attributes of articles impacting contemporary accounting literature

24
Attributes of articles impacting contemporary accounting literature* LAWRENCE D. BROWN State University of New York at Buffalo lOHN C. GARDNER State University of New York at Buffalo MIKLOS A. VASARHELYI AT&T Bell Laboratories Abstract. This paper describes and analyzes four attribute dimensions that have impacted contemporary accounting literature (CAL) between 1976 and 1984, and develops a model that predicts attribute levels in 1985 and 1986. The attribute dimensions studied are: accounting area, research method, school of thought, and geographical focus. Publication counts and citation analyses are performed on a data set of 1,110 accounting articles. The results suggest that linear trends exist over time in the publication and citation measures of selected attributes of accounting papers; that new or emerging attribute areas are more likely to be influential than are papers published in established areas; and that it is easier to predict the relative influence of publications that will exhibit certain attributes than it is to predict the number of papers that will be published with these attributes. Risumi. Les auteurs d6crivent et analysent quatre aspects de preoccupation qui ont eu des repercussions sur les publications contemporaines en comptabilit^ entre 1976 et 1984 et crdent un module permettant de pr6voir les niveaux de pr6occupation des ann6es 1985 et 1986. Les aspects de preoccupation faisant l'objet de l'etude sont les suivants: le domaine comptable, les mfithodes de recherche, les 6coles de pens^e et I'optique geographique. Les auteurs procMent au d6nombrement des publications et h. l'andyse des citations sur un corpus de 1110 articles comptables. Les r^sultats donnent ^ penser qu'il existe des tendances Iin6aires, dans le temps, dans les mestires des publications et des citations relatives ^ des preoccupations choisies dans les articles comptables; que les publications dans des domaines de preoccupation nouveaux ou en emergence sont plus susceptibles d'influencer que ne le sont les publications dans des domaines etablis; et qu'il est plus facile de predire I'influence relative des publications portant sur certaines preoccupations qu'il ne Test de preidire le nombre de documents qui seront publies relativement ^ ces preoccupations. A science which hesitates to forget its founders is lost. — Merton (1968, p. 1) Introduction In accounting, as in other disciplines in the social sciences, a discipline's jour- nals represent its principal research communication network. When knowledge * The authors are grateful for the assistance of Jee Jang and Georgiana Hsu of the University at Buffalo and Columbia University, respectively. We also appreciate the helpful comments of Messod Beneish, the two anonymous reviewers and the editor, Haim Falk. Contemporary Accounting Research Vol. 5 No. 2 pp. 793-815

Upload: lawrence-d-brown

Post on 30-Sep-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Attributes of articles impacting contemporary accounting literature

Attributes of articles impactingcontemporary accounting literature*

LAWRENCE D. BROWN State University of New York at Buffalo

lOHN C. GARDNER State University of New York at Buffalo

MIKLOS A. VASARHELYI AT&T Bell Laboratories

Abstract. This paper describes and analyzes four attribute dimensions that have impactedcontemporary accounting literature (CAL) between 1976 and 1984, and develops a modelthat predicts attribute levels in 1985 and 1986. The attribute dimensions studied are:accounting area, research method, school of thought, and geographical focus. Publicationcounts and citation analyses are performed on a data set of 1,110 accounting articles. Theresults suggest that linear trends exist over time in the publication and citation measures ofselected attributes of accounting papers; that new or emerging attribute areas are morelikely to be influential than are papers published in established areas; and that it is easier topredict the relative influence of publications that will exhibit certain attributes than it is topredict the number of papers that will be published with these attributes.

Risumi. Les auteurs d6crivent et analysent quatre aspects de preoccupation qui ont eudes repercussions sur les publications contemporaines en comptabilit^ entre 1976 et 1984et crdent un module permettant de pr6voir les niveaux de pr6occupation des ann6es 1985 et1986. Les aspects de preoccupation faisant l'objet de l'etude sont les suivants: le domainecomptable, les mfithodes de recherche, les 6coles de pens^e et I'optique geographique.Les auteurs procMent au d6nombrement des publications et h. l'andyse des citations surun corpus de 1110 articles comptables. Les r^sultats donnent ^ penser qu'il existe destendances Iin6aires, dans le temps, dans les mestires des publications et des citationsrelatives ^ des preoccupations choisies dans les articles comptables; que les publicationsdans des domaines de preoccupation nouveaux ou en emergence sont plus susceptiblesd'influencer que ne le sont les publications dans des domaines etablis; et qu'il est plusfacile de predire I'influence relative des publications portant sur certaines preoccupationsqu'il ne Test de preidire le nombre de documents qui seront publies relativement ^ cespreoccupations.

A science which hesitates to forget its founders is lost.— Merton (1968, p. 1)

IntroductionIn accounting, as in other disciplines in the social sciences, a discipline's jour-nals represent its principal research communication network. When knowledge

* The authors are grateful for the assistance of Jee Jang and Georgiana Hsu of the University atBuffalo and Columbia University, respectively. We also appreciate the helpful comments ofMessod Beneish, the two anonymous reviewers and the editor, Haim Falk.

Contemporary Accounting Research Vol. 5 No. 2 pp. 793-815

Page 2: Attributes of articles impacting contemporary accounting literature

794 L.D. Brown J.C. Gardner M.A. Vasarhelyi

regarding a discipline and related disciplines is accumulated, changes in thediscipline's research are inevitable. The evolution of a discipline's research canbe classified along a variety of attribute dimensions. As a discipline evolves, theunderlying attributes of its research generally are altered. Indeed, the accountingdiscipline has undergone a virtual revolution during the past two decades (Beaver(1981)).

This paper describes and analyzes four attribute dimensions that have impactedcontemporary accounting literature (CAL) between 1976 and 1984, and tests theability of our model to predict attribute levels one year ahead. The study pro-ceeds as follows. The next two sections discuss articles that have examined adiscipline's evolution and the data base utilized in our study, respectively. Thefollowing two sections respectively describe the methodology and present theresults. The last section summarizes the paper.

Attribute studiesThe history of a discipline or its subarea(s) can be examined by mapping theattributes of publications within the area(s) over time. For example, Dyckmanand Zeff (1984) used publication analysis to examine the composition by attri-bute areas of articles published in The Accounting Review and the Journal ofAccounting Research. More recently, Vasarhelyi, Bao and Berk (1987) describedthe attribute composition of six accounting research joumals. Publication countsprovide an excellent historical perspective of the discipline's (or subarea's) evolu-tion of thought, and we use them for this purpose. Nevertheless, publicationcounts are not useful for measuring how the influence of a particular discipline'sattributes have varied over time. Citation analysis is useful for this purpose, sowe employ this technique as well.

Garfield (1979, pp. 72-73) has demonstrated the usefulness of citation-basedattribute studies by utilizing the procedure to trace the development and valida-tion of DNA genetic coding theory.' Comparing his results to Asimov's (1963)account of the history of published DNA research, Garfield shows that citationanalysis: (1) closely duplicates Asimov's historical account; (2) identifies themost important contribution noted by Asimov as the most highly cited event; (3)identifies the researchers credited by Asimov for their contributions to be themost heavily cited; and (4) identifies an additional important event Asimovoverlooked (but which he later verified). Garfield's results suggest that citationanalysis is a useful way to trace the impact of various attributes of a disciplineover time.^

Citation analysis can be used to measure the impact that various classifications

1 DNA research is the study of deoxyribonucleic acid, the genetic material that exists within humancells.

2 Garfield (1979, pp. 24, 78, 149) discusses the use of impact factor analysis. Consistent with itsmeaning in the citation literature, this manuscript uses the term impact to mean infiuence orimportance.

Page 3: Attributes of articles impacting contemporary accounting literature

Attributes Impacting Accounting Literature 795

of research have had upon subsequent research.^ Previous research in accountinghas applied citation impact measures to determine influence along various dimen-sions. McRae (1974) used citation analysis to define the accounting informationnetwork of the flow of messages between the accounting knowledge system andother knowledge systems, and the flow of messages within the accounting knowl-edge system itself. Brown and Gardner (1985) applied citation analysis to mea-sure the impact of four accounting journals, and to identify those journals' mostinfluential papers on accounting research. Brown, Gardner and Vasarhelyi (1987)used citation analysis to evaluate the contributions of Accounting, Organizationsand Society dtiring the time period 1976-1984, while Gamble, O'Doherty andHyman (1987) used the procedure to study the influence of agency theory researchin accounting. Examples of citation studies in other disciplines include Hamel-man and Mazze (1973) in marketing; Carpenter and Narin (1973) in physics,chemistry, and biology; Ellis, Hepburn and Oppenheim (1978) in patents; andMunigesan and Moravcsik (1978) in theoretical physics.

The present study utilizes a data base that combines attributes (as noted above)of 1,110 individual accounting articles with citations to those articles. Our database allows for a broader analysis of the attributes of accounting articles than hasbeen accomplished heretofore. (See Brown, Gardner and Vasarhelyi (1987) andGamble, O'Doherty and Hyman (1987) for more limited analyses.) Our studyalso is the first to use a model that predicts levels of publications and impactfactors along various attribute dimensions.

Attribute classifications and citation measuresIn an effort to determine the impact of selected attributes of accounting articles,we merge the Brown and Gardner (1985, 1985a) citation data base with theVasarhelyi and Berk (1984) attribute data base. The Brown and Gardner database includes citations by major articles published in CAL to major articlespublished in CAL.'' CAL is defined for the purpose of this study as all mainarticles published between 1976 and 1986 in four journals: Accounting, Organi-zations and Society, The Accounting Review, Journal of Accounting and Econ-omics, and the Journal of Accounting Research.^ The Vasarhelyi and Berk

3 Brown and Gardner (1985,1985a) discuss some of the limitations of citation analysis for mea-suring research impact. These limitations include: (1) some citations may be negative ratherthan positive, (2) citations may be biased in favor of popular authors (halo effect), (3) citationsmay be by authors citing themselves (self-cites), (4) citations may be to hot topics which generatemany citations for only a short period of time, and (5) simple citation counts do not considerthe paper's age. We mitigate these problems by collecting citations over a long time period, usingimpact factors that adjust for the paper's age, and aggregating over several journals.

4 For an extensive discussion of the citation data base, see Brown and Gardner (1985).5 Three of these journals were ranked consistently in the top ten of the accounting research journals

(Howard and Nikolai (1983)). The Journal of Accounting and Economics was not mentionedin the Howard and Nikolai study due to its "newness," but we include it because we believe it hasbecome a major outlet for accounting research. Our failure to include citations by journalsdevoted to special areas of accounting research (e.g., tax; MIS) biases our results against thesetopical areas.

Page 4: Attributes of articles impacting contemporary accounting literature

796 L.D. Brown J.C. Gardner M.A. Vasarhelyi

attributes data base includes classifications of these same articles along a varietyof attribute dimensions.* The merged data base consists of attributes of CALarticles that were cited by other CAL articles.^ The impact of any particularattribute is measured using the number of citations to articles possessing thatattribute.

The four attribute dimensions analyzed in this study are: accounting area,research method, school of thought, and geographical focus (i.e., whether theissue addressed is U.S. or non-U.S. related).^ More specifically, articles wereclassified into one of five accounting areas: tax, financial, managerial, auditing,or information systems; one of eight research methods: intemal logic, simula-tion, primary archival (e.g., CRSP, COMPUSTAT), secondary archival (e.g.,literature reviews), case study, field study, lab. study, or opinion survey; oneof nine schools of thought: behavioral-HIP (Human Information Processing),behavioral-other, capital markets, time series, information economics/agencytheory, math, programming, accounting theory, accounting history, or institu-tional; and one of two geographical foci - whether the primary issue it addressedwas U.S. or non-U.S. related. In cases where a paper focused on several account-ing disciplines, research methods, etc., Vasarhelyi and Berk (1984) determinedthe dominant discipline by analyzing both the paper's comparative space devotedto each discipline as well as the comparative number of references to each disci-pline.^ Brown and Vasarhelyi (1985) used a similar procedure to examine andvalidate the Vasarhelyi and Berk (1984) classifications.'°

MethodologyThe first technique used to trace the evolution of accounting article attributes isrelative publications (RP) analysis. This metric is calculated as:

m

RP,-, = n,,/ I "i, (1)( = 1

6 In addition to the four journals comprising CAL, the Vasarhelyi and Berk data base includesattribute classifications of main articles in two other journals: Journal of Accounting, Auditingand Finance and Auditing: A Journal of Practice and Theory.

1 The merger resulted in the classification of 1,110 articles and 5,385 citations to those articles.8 The Vasarhelyi and Berk data base classifies articles along 12 dimensions. In addition to the four

attribute classifications chosen for this study, the eight others are inference style, mode ofreasoning, mode of analysis, information, treatment, objective, applicability, and foundationdiscipline. Some of the classifications overlap (e.g., mode of reasoning, research method), andothers have small sample sizes (e.g., treatment). The four classifications selected have theminimum definitional overlap and maximum sample sizes.

9 For example, Foster (1977) could be classified as either using a time-series or capital marketsresearch method. The article was classified as time-series because more space was devoted to thediscussion of time-series than capital markets and more references were made to the time-seriesthan to the capital markets literature. Articles with no dominant classification were treated asmissing values.

10 Some articles were excluded from three of the four attribute dimensions examined, because theirtopic could not be classified in any of the areas within the attribute dimension or, as noted infootnote 9, they related equally to more than one area. Article exclusions approximated 8.5percent for accounting area, two percent for school of thought, and 34 percent for geographicalfocus.

Page 5: Attributes of articles impacting contemporary accounting literature

Attiibutes Impacting Accounting Literature 797

where the relative publications of articles with attribute i in period t (RP,,) equalsthe number of year t publications with attribute / (n,,), divided by the sum of allarticles published in year / classified along that attribute dimension (e.g., m = 9for school of thought because this attribute has nine classifications). In this way,the relative publications of all categories of a particular attribute are placed on acommon footing, in that they sum to 1.000.

The second technique used to frace the evolution of accounting article attri-butes is relative impact factor (RIF) analysis, a type of citation analysis. Thistechnique measures an attribute's infiuence on a per article basis. This relativeimpact factor can be expressed as:

i )J (2), i nyj/[( .f [ci,/^i n,;,)Jwhere the relative factor of attribute i in year t (RIF,,) equals year t citations toarticles with attribute i published between 1976 and year t (c,,), divided by thesum of all articles with attribute / published between 1976 (y = 1) and year t,

where 1976 < f < 1986.

This number is then divided by the sum of all (m) impact factors for the particularattribute dimension to provide a relative impact measure between attributes.'' Inthis way, the relative impact factors of all categories of a particular attributedimension sum to 1.000.

These two metrics, relative publications (RP) and relative impact factors (RIF),are used in three ways. First, the values of the metrics for each attribute dimen-sion are presented for each of the nine years, 1976 to 1984 (hereafter, examina-tion period), along with the mean and the standard deviation of the distribution.These descriptive statistics are useful for "seeing" whether temporal or cross-sectional pattems exist in the data. Second, the results of generalized leastsquares (GLS) regressions that relate values of RP and RIF to time (i.e., 1976equals one,..., 1984 equals nine) are presented.'^ The regressions are used todetermine the significance of the temporal trend (i.e., the significance of theslope term), and to examine the predictive power of the historical analysis.Third, predictions for 1985 and 1986 (hereafter, prediction period) are formu-lated and compared to the actual 1985 and 1986 values. The predictive accuracyof the model is useful for ascertaining whether the short term evolution ofselected accounting article attributes can be predicted, and for comparing thepredictive validity of publication analysis with that of impact factor analysis.

11 m equals five for accounting area, eight for research method, nine for school of thought, and twofor geographical focus.

12 For a given year, the RP,s and RIF/S sum to 1 .(KK), so the RP, and RIR regressions are notindependent. Consequently, we used a generalized least squares procedure, the SAS SYSLIN-SUR (seemingly unrelated regressions) program, to estimate the model parameters. This estima-tion procedure is equivalent to the equation-by-equation application of ordinary squares becauseXi = X for every i = l,2,...,N (Pindyck and Rubinfeld (1981, pp. 348-349)).

Page 6: Attributes of articles impacting contemporary accounting literature

798 L.D. Brown J.C. Gardner M.A. Vasarhelyi

Results: Reiative publications and relative impact factors

Accounting areaTable 1 displays per year relative publications, relative impact factors, andtemporal GLS regressions for articles classified by accounting area during theexamination period. Panel A reveals that financial accounting has been the mostdominant publication area in CAL in each year (averaging approximately 50percent); that managerial accounting has been, on average, the second mostoft-published accounting area (approximately 25 percent); that auditing, on aver-age, has been the third most oft-published accounting area (approximately 19percent); and that information systems and tax have published the smallest per-centage of articles (on average, the two combined are approximately six percentof the total). Panel B presents the regressions of relative publications in eachaccounting area on time. The F-values of the regressions denote the lack of asignificant relationship between relative publications and time, confirming thePanel A pattems of relatively steady percentages of temporal publications in eachaccounting area.

The relative impact measures in Panels C and D provide quite different insightsregarding the five accounting areas. Panel C shows that financial accountingarticles have exerted a greater impact than managerial and auditing articles in sixand eight years, respectively, while managerial and auditing articles have beenmore infiuential than information systems articles in four and two years, respec-tively. These findings contrast with the relative publication results shown inPanel A. In addition, the results show that the range between the highest andlowest relative publication measure for accounting area has remained fairly con-stant (48.7 percent in 1976 versus 56.3 percent in 1984), whereas the range ofrelative impact factors has narrowed substantially (from 69.0 percent in 1976 to27.9 percent in 1984). The convergence of relative impact factors suggests thateach accounting area has drawn upon its previous research rather consistentlyacross areas in recent years, and that no one area has been disproportionatelyimpacted by literature outside of CAL. '^

Panel D reveals that a significant (defined as the ten percent level throughoutthis study) relationship exists between the relative impact factor of four account-ing areas and time. More specifically, the financial and managerial areas havebecome less infiuential over the examination period, while the auditing andinformation systems areas have become more infiuential. Tax articles evidenceno significant temporal trend.

In sum, the Table 1 findings suggest that the percentage of publications in eachaccounting area has been nearly constant; recent papers (1979-1984) in theauditing and information systems areas have been more infiuential than earlierpapers in these areas (1976-1978); and recent papers (1979-1984) in the finan-

13 This statement implicitly assumes that the total number of (CAL and non-CAL) citations by eacharticle is independent of the accounting area.

Page 7: Attributes of articles impacting contemporary accounting literature

Attributes Impacting Accounting Literature 799

cial and managerial areas have been less influential than earlier papers in theseareas (1976-1978).

Research methodTable 2 presents relative publications, relative impact factors, and temporal GLSregressions for articles classified by research method. Panel A reveals that, incontrast to the findings for accounting area, no research method has dominatedCAL over the entire examination period. Nevertheless, three methods (i.e.,intemal logic, primary archival, and lab. studies) have been used more often thanthe other five in every year with primary archival being the most popular methodbetween 1982 and 1984. Panel B shows that, with the exception of simulationand secondary arehival studies, the relative publications exhibit significant tem-poral trends. More specifically, a significant downward trend has occurred inintemal logic, case studies, and field studies, while a significant upward trend isevident in primary archival, lab. studies, and opinion surveys.

The relative impact factors in Panel C show that the impact per article ofprimary archival and lab. studies appears to be on the decline, while that ofsecondary arehival and field studies is on the increase. Panel D confirms theseobservations, evidencing significant slopes for the attendant temporal GLS re-gressions. The impact factor trends contrast with the publication trends - pri-mary archival and lab. study publications were increasing and field studies weredecreasing.

The combined publications and impact fact results suggest that, on average, asthe use of a research method matures in the literature, the per article impact ofpapers using that research method declines. Altematively, as the area matures,the probability of any particular article being cited declines.''* These findings areconsistent with the results of Meadows and O'Connor (1971) who maintain thatthe number of citations per article increases (decreases) in the early (latter) stagesof an area's development. We examine below whether the observed trends inrelative publications and impact factors of the research methods continues in the1985 and 1986 period.

School of thoughtTable 3 presents the relative publications, relative impact factors, and temporalGLS regressions for articles classified by school of thought. Panel A suggeststhat relative publications have increased in two areas, behavioral-HIP and capitalmarkets, and have declined in two others, math, programming and accountingtheory. Panel B provides statistical evidence confirming the Panel A data; namely,behavioral-HIP and capital markets (math, programming and accounting theory)have significant positive (negative) slop»es.

14 One implication of these results is that it is easier to publish less-influential papers using a popularrather than an unpopular research method. Altematively, papers using unpopular researchmethods must make important contributions in order to be published.

Page 8: Attributes of articles impacting contemporary accounting literature

ON "n CN O OS—' •* VO «-l -H

O O O O O

o o o o o

o o o o o

O f* ^ "^ OS

o T) <N — qo o o o d

—t ON m ^ -Hm ^ c* •'t fNq Tj^ r* -H o

d d d d d

q 5 (N (N qd d d d d

§ •* O\ ^ -H

r^ fi fi O«n •-' fs q

o o d o d

00 00 VO >n 00-H •* <N in 5q in r4 —' qd d d d d

CN ^ r^ m (Ni^ ov —• •* r~.q •* cN - q

o o o o o

^ O CN Q C<1tN O t~ *. —q «n m q qd d d d d

d d d d d

r^ Os in ON ^O (N oo m 00t^ oo Tt m vc

d d d d d

d d d d d

Page 9: Attributes of articles impacting contemporary accounting literature

-a

i

O

O — O O —d d d d d

o o o o o

C - 0 0 >

d d d d d

d d d d d

00 <N r ) <s >o..4 \o r r r--

o fs c^ <N • -d d d d d

d d d d d

r<i v^ r^ -^ <3\O <N ^- -H rn

d d d d d

f ^ 1 O fO»n o >-• >n t^O r* (N — <So o o o o

§ c4 m ON 300 r^ 00 g\"^ m O O

d d d d d

d d d d d

d d d d d

•a'S sp

iBMt:s-

d d d d d

t ^ r o 00Q ^ (s >n r« - - . ^ 00 CSo rj-' IT) vd t~:

do o od d d dI I

I lltl

>. x>

•a o

S 2

i I.La

^ II u II

fillsi.if

•S

•S + ^ +

in?I " I "-

Page 10: Attributes of articles impacting contemporary accounting literature

2

C/3 T3 oooooooo

CO rO d "" fN \O r—coocopoo—*o

0 O g f ^ O O 0 0

oooooooo

S ^ O ON ON O^ "^ f^^- ^ m —' 1—I m \C

( N O r n O O O t N OOOOOOOOO

OOOOOOOO

j O O O O O ^ G O C Nr^ •-< 00 c^ o O '"•f ^ O f S O O O ' — ' ' — 'oooooooo

•^(Nroc^^^CNro'*^•^r4coco^H<Noo»nroOcoooOt—pdddddddd

roO<sppp-Hd

o o r 2

dddddddd

oooooooo

8!?8SSSSooddooodo

o o o o o o o oII III

Eiii

Page 11: Attributes of articles impacting contemporary accounting literature

sCD

oooooooo

poooooooo

oooooooo

oooooooo

oooooooo

I O t- — (1 O CS (S in 1

oooooooo

rJO — 3«®(S

oooooooo

oooooooo

dd

oooooooo

•a

. ili S » 2i 8 a S •. <» U (1. I

oooooooo

d dI I

o o jp r o —

d S S d d oI I I

li'h

Page 12: Attributes of articles impacting contemporary accounting literature

—) S '9^ OH uH 2 t»

0 0 0 0 OOOOO

^i f ^ ^ ^ ( ^ f ^ ^ ^ g ^ ( ^ ^ ^

fSOOTtOO

qggS(O d> d S

^ f^ ON CN *O O O fN ON

— (N(Np pp^- 'pOdddd ddddd

Qop

G0tn<O

d d d d

ON O t* 3" (N O^ ^ ^

ddd d d d d d

0 0 0 0 OOOOO

dddd ddddd

dddd ddddd

pp p p p poddd ddddd

^ • .

13

L -g -g §-.11 < I § b i

^ ^ 00 C^ ^ f "" ^ ^ ^ ^ ^ ^ ^ ?

dddd ddddd

, 0 0 0 fs-"moo

OOOO OOOOoddd d d o o

I I I I

S ca BO L

Page 13: Attributes of articles impacting contemporary accounting literature

ON

dddd ddddd

mo—'Ov inint--<in" - ^ ^ r j " ^ — " q q q qdddd ddddd

qdddd ddddd

oc-^NOTf in—*oo(Noom^oo^ ON-"inr-in« & — <S rjqqqqdddd ddddd

O m ^ 5oono^dddd ddddd

n^oo oi^ooovo•nooovo vooom — r-^^ ^^ r.^ ^^ ^-4 C^ O^ GC>

dddd ddddd

OOOO OOOOO

osNOin-t ONOOCSOOI

OOOO OOOOO

fT) o O 00 NO oo O ^ ^C|)g — ON OOOOmQO

dddd ddddd

qq qqqqdddd ddddd

OOOO OOOOO

• ? • >

s I 00 U H ^

OOOOOOOO-"mf«^O^dddd ddddd

ddddI I I

dddodII .s

•a +

Page 14: Attributes of articles impacting contemporary accounting literature

806 L.D. Brown J.C. Gardner M.A. Vasarhelyi

Panels C and D show that the relative impact factors of the nine schools ofthought evidence no definitive time trends. Thus, while substantial temporalshifts have occurred in relative publications, substantial temporal shifts have notoccurred in relative impact factors.

Geographical focusTable 4 presents the relative publications, relative impact factors, and temporalGLS regressions for the geographical focus of the article. Panels A and Brespectively reveal that U.S.-related literature has consistently dominated CAL,and that significant linear publication trends do not exist. Panels C and D respec-tively reveal that the per article impact of U.S.-related papers dominates non-U.S. related papers, and that significant temporal trends do not exist.

Results: Predictions of relative publications and relative impact factorsTables 5 to 8 present 1985 and 1986 prediction values for the four attributionaldimensions. The 1985 (1986) predictions, based on the temporal GLS regressionof relative publication (RP) and relative impact factor (RIF) values for 1976through 1984 (1977 through 1985), are used to validate the previously discussedtrends that were observed with the 1976 through 1984 data.'^ The attribute'smean and standard deviation for the regression period are used as the predictedvalue and standard error, respectively, for attribute areas that do not exhibit asignificant temporal trend in the 1976 through 1984 (1977 through 1985) regres-sion.'^ The tables contain the actual value, the predicted value, the standarderror, the 95 percent confidence interval of the prediction, and the number ofstandard deviations the actual value differs from its predicted value.'^

For the sake of parsimony, we henceforth refer to predictions within onestandard deviation of the actual value as being "very adequate;" those betweenone and two standard deviations from the actual value as "somewhat adequate;"and those greater than two standard deviations from the actual value as "inade-quate." In addition, shifts in predictive accuracy between very adequate andinadequate from 1985 to 1986 or vice versa are referred to as "inconsistent"predictions. Either no shift between classifications or a shift of one classification(i.e., from very adequate to somewhat adequate; from somewhat adequate toinadequate) will be referred to as "consistent" predictions.

15 Predictions of 1986 RP and RIF values were also performed using the 1976 through 1985 data.The results were not qualitatively different than those reported in the text.

16 With the exception of the RP regressions in accounting area and the RP and RIF regressionsin geography, the regressions within each attribute dimension had mixed results (i.e., theregressions within the dimensiotis were neither all significant nor all insignificant). Therefore,predictions in the mixed results dimetisions do not sum to 1.000 (i.e., the slope parameters fromthe significant regressions plus the means from the insignificant regressions do not sum to 1.000).

17 Using the standard deviation of a within-sample observation induces a conservative bias to ourresults. An unbiased estimate of the confidence band associated with the predictions wouldaccount for the fact that the standard deviation of a holdout observation exceeds that of a within-sample observation.

Page 15: Attributes of articles impacting contemporary accounting literature

r-

ii

en fno qd d

00 ^ -

d d

00 —«

d d

s:200 ^^

d d

m r-oo —

r-i 0000 ^-*

r- ro^" ooa\ o

00 00

CO rod d

o oI

5

00 ^ -

d d

(N 00

d d

d d

D

IDZ

99d d

« g=1 z

i I

\.s

Page 16: Attributes of articles impacting contemporary accounting literature

808 L.D. Brown J.C. Gardner M.A. Vasarhelyi

TABLE 51985 and 1986 predictions of attribute valuesAccounting area

A. RP predictions"1985TaxFinancialManagerialAuditingInfo, systems

1986TaxFinancialManagerialAuditingInfo, systems•=

B. RIF predictions"1985TaxFinancialManagerialAuditingInfo, systems

1986TaxFinancialManagerialAtiditingInfo, systems

A y->tiinl

value

0.0420.4580.2920.2000.008

0.0000.4760.2760.2290.019

0.0460.2380.1770.2460.293

0.0430.2020.2120.1550.388

rrCQlCtCG

value

0.0230.4990.2540.1880.036

0.0250.4940.2450.2000.006

0.0270.1230.1620.2740.398

0.0330.1730.1350.2750.373

Std.1 , -.,, ,1 ciTor

(Dev.)

0.0190.0450.0620.0560.019

0.0210.0500.0510.0470.011

0.0340.0700.0510.0360.080

0.0330.0480.0490.0330.080

95% confidenceinterval

Lower

-0.0210.3950.1110.059

-0.008

-0.0230.3790.1270.092

-0.019

-0.051-0.038

0.0440.1910.213

-0.0430.0620.0220.1990.188

Upper

0.0670.6030.3970.3170.080

0.0730.6090.3630.3080.031

0.1050.2840.2800.3570.582

0.1090.2840.2480.3510.557

Number of1 Std. dev.l

predicted

1.000.910.610.211.47

1.190.360.610.621.18

0.561.660.290.781.32

0.300.601.573.640.19

'Predictions of relative publications, using the model described in footnote b of Tables 1-4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for the 1986publications.

'' Predictions of relative impact factors, using the model described in footnote d of Tables 1 -4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for 1986 predictions.

'The 1977 Arough 1985 regression was significant at the ten percent level, as opposed to aninsignificant 1976 through 1984 regression. Therefore, the GLS regression is used for the 1986predictions.

Accounting areaPanel B of Table 1 revealed no significant temporal trends in the relative publica-tions of articles classified by accounting area during the examination period.Panel A of Table 5 presents predictions of relative publications classified byaccounting area for the 1985 and 1986 prediction period, using the means andstandard deviations of the within-sample estimates in Panel B of Table 1. Predic-tions of publications in the financial, managerial, and auditing areas are veryadequate in both years, and those for tax and information systems are (at least)

Page 17: Attributes of articles impacting contemporary accounting literature

Attributes Impacting Accounting Literature 809

TABLE 61985 and 1986 predictions of attribute valuesResearch method

A. RP predictions'1985Intemal logicSimulationPrimary archivalSecondary archivalCase studiesField studiesLab. studiesOpinion survey

1986Intemal logicSimulationPrimary archivalSecondary archivalCase studiesField studies'*Lab. studiesOpinion survey

B. RIF predictions"1985Intemal logicSimulationPrimary archivalSecondary archivalCase stucUesField studiesLab. studiesOpinion survey

1986Internal logicSimulationPrimary archival'*Secondary archivalCase studiesField studies'*Lab. studies'*Opinion survey'

value

0.2300.0480.3650.0240.0000.0400.2060.087

0.2680.0710.4110.0270.0090.0090.1160.089

0.1160.0940.1850.2090.0630.1180.1170.097

0.0910.1130.1190.2550.1280.1540.0630.078

.

value

0.1890.0370.4210.049

-0.0070.0060.2130.111

0.1690.0380.4150.0450.0000.0260.2320.108

0.1170.0650.1670.2280.0590.2770.0530.077

0.0960.0760.2040.2260.0660.1530.1680.057

Std.

error(Dev.)

0.0340.0170.0280.0270.0110.0090.0250.013

0.0340.0170.0300.0280.0040.0160.0210.014

0.0770.0540.0280.0350.0420.0350.0690.040

0.0300.0490.0250.0310.0360.0710.1010.018

95% confidenceinterval

Lower

0.111-0.002

0.356-0.013-0.032-0.015

0.1550.081

0.091-0.001

0.346-0.020-0.009-0.011

0.1840.076

-0.061-0.060

0.1020.147

-0.0380.196

-0.106-0.015

0.027-0.037

0.1460.155

-0.017-O.OU-0.065

0.015

Upper

0.2670.0760.4860.1110.0180.0270.2710.141

0.2470.0770.4840.1100.0090.0630.2800.140

0.2960.1890.2320.3090.1560.3580.2120.169

0.1650.1890.2620.2970.1490.3170.4010.099

Number of|Std. dev.

predicted

1.210.651.970.930.643.620.291.78

2.911.940.130.642.251.065.521.36

0.010.540.640.540.104.560.920.50

0.160.763.400.941.720.011.041.16

'Predictions of relative publications, using the model described in footnote b of Tables 1-4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for the 1986publications.

"Predictions of relative impact factors, using the model described in footnote d of Tables 1-4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for 1986 predictions.

' The 1977 through 1985 regression was significant at the ten percent level, as opposed to aninsignificant 1976 through 1984 regression. Therefore, the GLS regression is used for the 1986predictions.

"The 1977 through 1985 regression was insignificant at the ten pereent level, as opposed to aninsignificant 1976 through 1984 regression. Therefore, the mean for the 1977 through 1985 periodis used for the 1986 predictions.

Page 18: Attributes of articles impacting contemporary accounting literature

810 L.D. Brown J.C. Gardner M.A. Vasarhelyi

TABLE 71985 and 1986 predictions of attribute valuesSchool of thought

A. RP predictions"1985Behavioral-HIPBehavioral-OtherCapital marketsTime seriesInfo, economics/

Agency theoryMath, programmingAccounting theoryAccounting historyInstitutional

1986Behavioral-HIP"*Behavioral-OtherCapital marketsTime series'Info, economics/

Agency theoryMath, programmingAccounting theoryAccounting historyInstitutional

B. RIF predictions'"1985Behavioral-HIPBehavioral-OtherCapital marketsTime series

value

0.1280.2230.2770.043

0.0430.0110.1700.0110.096

0.1450.1200.2410.024

0.0720.0120.2770.0480.060

0.1440.0960.2220.222

value

0.2330.2150.3290.065

0.039-0.010

0.0570.0310.062

0.1640.2110.3200.050

0.039-0.014

0.0800.0310.063

0.1340.1030.2130.197

Std.

error(Dev.)

0.0410.0460.0420.008

0.0390.0130.0250.0230.032

0.0640.0420.0410.006

0.0390.0130.0350.0230.033

0.0590.0420.1600.110

95% confidenceinterval

Lower

0.1380.1090.2320.047

-0.051-0.040-0.001-0.022-0.012

0.0160.1140.2250.036

-0.051-0.044-0.001-0.022-0.013

-0.0020.006

-0.156-0.057

Upper

0.3280.3210.4260.083

0.1290.0200.1150.0840.136

0.3120.3080.4150.064

0.1290.0160.1610.0840.139

0.2700.2000.5820.451

*Jumber ofStd. dev.

)redicted

2.540.171.262.75

0.101.614.460.871.06

0.302.171.934.33

0.852.005.630.740.09

0.170.170.060.23

somewhat adequate in each of the two years. Moreover, the relative publicationpredictions for all five accounting areas are consistent.

Panel D of Table 1 revealed significant temporal trends, suggesting the grow-ing influence of auditing and information systems articles, and the declininginfluence of financial and managerial articles. Panel B of Table 5 reveals that thepredictions of relative influence are generally consistent with these temporaltrends, with all predictions except auditing in 1986, being (at worst) very ade-quate in one year and somewhat adequate in the other year. Moreover, with theexception of the 1986 auditing prediction, predictions of relative impact areconsistent between the two years.

Research methodPanel A of Table 6 provides evidence that utilizing 1976-1984 (1977-1985)trends is not especially useful for predicting 1985 (1986) relative publication

Page 19: Attributes of articles impacting contemporary accounting literature

Attributes Impacting Accounting Literattire 811

TABLE 7 (Concluded)

Info, economics/Agency theory

Math, programmingAccounting theoryAccounting historyInstitutional

1986Behavioral-HIPBehavioral-OtherCapital markets'Time seriesInfo, economics/

Agency theoryMath, programming°Accounting theoryAccotinting historyInstitutional

. . .Actualvalue

0.1490.0170.0720.0160.061

0.1080.1040.2030.150

0.1590.0000.0790.1200.077

value

0.1510.0540.0740.0190.055

0.1500.0910.2420.221

0.1670.0160.0620.0210.062

Std.

ciTor(Dev.)

0.0800.0340.0500.0240.038

0.0320.0180.0360.081

0.0570.0170.0300.0230.032

95% confidenceinterval

Lower

-0.033-0.024-0.041-0.036-0.033

0.0760.0490.1590.034

0.036-0.023-0.007-0.032-0.012

Upper

0.3350.1320.1890.0740.143

0.2240.1330.3250.408

0.2980.0550.1310.0740.136

Number of1 Std. dev.l

predicted

0.031.090.040.120.16

1.310.721.080.88

0.140.940.574.300.47

"Predictions of relative publications, using the model described in footnote b of Tables 1-4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for the 1986publications.

'' Predictions of relative impact factors, using the model described in footnote d of Tables 1-4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for 1986 predictions.

''The 1977 through 1985 regression was significant at the ten percent level, as opposed to aninsignificant 1976 through 1984 regression. Therefore, the GLS regression is used for the 1986predictions.

•"The 1977 through 1985 regression was insignificant at the ten percent level, as opposed to aninsignificant 1976 through 1984 regression. Therefore, the mean for the 1977 through 1985 periodis used for the 1986 predictions.

levels. More specifically, otir predictions are very adequate in both years for onlyone research method, secondary archival, and they are inadequate for four othersin (at least) one year (i.e. intemal logic, case studies, field studies, and lab.studies). Two areas (i.e., case studies and lab. studies) exhibit inconsistentpredictions between the two years.

Panel B of Table 6 reveals whether the 1976-1984 (1977-1985) temporalpattems of relative impact measure revealed in Panel D of Table 2 extend to 1985(1986). Predictions are very adequate in both years for three research methods(i.e., intemal logic, simulation, and secondary archival); somewhat adequate in(at least) one year for three research methods (i.e., case studies, lab. studies, andopinion survey); and inadequate in (at least) one year for two research methods(i.e., primary archival and field studies). Moreover, inconsistent results areobtained for the primary archival and field study research methods.

In sum, the best (combined relative publication and impact factor) resultspertain to the secondary archival research method and the worst results pertain to

Page 20: Attributes of articles impacting contemporary accounting literature

812 L.D. Brown J.C. Gardner M.A. Vasarhelyi

TABLE 81985 and 1986 predictions of attributeGeographical focus

A. RP predictions"1985U.S.Non-U.S.

1986U.S.Non-U.S.

B. RIF predictions"1985U.S.Non-U.S.

1986U.S.Non-U.S.

Actualvalue

0.8290.171

0.7560.244

0.7390.261

0.6710.329

values

Predictedvalue

0.8660.134

0.8570.143

0.7850.215

0.7560.244

Std.error(Dev.)

0.0340.034

0.0300.030

0.1410.141

0.1160.116

95% confidenceinterval

Lower

0.7870.056

0.7780.074

0.460-0.110

0.488-0.023

Upper

0.9440.212

0.9160.212

1.1100.540

1.0240.511

Number ofi Std. dev. 1Actual vs.predicted

1.091.09

3.673.67

0.330.33

0.730.73

"Predictions of relative publications, using the model described in footnote b of Tables 1-4, with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for the 1986publications.

•" Predictions of relative impact factors, using the model described in footnote d of Tables 1 - 4 , with1976 through 1984 data for the 1985 predictions, and 1977 through 1985 data for 1986 predictions.

field Studies. Moreover, we are better able to predict the relative importance ofarticles using a particular research method than we are to predict the number(proportion) of articles that use the method. Finally, consistent predictions areobtained for both the relative publications and the relative impact factors.

School of thoughtThe usefulness of temporal frends between 1976 and 1984 (1977 and 1985) inpredicting 1985 (1986) relative publications are shown in Table 7. The predictiveresults for the relative publications segmented by school of thought classifica-tions generally are inadequate. More specifically. Panel A reveals that very ade-quate predictions of relative publications are obtained for only two schools ofthought (i.e., information economics/agency theory, and accounting history),while inadequate predictions (in at least one year) are obtained for five schools ofthought (i.e., behavioral-HIP, bdiavioral-other, time series, math, programming,and accounting theory). Inconsistent results are observed for relative publicationpredictions for two schools of thought, behavioral-HIP and behavioral-other.

Panel B of Table 7 shows that predictions of the relative influence of fiveschools of thought are very adequate in both years (i.e., behavioral-other, timeseries, information economics/agency theory, accounting theory, and institu-

Page 21: Attributes of articles impacting contemporary accounting literature

Attributes Impacting Accounting Literature 813

tional); that only one prediction is inadequate in either year (i.e., accountinghistory in 1986); and that only one inconsistency is observed (i.e., accountingtheory). In sum, our best (combined publications and impact factors) school ofthought predictions are for information economics/agency theory and institu-tional; we can better predict the relative infiuence of publications than we can thenumber (percentage) of publications; and, we generally obtain consistent predic-tions for both relative publications and impact factors.

Geographical focusPanel A of Table 8 reveals somewhat adequate (inadequate) predictions of rela-tive publications for 1985 (1986). However, Panel B of Table 8 reveals veryadequate predictions of relative influence in both 1985 and 1986. Once again,predictions of relative influence are more accurate than are those of relativepublications, and the impact factor predictions are more consistent than thepublication predictions.

SunnnaryThis paper describes and analyzes the attribute dimensions of articles that haveimpacted contemporary accounting literature (CAL) for the period 1976-1984(examination period) and tests the ability of a generalized least squares regres-sion model to predict attribute levels in 1985 and 1986 (prediction period).'* Anarticle's attributes are defined along four dimensions: accounting area, researchmethod, school of thought, and geographical focus. Two methods are used toexamine the evolution of these attributes, publication analysis and citation anal-ysis. The principal results follow.

Financial accotmting has been the dominant publication area in CAL duringthe examination period, followed (in order) by managerial, auditing, informationsystems, and tax. The percentage of publications classified by accounting areahave not exhibited significant temporal trends. The relative impact of financialand tnanagerial articles has declined during the examination period, while therelative impact of auditing and information systems articles has increased. '^ Theprediction results for impact factors and publications, in general, reveal similaraccuracy and consistency.

No single research method has dominated CAL over the nine-year examina-tion period. However, intemal logic, primary archival, and lab. studies havebeen used more often than the other five methods in each of the nine years, andprimary archival was the most popular method between 1982 and 1984. Publica-tions utilizing the primary archival, lab. studies, and opinion survey methodsincreased significantly during the examination period; articles using intemal

18 With only two years of prediction data, the authors feel that definitive statements cannot be maderegarding the predictive consistency of the model. Future research using more temporal datamay be able to provide more definitive explatiations as to why certain attributes are relatively easy(hard) to predict.

19 This may suggest that the state of the art is shifting rapidly in the auditing and informationsystems areas.

Page 22: Attributes of articles impacting contemporary accounting literature

814 L.D. Brown J.C. Gardner M.A. Vasarhelyi

logic, case studies, and field studies decreased significantly; and no significanttrends were evident in studies using the simulation and secondary archival meth-ods. Articles using the secondary archival and field study methods became moreinfiuential during the examination period; articles using the primary archival andlab. study methods became less infiuential; and no significant trends were appar-ent in the impact of articles using the other four research methods. Regarding theprediction evidence, our best (combined publication and impact factor) resultspertain to secondary archival, and our worst results pertain to field studies.Moreover, we are better able to predict the relative infiuence of articles usingparticular research methods than we are to predict the number (percentage) ofarticles using that method. In addition, consistent predictions were obtained forrelative publications and impact factors.

Five of nine schools of thought evidenced temporal stability of publicationsduring the examination period, two had increasing relative publications (i.e.,behavioral-HIP and capital markets), and two had decreasing relative publica-tions (i.e., math, programming and accounting theory). Predictions of relativepublications were poor, with only two classifications being predicted within onestandard deviation of their actual values in both prediction years. Predictions ofthe relative impact factors were very accurate, with five areas having predictionsthat fell within one standard deviation of the actual number in both predictionyears. In addition, three of the schools of thought were predicted (at least) withintwo standard deviations of their actual values in 1985 and 1986. Similar toresearch method, predictions of relative impact factors for schools of thoughtwere more accurate than predictions of relative publications, and both impactfactors and publications evidenced similar predictive consistency.

In geographical focus, publications of articles addressing U.S. issues domi-nated non-U.S. issues; articles addressing U.S. issues exhibited a greater impactper article; and predictions of relative impact factors were more consistent thanpredictions of the number (proportion) of pubhshed articles.

In sum, we have shown that temporal trends exist in publications and impactfactors of selected attributes of accounting papers: accounting area, researchmethod, school of thought, and geographical focus. Moreover, we have shownthat the relative amount of future publications in an attribute area can be pre-dicted with limited success, and that the relative importance of these publicationscan be predicted with considerable success.

ReferencesAsitnov, I., The Genetic Code (New York: New American Library, 1963).Beaver, W. A., Financial Reporting: An Accounting Revolution (Etiglewood Cliffs:

Prentice-Hall, 1981).Brown, L.D. and J.C. Gardtier, "Using Citation Analysis to Assess the Impact of Joumals

and Articles oti Contemporary Accoutititig Research (CAR)," Journal of AccountingResearch (Spring 1985) pp. 84-109.

and , "Applying Citation Atialysis to Evaluate the Research Contributiotisof Accountitig Faculty and Doctoral Programs," The Accounting Review (April 1985a)pp. 262-277.

Page 23: Attributes of articles impacting contemporary accounting literature

Attributes Impacting Accounting Literature 815

, and M.A. Vasarhelyi, "An Analysis of the Research Contributiotis ofAccounting, Organizations and Society, 1916-19S4," Accounting, Organizations andSociety (1987 No. 2) pp. 193-204.

Brown, L.D. and M.A. Vasarhelyi, Accounting Research Directory: The Database ofAccounting Literature (New York: Marcus Wiener Publishing, 1985).

Carpenter, M.P. and F. Narin, "Clustering of Scientific Journals," Journal of the Ameri-can Society for Information Science (November 1973) pp. 425-436.

Dyckman, T.R. and S.A. Zeff, "Two Decades of the Journal of Accounting Research,"Journal of Accounting Research (Spring 1984) pp. 225-297.

Ellis, P., G. Hepbum and C. Oppenheim, "Studies on Patent Citation Networks," Jourrmlof Documentation (March 1978) pp. 12-20.

Foster, G., "Quarterly Accounting Data: Time-Series Properties and Predictive-AbilityResults," The Accounting Review (January 1977) pp. 1-21.

Gamble, G.O., B. O'Doherty and L.M. Hyman, "The Development of Agency Thought:A Citation Analysis of the Literature," The Accounting Historians' Journal (Spring1987) pp. 7-26.

Garfield, E., Citation Indexing - Its Theory and Application in Science, Technology,and Humanities (New York: John Wiley and Sons, 1979).

Hamelman, P.E. and E.M. Mazze, "Cross Referencing Between AMA Journals andOther Publications," Journal of Marketing Research (May 1973) pp. 215-219.

Howard, T.P. and L.A. Nikolai, "Attitude Measurement and Perceptions of AccountingFaculty Publication Outlets," The Accounting Review (October 1983) pp. 765-776.

McRae, T.W., "A Citation Analysis of the Accounting Information Network," Journalof Accounting Research (Spring 1974) pp. 80-92.

Meadows, A.J. and J.G. O'Connor, "Biographical Statistics as a Guide to Growth Pointsin Science," Science Studies (January 1971) pp. 95-99.

Merton, R.K., Social Theory and Social Structure (New York: The Free Press, 1968).Munigesan, P. and M.J. Moravcsik, "Variation of the Nature of Citation Measures with

Journals and Scientific Specialties," Journal of the American Society for InformationScience (May 1978) pp. 141-147.

Pindyck, R.S. and D.L. Rubinfeld, Econometric Models and Economic Forecasts,Second Edition (New York: NcGraw-Hill Book Company, 1981).

Vasaihelyi, M.A., D.H. Bao and J. Berk, "Patterns in the Evolution of ContemporaryAccounting Literature," (Unpublished manuscript, Columbia University, 1987).

Vasarhelyi, M.A. and J. Berk, "The Multiple Taxonomies of Accounting Research,"(Unpubhshed manuscript, Columbia University, 1984).

Page 24: Attributes of articles impacting contemporary accounting literature