Methodological Conception
Chapter
Abstract
Data collection in the field is indispensable to the progress of management accounting research.^{389} There exists a variety of research methods and procedures to initially approach the present research problems and subsequently collect, analyze, and interpret the empirical data. According to CRESWELL (2003), for designing a research strategy, three interrelated framework elements need to be considered that lead to a research design, which tends to be more quantitative or qualitative in nature:
In the following, this framework will be employed to design a strategy for the present research.

• The assumptions about knowledge claims made by the researcher, i.e., what will be learned during the research and how will that be achieved?

• The strategies of inquiry, i.e., the general research procedures employed.

• The concrete methods for data collection, analysis, and interpretation.^{390}
Keywords
Hierarchy Level German Translation Methodological Conception Management Accounting Research Formative Measurement Model
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.
References
 389.Cf. Modell (2005), p. 231; Kaplan (1986), p. 429.Google Scholar
 390.In the following cf. Creswell (2003), pp. 3–21.Google Scholar
 391.Cf. Patton (1997), p. 267.Google Scholar
 392.Cf. Denzin/ Lincoln (2005), p. 183. The authors further include axiology, which asks: “How will I be as a moral person in the world?” Denzin/Lincoln (2005), p. 183. As this aspect does not refer to the actual research problem, it will not be included in the subsequent analysis.Google Scholar
 393.Cf. Denzin/ Lincoln (2005), p. 21.Google Scholar
 395.Cf. Denzin/ Lincoln (2005), pp. 183f.Google Scholar
 399.Cf. Creswell (2003), p. 8.Google Scholar
 400.Greenwood/ Levin (2000), p. 92.Google Scholar
 401.Kaplan (1964), p. 36.Google Scholar
 402.Cf. Smith (2003), p. 5.Google Scholar
 403.Here and in the following cf. Creswell (2003), pp. 7f.Google Scholar
 404.Creswell (2003), p. 7.Google Scholar
 405.Cf. Cherryholmes (1992), p. 13. One of the first descriptions of pragmatism appears in 1905 by Peirce: “The word pragmatism was invented to express a certain maxim of logic. […] The maxim is intended to furnish a method for the analysis of concepts. […] The method prescribed in the maxim is to trace out in the imagination the conceivable practical consequences — that is, the consequences for deliberate, selfcontrolled conduct — of the affirmation or denial of the concept.” Peirce (1905), p. 494.Google Scholar
 406.Kaplan (1964), p. 42.Google Scholar
 407.Creswell (2003), p. 12.Google Scholar
 409.Cf. Creswell (2003), p. 153; Kaplan/Duchon (1988), p. 572.Google Scholar
 410.Cf. Kaplan/ Duchon (1988), p. 572.Google Scholar
 411.Cf. Creswell (2003), p. 14.Google Scholar
 412.Smith (2003), p. 100.Google Scholar
 413.Cf. Smith (2003), p. 100; Birnberg/Shields/Young (1990), p. 35. Internal validity describes “[…] the credibility of the causal relationships between independent and dependent variables inferred from data.” Modell (2005), p. 236.Google Scholar
 414.Cf. Smith (2003), p. 103. External validity describes the degree to which the empirical findings of a specific research study can be generalized for different populations and contexts. Cf. Modell (2005), p. 234.Google Scholar
 415.Each of these data gathering methods has its own advantages and disadvantages. A more detailed description about the possibilities and drawbacks of conducting survey research is provided by Smith (2003), pp. 117–120.Google Scholar
 416.Birnberg/ Shields/ Young (1990), p. 38.Google Scholar
 417.Kaplan/Duchon (1988), p. 573.Google Scholar
 418.Creswell (2003), p. 14.Google Scholar
 419.Miller/ Hengst/ Wang (2003), p. 223.Google Scholar
 420.Miller/ Hengst/ Wang (2003), p. 224.Google Scholar
 421.Cf. Creswell (2003), p. 14.Google Scholar
 422.Yin (2003), p. 13; further cf. Schäffer/Brettel (2005), p. 43.Google Scholar
 423.Cf. Creswell (2003), p. 181.Google Scholar
 424.Kaplan/ Duchon (1988), p. 573.Google Scholar
 425.Cf. Creswell (2003), p. 15. Mixed strategies of inquiry have further been labeled multimethod or combined strategies of inquiry. Cf. Birnberg/Shields/Young (1990), p. 33.Google Scholar
 426.Here and in the following cf. Creswell (2003), pp. 210–213.Google Scholar
 428.Smith (2003), p. 103.Google Scholar
 429.Yin (2003), p. 13.Google Scholar
 430.Cf. Mayring (2001), p. 10.Google Scholar
 431.Birnberg/ Shields/ Young (1990), p. 33. Various researchers have recently made similar calls to combine quantitative with qualitative research. Cf. Modell (2005), pp. 233f.; Mayring (2001), pp. 10f.Google Scholar
 432.Cf. Morse (1991), pp. 121f.Google Scholar
 434.Denzin/ Lincoln (2005), p. 183.Google Scholar
 436.Cf. Bollen (1989), p. 4 and the literature cited there.Google Scholar
 437.Homburg/ Hildebrandt (1998), p. 17.Google Scholar
 438.SEM is generally seen as an extension of firstgeneration methods such as multivariate regression or factor analysis. Loosening some of the more stringent assumptions and constraints of SEM would lead to a first generation method. Cf. Chin/ Newsted (1999), p. 308.Google Scholar
 439.Rigdon (1998), p. 251.Google Scholar
 440.Cf. Chin (1998b), p. 297.Google Scholar
 441.Chin/ Newsted (1999), p. 308.Google Scholar
 442.In contrast to SEM, regression analysis assumes that all independent variables are uncorrelated. If such dependencies are evident in the empirical data, the estimators of the regression parameters are generally distorted (problem of multicollinearity). Cf. Backhaus et al. (2006), pp. 89–92.Google Scholar
 443.Cf. chapter D4.1 for the differences between reflective and formative measurement models; further cf. Ittner/ Larcker/ Randall (2003), pp. 200–202.Google Scholar
 444.Cf. Churchill Jr./ Iacobucci (2005), p. 283; Bagozzi (1994), p. 331; Carmines/Zeller (1979), p. 26; Churchill Jr. (1979), p. 66.Google Scholar
 445.Drolet/ Morrison (2001), p. 198.Google Scholar
 446.“Backtranslation is typically used to verify semantic equivalence (SE) of a translated measure to the original scale.” Mallinckrodt/ Wang (2004), p. 368; further cf. Brislin (1980) on the details for conducting translationbacktranslation procedures.Google Scholar
 448.On the downside, the raw score on a variable only takes meaning when compared to some standard. For example, from a mean value of five on a sevenpoint Likert scale, one cannot infer that the value is high or low. Cf. Churchill Jr./ Iacobucci (2005), p. 276.Google Scholar
 450.Cf. Steiners (2005), pp. 102f.Google Scholar
 451.Cf. Steiners (2005), pp. 102f.Google Scholar
 452.Cf. Sager (1994), pp. 78f.; Mowday/Steers/Porter (1979).Google Scholar
 453.Winata/ Mia (2005), p. 29; further cf. chapter B1.1.3.Google Scholar
 454.Cf. Winata/ Mia (2005), p. 30.Google Scholar
 455.Cf. Chenhall (2003), p. 134 and the literature cited there.Google Scholar
 456.Brownell (1979), p. 63.Google Scholar
 457.Cf. Viator (2001), p. 81 and p. 90.Google Scholar
 458.Cf. Kalbers/ Fogarty (1995), p. 83; Gregson/Wendell/Aono (1994), p. 152. 459 The transformation formula for the first indicator, measured on a sevenpoint Likert scale, was: (Value1)/6. For the second indicator, the transformation formula was: (Value1)/4.Google Scholar
 460.Cf. Hinkin/ Schriesheim (1989), p. 567.Google Scholar
 461.Cf. Raven/ Schwarzwald/ Koslowsky (1998), p. 330.Google Scholar
 462.Cf. Hinkin/ Schriesheim (1989), p. 567.Google Scholar
 463.Cf. Raven/ Schwarzwald/ Koslowsky (1998), p. 330.Google Scholar
 464.Cf. Raven/ Schwarzwald/ Koslowsky (1998), p. 330.Google Scholar
 465.Cf. Hinkin/ Schriesheim (1989), p. 567.Google Scholar
 466.Cf. Hinkin/ Schriesheim (1989), p. 567.Google Scholar
 467.Cf. Spector (1988), p. 340.Google Scholar
 468.Cf. Schwarzer/ Jerusalem (1999), p. 13. The reported indicators are based on the revised tenindicator version of a 20indicator construct originally developed in 1981 and published in 1986 by the same authors. Cf. Jerusalem/Schwarzer (1986).Google Scholar
 469.Brownell/ Dunk (1991), p. 695.Google Scholar
 470.Cf. Daft/ Macintosh (1981), p. 215; Withey/Daft/Cooper (1983), p. 59.Google Scholar
 471.Cf. Withey/ Daft/ Cooper (1983), p. 57.Google Scholar
 474.Cf. Kinnear/ Taylor (1991), p. 352; Hunt/Sparkman Jr./Wilcox (1982), pp. 265f.Google Scholar
 477.For the importance of cover letters cf. Diamantopoulos/ Schlegelmilch (1996), p. 523.Google Scholar
 478.Cf. Fox/ Crask/ Kim (1988), pp. 474f.Google Scholar
 482.The normal distribution assumption was tested with a KolmogorovSmirnov test for each indicator, as well as a test for skewness. Cf. Bortz (2005), pp. 74–77.Google Scholar
 483.Cf. McClave/ Benson/ Sincich(2005),pp. 1095f.Google Scholar
 484.A detailed description of the KW is provided by Sachs (2006), pp. 394–401.Google Scholar
 485.U describes the number of times a value in the first group precedes a value in the second group, when values are sorted in ascending order. Cf. Sachs (2006), pp. 381f.Google Scholar
 486.Cf. Sachs (2006), p. 382.Google Scholar
 487.The null hypotheses is rejected when the value U is smaller or equal to a critical value, which can be obtained from a table provided by, for example, Sachs (2006), p. 384.Google Scholar
 489.Podsakoff et al. (2003), p. 881; further cf. Hurrle/Kieser (2005), p. 596. These biases are also referred to as social desirability bias and selfserving attributions.Google Scholar
 490.Cf. Golden (1992), p. 438.Google Scholar
 491.Cf. Golden (1997), p. 1245.Google Scholar
 492.For the problems associated with differences in the functional background of key informants cf. Hurrle/ Kieser (2005), p. 595Google Scholar
 493.“Method variance refers to variance that is attributable to measurement method rather than to the construct of interest. The term method refers to the form of measurement at different levels of abstraction, such as the content of specific items, scale type, response format, and the general context […]. At a more abstract level, method effects might be interpreted in terms of response biases such as halo effects, social desirability, acquiescence, leniency effects, or yeaand naysaying.” Bagozzi/ Yi (1991), p. 426; for a detailed discussion of the topic cf. Podsakoff et al. (2003) and Podsakoff/Organ (1986).Google Scholar
 494.This is referred to as the consistency motif. Cf. Podsakoff et al. (2003), p. 881.Google Scholar
 495.Cf. Podsakoff et al. (2003), p. 888; Drolet/Morrison (2001), p. 201.Google Scholar
 496.Cf. Churchill Jr./ Iacobucci (2005), pp. 254–261; Podsakoff et al. (2003), p. 888.Google Scholar
 497.Cf. Armstrong/ Overton (1977), p. 397.Google Scholar
 498.Cf. Coviello/ Jones (2004), p. 493. A sample can be defined as a finite fraction of a statistical population for gaining an understanding of the whole. The sampling process refers to the process of selecting sample elements. Cf. Churchill Jr./Iacobucci (2005), pp. 322–324.Google Scholar
 499.Churchill Jr./ Iacobucci (2005), p. 324.Google Scholar
 500.Cf. Churchill Jr./ Iacobucci (2005), pp. 326–329; Galtung (1969), p. 56.Google Scholar
 501.Cf. Churchill Jr./ Iacobucci (2005), pp. 326f.Google Scholar
 502.Churchill Jr./ Iacobucci (2005), p. 328.Google Scholar
 503.Churchill Jr./ Iacobucci (2005), p. 327.Google Scholar
 504.Cf. Churchill Jr./ Iacobucci (2005), p. 328; Coviello/Jones (2004), p. 493.Google Scholar
 505.Cf. Chin/ Newsted (1999), p. 308.Google Scholar
 506.Rigdon (1998), p. 251.Google Scholar
 507.Cf. Tenenhaus et al. (2005), p. 166; Chin/Newsted (1999), p. 321.Google Scholar
 508.Anderson/ Gerbing (1982), p. 453.Google Scholar
 510.Own compilation following Ringle (2004), p. 281.Google Scholar
 511.Here and in the following cf. Fornell/ Cha (1994), p. 58.Google Scholar
 512.Cf. Chin/ Newsted (1999), p. 321.Google Scholar
 513.Chin/ Newsted (1999), p. 322.Google Scholar
 514.Cf. Diamantopoulos/ Winklhofer (2001), p. 274; further cf. chapter D2.Google Scholar
 515.Own compilation following Fornell/ Bookstein (1982), p. 441.Google Scholar
 516.Cf. Jarvis/ MacKenzie/ Podsakoff (2003), pp. 199f.; further cf. Nunnally (1978).Google Scholar
 517.It may, however, adversely affect construct reliability and/or lead to identification problems if the number of remaining indicators is too low. Cf. Bagozzi (1994), p. 331; Churchill Jr. (1979), p. 66.Google Scholar
 518.MacCallum/ Browne (1993), p. 533.Google Scholar
 519.Cf. Fornell/ Bookstein (1982), p. 441.Google Scholar
 520.Bollen/ Lennox (1991), p. 308; further cf. Jarvis/MacKenzie/Podsakoff (2003), p. 202.Google Scholar
 521.Nunnally/ Bernstein (1994), p. 489.Google Scholar
 522.Here and in the following cf. Fornell/ Cha (1994), p. 59. Exogenous and endogenous LVs are operationalized via manifest indicators, whose error terms are designated with the parameters δ, ε, and ξ.Google Scholar
 523.Cf. Chin (1998b), p. 307.Google Scholar
 524.Cf. Fornell/ Bookstein (1982), p. 442.Google Scholar
 525.Cf. Jarvis/ MacKenzie/ Podsakoff (2003), p. 212; Galtung (1969), p. 281.Google Scholar
 526.Fornell/ Bookstein (1982), p. 442; further cf. Rossiter (2002), p. 316.Google Scholar
 528.Therefore, different algorithms like Maximum Likelihood, Generalized Least Squares, Weighted Least Squares, or Unweighted Least Squares exist. Most of them lead to consistent estimators. Only the estimators of the Unweighted Least Sqares algorithm tend to be less efficient. It therefore requires a lower sample size as compared to the other algorithms. Cf. Rigdon (1998), p. 265.Google Scholar
 529.Chin (1998b), p. 299.Google Scholar
 530.Cf. Chin (1998b), p. 297.Google Scholar
 531.Cf. Dillon/ Kumar/ Mulani (1987), p. 128.Google Scholar
 532.Cf. Fornell/ Bookstein (1982), p. 440.Google Scholar
 534.Wold (1982a), p. 200; further cf. Wold (1982b); Wold (1982c); Wold (1975).Google Scholar
 535.Chin (1998b), p. 301.Google Scholar
 536.A description of the formal PLS model specification is provided by Chin/ Newsted (1999), pp. 321–326.Google Scholar
 537.In the following cf. Chin (1998b), pp. 301f.Google Scholar
 538.Chin/ Marcolin/ Newsted (2003), p. 197.Google Scholar
 539.Chin (1998b), p. 302.Google Scholar
 540.Cf. Chin/ Marcolin/ Newsted (2003), p. 197.Google Scholar
 541.Chin (1998b), p. 303.Google Scholar
 542.Cf. Chin (1998b), p. 295.Google Scholar
 543.Cf. Chin/ Newsted (1999), p. 312.Google Scholar
 544.Cf. Chin/ Newsted (1999), pp. 335f.; Chin (1998b), p. 311.Google Scholar
 546.Cf. Chin/ Newsted (1999), p. 314; Mueller (1996), p. 26 and p. 57.Google Scholar
 547.Cf. Chin/ Marcolin/ Newsted (2003), p. 197.Google Scholar
 548.Fornell/ Bookstein (1982), p. 450.Google Scholar
 549.Cf. Chin/ Newsted (1999), p. 328.Google Scholar
 551.Cf. Viswanathan (2005), p. 3.Google Scholar
 552.Cf. Churchill Jr./ Iacobucci (2005), p. 291.Google Scholar
 553.Cf. Churchill Jr. (1979), pp. 68–72. For the evaluation criteria of formative measurement models cf. Götz/LiehrGobbers (2004), pp. 728–730.Google Scholar
 554.Viswanathan (2005), p. 18.Google Scholar
 555.Nunnally (1978), p. 93Google Scholar
 556.Cf. Tenenhaus et al. (2005), p. 163; Götz/LiehrGobbers (2004), p. 727; Weber/Willauer/Schäffer (2003), p. 374.Google Scholar
 557.For a detailed description cf. Backhaus et al. (2006), pp. 259–336; Hüttner/Schwarting (1999).Google Scholar
 558.Cf. Hüttner/ Schwarting (1999), p. 397. In contrast, the varimax rotation assumes linear independency between factors. This restriction is considered inapt for this dissertation. However, “[…] if identification of basic structuring of variables into theoretically meaningful subdimensions is the primary concern of the researcher, as is often the case in an exploratory factor analysis, almost any readily available method of rotation will do the job.” Kim/Mueller (1978), p. 50.Google Scholar
 559.Cf. Kim/ Mueller (1978), p. 43.Google Scholar
 560.Cf. Backhaus et al. (2006), p. 295. The eigenvalues do not represent the percentage of variance explained, but are measures for the amount of variance of the factor in relation to total variance.Google Scholar
 561.Cf. Homburg/ Giering (1998), p. 128.Google Scholar
 562.Cf. Hulland (1999), p. 198; Hoyle (1999), p. 327.Google Scholar
 563.Cf. Hulland (1999), p. 198.Google Scholar
 564.A detailed description of the two approaches is provided by Chin (1998b), pp. 318–320.Google Scholar
 565.Cf. Efron/ Gong (1983), p. 38. “Bootstrapping produces samples consisting of the same number of units as in the original sample. The number of resamples has to be specified. The default is 100 but a higher number (such as 200) may lead to more reasonable standard error estimates.” Tenenhaus et al. (2005), p. 176.Google Scholar
 566.Cf. Homburg/ Giering (1998), p. 125. Statistically significant on the one (five) percent significance level implies an average risk of one (five) in a hundred of dismissing an accurate hypothesis. Cf. Haller/Krauss (2002), p. 2; Schneider/Kornrumpf/Mohr (1993), p. 246.Google Scholar
 567.Cf. Fornell/ Larcker (1981), p. 45.Google Scholar
 568.Cf. Fornell/ Larcker (1981), p. 45; further cf. Bagozzi/Yi (1988), p. 82 for the advantages of the internal consistency measure.Google Scholar
 569.Cf. Nunnally (1978), p. 245; Bagozzi/Yi (1988), p. 80.Google Scholar
 570.Here and in the following cf. Fornell/ Larcker (1981), pp. 45f.Google Scholar
 571.Cf. Bagozzi/ Yi (1988), p. 81; Fornell/Larcker (1981), p. 46.Google Scholar
 572.Hulland (1999), p. 199.Google Scholar
 573.Cf. Hulland (1999), p. 200.Google Scholar
 574.Cf. Backhaus et al. (2006), p. 97; Chin (1998b), p. 323.Google Scholar
 575.Chin (1998a), p. xii; further cf. Lohmöller (1989), pp. 60f.Google Scholar
 576.Cf. Homburg/ Giering (1998), p. 125.Google Scholar
 577.Cf. Chin (1998b), p. 316.Google Scholar
 578.Cf. Cohen (1988), p. 413.Google Scholar
 580.Geisser (1975), p. 320.Google Scholar
 581.Chin (1998b), p. 317.Google Scholar
 582.Cf. Fornell/ Cha (1994), p. 64.Google Scholar
 583.Chin (1998b), p. 317. However, as discussed above, jackknifing standard errors are disregarded in this research. Instead, the more accurate bootstrapping estimates are used.Google Scholar
 584.Chin (1998b), p. 317.Google Scholar
 585.Cf. Götz/ LiehrGobbers (2004), p. 731.Google Scholar
 586.Chin (1998b), p. 318.Google Scholar
 587.Chin (1998b), p. 318. Q^{2} values were calculated using Visual PLS, as SmartPLS V2.0 M2 did not offer these applications.Google Scholar
 588.Cf. Chin (1998b), p. 318.Google Scholar
 589.Cf. Shrout/ Bolger (2002), p. 422 and p. 425.Google Scholar
 591.Cf. Baron/ Kenny (1986), p. 1177. More specifically, path c’ refers to the relationship between variables X and Y in an alternative model that has no mediating variable M.Google Scholar
 592.Cf. Iacobucci/ Duhachek (2003), p. 7.Google Scholar
 593.Cf. Sobel (1982), pp. 292–299.Google Scholar
 595.MacKinnon/ Krull/ Lockwood (2000), p. 174.Google Scholar
 596.Shrout/ Bolger (2002), p. 434.Google Scholar
 598.Own compilation following Eggert/ Fassott/ Helm (2005), p. 107.Google Scholar
 599.Cf. Chin/ Marcolin/ Newsted (2003), p. 199. Standardizing should be used when no emphasis is given exante to any specific indicator, which is the case in the present research. In contrast, centering only sets product indicators to a mean of one. Centering should only be used to perpetuate the scale metric or if some indicators are deemed more important than others.Google Scholar
 601.Cf. Baron/ Kenny (1986), p. 1174.Google Scholar
 602.Cf. Chin/ Marcolin/ Newsted (2003), p. 195.Google Scholar
 603.Cf. Cohen (1988), p. 413.Google Scholar
 604.Chin/ Marcolin/ Newsted (2003), p. 211.Google Scholar
Copyright information
© Deutscher UniversitätsVerlag  GWV Fachverlage GmbH, Wiesbaden 2008