Scholar
  94. 529.
    Cf. Green/ Hall (1984), p. 47. For example, Green and Hall (1984) refer to a study which integrated 345 primary tests. To question their significance, one would need approximately 65,000 non-significant results.Google Scholar
  95. 530.
    Cf. Stamm/ Schwarb (1995), p. 22.Google Scholar
  96. 531.
    Cf. Barrilleaux-Pizzolatto/ Chhokar (1985), p. 19.Google Scholar
  97. 532.
    Chambers (2004), p. 36. Chambers refers to two studies, e.g., Bangert-Drowns et al. (1985) who located around 500 articles but over 90% were not suitable as well as Gillingham and Guthrie (1987) who reported that their sample could just use 6% of the literature found due to the inability to calculate effects from the lack of data provided. Cf. Chambers (2004), p. 36.Google Scholar
  98. 533.
    Cf. Stamm/ Schwarb (1995), p. 20f; Barrilleaux-Pizzolatto/Chhokar (1985), p. 18; Chambers (2004), p. 36; Rosenthal/DiMatto (2001), p. 68.Google Scholar
  99. 534.
    Cf. Chambers (2004), p. 36.Google Scholar
  100. 535.
    Smith et al. (1980), p. 47.Google Scholar
  101. 536.
    Cf. Stamm/ Schwarb (1995), p. 21; Barrilleaux-Pizzolatto/Chhokar (1985), p. 18.Google Scholar
  102. 537.
    Cf. Chambers (2004), p. 36.Google Scholar
  103. 538.
    Cf. Stamm/ Schwarb (1995), p. 20; Rosenthal/DiMatto (2001), p. 67. These studies can be detected by their sample descriptions.Google Scholar
  104. 539.
    Cf. Lipsey/ Wilson (2001), p. 7.Google Scholar
  105. 540.
    Cf. Onwuegbuzie et al. (2003), p. 38.Google Scholar
  106. 541.
    Barrilleaux-Pizzolatto/ Chokar (1985), p. 20.Google Scholar
  107. 544.
    Koslowsky and Sagie (1993) explicate that relative percentages of variance explained by each one of the correctable artifacts exists. They assert that 80% of total artifactual variance in a study can be ascribed to sampling error variance. Therefore, the authors suggest that in the case where meta-analysis only accounts for sampling error, one could (should) lower the cutoff point for determining the existence of moderators from 75% to 60% of the sample correlation variance (implying that measurement error and range restriction make up 15% of the total sample variance). Cf. Koslowsky/ Sagie (1994), p. 562f.Google Scholar
  108. 548.
    Cf. McGahan/ Porter (1997), p. 23. Segement-specific effects encompassed differences between business-segments, e.g., differences in market share, differentiation, organizational processes, heterogeneity in fixed assets, etc.Google Scholar
  109. 549.
    Cf. Wu (2006), p. 167.Google Scholar
  110. 550.
    Cf. Sundaramurthy et al. (2005), p. 502.Google Scholar
  111. 551.
    Cf. Geykens et al. (2006), p. 525.Google Scholar
  112. 552.
    Sharma/ Vredenburg (1998), p. 735.Google Scholar
  113. 553.
    Spanos/ Lioukas (2001), p. 915.Google Scholar
  114. 555.
    Cf. Ray et al. (2004), p. 24.Google Scholar

Copyright information

© Gabler | GWV Fachverlage GmbH, Wiesbaden 2008

Personalised recommendations