Who Uses Financial Statements

ACCOUNTING HORIZONS American Accounting Association Vol.31,No.3DOI:10.2308/acch-51736 September2017

pp.55–68

Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR

Michael S.Drake

Brigham Young University

Phillip J.Quinn

University of Washington

Jacob R.Thornock

Brigham Young University

SYNOPSIS:We link EDGAR requests for financial statements originating from a particular U.S.ZIP code to demographic characteristics of that ZIP code.We focus on four demographics:income,household characteristics, education,and local conditions.Overall,we find each of the four demographics explain significant cross-sectional variation in EDGAR financial statement use.On a relative basis,we find that education has significantly more explanatory power for financial statement usage than does income or household characteristics.In our examination of specific demographic factors,we find that EDGAR financial statement usage is higher in areas with major cities, more accounting and finance jobs,higher capital gains and dividend income,greater access to broadband internet,a top100business school,or higher rates of college-educated https://www.sodocs.net/doc/db3306518.html,age is lower in ZIP codes with more fixed income,business income,retirees,unemployed workers,homeowners,or households with children.Overall,these results provide a general portrait of the users of financial statements hosted online on EDGAR.

Keywords:?nancial accounting;?nancial accounting usage;EDGAR;demographic.

INTRODUCTION

T his study presents a demographic portrait of the users of?nancial statements hosted online on EDGAR.We combine novel data that measure the amount of?nancial statement usage in EDGAR together with IP address data that measure the user’s physical location.1We combine these data with ZIP-code-level data from the U.S.Census Bureau(Census) and Internal Revenue Service(IRS),including demographic variables based on household,income,and education.2We use these data to explore whether?nancial statement usage varies in areas with different demographic attributes,which allows us to paint a general demographic portrait of?nancial statement users in the United States,as measured by the demographics of EDGAR users.

Our study is motivated by several related observations.First,accounting academics have expressed concern that prior research over emphasizes the need for?nancial statements to be correlated with capital market prices and volume(e.g.,Lev

We are grateful to Lynn L.Rees(editor)and two anonymous reviewers for excellent suggestions that improved the paper.We also thank David Wood for help with the heat map.We thank John Barrios,John Campbell,Cory Cassell,Peter Demerjian,Rebecca Files,Jeff Hoopes,Stephanie Rasmussen,and Katie Spangenberg,who provided helpful feedback on earlier drafts of this paper.We also thank Eric Haberkorn for outstanding computational assistance. Finally,we are grateful to the Marriott School of Management and the Foster School of Business for?nancial support.

Editor’s note:Accepted by Lynn L.Rees.

Submitted:June2016

Accepted:January2017

Published Online:April2017

1An important limitation is that our EDGAR data allow us to observe individuals‘‘accessing’’?nancial reports,but not the actual‘‘use’’of those reports.

Thus,an underlying assumption is that individuals who look at?nancial reports online,or download them,do so with a speci?c purpose in mind(i.e., that EDGAR access to?nancial statements proxies for?nancial statement usage).

2For example,for51566in Red Oak,Iowa,we combine all user access to?nancial statements via EDGAR originating from51566to the average household demographics for that particular ZIP code.

55

1989;Ball 2013),and have,in effect,called for new forms of accounting research that examine the usage of the ?nancial statements without necessarily tying that usage to market activity.Second,many have asserted that accounting reports can be viewed as a public good that is freely accessible to interested parties (e.g.,Beaver 1998;Gonedes and Dopuch 1974;Watts and Zimmerman 1979;Watts and Zimmerman 1986),and such a view often motivates policy.3Beyond anecdotes,however,we have very little demographic evidence on the types of individuals that use this public good.

Third,regulation is often directly aimed at ‘‘leveling the playing ?eld ’’to help ensure that individual investors are not at an informational disadvantage to institutional investors (Healy and Palepu 2001).Nevertheless,we have little evidence on the identity and attributes of these individual investors.Financial statement access in EDGAR is free of charge,which makes it uniquely suited to providing evidence on the activities of individual investors that are less likely to obtain ?nancial information from subscription-based providers.Discussions in prior literature about ?nancial statement users are often left to general statements about the ‘‘average,’’‘‘marginal,’’or ‘‘public ’’user.Regulators and standard setters likely have an end-user in mind for whom accounting disclosures are intended,but the concept is also general.For example,the Securities and Exchange Commission (SEC)states that it ‘‘requires public companies to disclose meaningful ?nancial and other information to the public ’’(emphasis added),which ‘‘provides a common pool of knowledge for all investors to use to judge for themselves whether to buy,sell,or hold a particular security ’’(emphasis added).4The Financial Accounting Standards Board (FASB)asserts in Objective 5of Statement of Financial Accounting Concepts No.8(FASB 2010)that ‘‘existing and potential investors,lenders,and others creditors ...are the primary users to whom general purpose ?nancial statements are directed.’’5These broad statements about end-users likely stem from the idea that general purpose ?nancial statements are intended to be just that—generally useful to a broad target audience.But an important and interesting question remains:To whom is the ?nancial accounting information generally useful?In other words,which types of individuals are consuming the information?

We use unique empirical data to examine the question of ‘‘who uses ?nancial statements?’’from a demographic perspective of EDGAR users.Our analyses rely on a revealed-preferences approach wherein we examine the demographic characteristics of users that reveal a preference for ?nancial statements by directly accessing the ?nancial statements hosted on the SEC’s EDGAR system.These data provide a proxy for ?nancial statement usage by providing the count of the number of annual and quarterly reports (i.e.,10-Ks and 10-Qs)downloaded directly from EDGAR in a particular geographic area (ZIP code).

Our analyses focus on four broad demographic categories:income,household characteristics,education,and local community conditions.On a relative basis,we ?nd that community characteristics explain the highest variation in ?nancial statement usage via EDGAR.6Aside from that category,we ?nd that education explains 32.8percent of the variation in ?nancial statement usage,which is signi?cantly greater than that explained by household characteristics and income,each of which explains approximately 28percent of the variation in ?nancial statement usage.

Next,we break each of the demographic categories into smaller attributes to examine in greater detail how they are associated with ?nancial statement usage via EDGAR.Regarding income,the usage of ?nancial statements via EDGAR is positively associated with dividend income and capital gains income and is negatively associated with interest income,retirement income,and unemployment income.These results are consistent with the idea that equity investors are more likely to use ?nancial statements,but that ?xed-income investors,including retirees that are more likely to invest in ?xed-income securities,are less likely to be accessing the statements.It is also interesting to observe that business owners are less likely to access reports.This ?nding may re?ect business owners’desire to invest discretionary income back into their own ventures.Alternatively,this ?nding may re?ect business owners’greater wealth,which allows them to use paid services for their ?nancial statement information.

Second,we examine the association of ?nancial statement usage via EDGAR with household characteristics,including homeownership,the presence of children in the household (i.e.,number of dependents),and whether a member of the household has served in the military.We ?nd that each of these variables is negatively related to ?nancial statement usage.The negative conditional association with homeownership is perhaps surprising given that homeownership represents a propensity to invest,albeit in real estate.Several plausible explanations exist for the negative association with children in the home,

3SEC Commissioner Luis Aguilar recently stated that ‘‘disclosure is,in economic terms,a ‘public good’in that its bene?ts are enjoyed broadly by the public ’’(see,https://https://www.sodocs.net/doc/db3306518.html,/news/speech/2013-spch120513-2laa#.VSKntvnF_h5).4See,https://https://www.sodocs.net/doc/db3306518.html,/Article/whatwedo.html

5

Although the SEC and the FASB discuss wanting to make ?nancial accounting information broadly available,the FASB notes in its Statement of Financial Accounting Concepts No.8(FASB 2010)that standard setters presume users of ?nancial statements have some knowledge of businesses and the transactions that businesses make (see Qualitative Characteristic 32).

6

This result is likely driven,however,by the positive association between community population and EDGAR downloads (i.e.,more populous ZIP codes will access more ?nancial statements on EDGAR).

56Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

including reduced free time and reduced discretionary income.We?nd weak evidence that households of military veterans tend to access?nancial statements less.

Third,we examine the association of?nancial statement usage via EDGAR with the levels of educational attainment in the ZIP codes,as self-reported to the U.S.Census Bureau via the American Community Survey(ACS).Financial statement usage is signi?cantly positively associated with the percent of individuals in a ZIP code that have a bachelor’s degree or graduate school education,relative to ZIP codes where users typically have not gone to college.Thus,our evidence is consistent with ?nancial statement usage increasing in higher education.

Finally,our analyses include general variables that capture elements of the local community conditions of each ZIP code, including the number of?nancial jobs,whether the region is a metropolitan area,whether the ZIP code is home to a top100 business school,the availability of broadband internet,and the local population.We?nd that,as a group,these local conditions explain62percent of the total variation in EDGAR usage and that each has a positive and signi?cant in?uence,with one exception.We?nd only modest evidence that EDGAR usage is associated with whether the ZIP code is in a metropolitan area.

In summary,we provide descriptive evidence on the types of individuals and households that access?nancial statements via EDGAR.To our knowledge,this study represents the?rst large-scale demographic analysis of individuals that access ?nancial statements.An improved understanding of the characteristics of the individuals that access?nancial reports from EDGAR provides the SEC,the federal agency that operates and maintains EDGAR,with information on who uses its public database.Our study on EDGAR users also provides a starting point for helping regulators and standard setters understand who is actually using the?nancial reports they regulate.In addition,our results speak to a particular form of?nancial literacy in which individuals access the freely available?nancial statements on EDGAR.Our results suggest that there are demographics of users who do and do not become?nancially informed by using the actual?nancial reports?led with the SEC and provided on EDGAR.

Our study is subject to several important caveats.First,there are many substitutes for EDGAR(e.g.,FactSet Research Systems Inc.,Bloomberg L.P.,Yahoo!Finance,and investor relations websites),which some demographic users might use more predominantly(e.g.,Lawrence,Ryans,Sun,and Laptev2017).Thus,our results represent demographic?nancial statement usage through a single online channel,EDGAR.Second,we measure?nancial statement usage using EDGAR downloads,rather than by observing individuals actually using10-K and10-Q?lings.Thus,insofar as individuals access speci?c EDGAR?lings,but do not use the?lings,our proxy for EDGAR usage will contain measurement error.Finally,our demographic measures capture demographic variation at the ZIP code level,rather than directly measuring the actual demographic differences of each particular user.We urge the reader to interpret our?ndings with these caveats in mind.

DATA AND RESEARCH DESIGN

EDGAR Request Data and Variables

Our data on?nancial statement usage come from the SEC’S EDGAR search database,which is used in several recent studies(e.g.,Drake,Roulstone,and Thornock2015;Drake,Roulstone,and Thornock2016;Lee,Ma,and Wang2015).The EDGAR search data are a record of the server log for the immense EDGAR database,which hosts all of the regulatory?lings mandated by the SEC,including annual and quarterly reports(10-K and10-Q),current reports(8-K),equity ownership updates (Form4),registration statements(S-1),and proxy statements(DEF14A),among others.The EDGAR search database is used in extant research to examine the extent,timing,and implications of investors’acquisition of?rm-level?nancial statements from EDGAR.To date,however,the EDGAR search data have been aggregated at the?rm level to examine investors’interest in a particular?rm(e.g.,Drake et al.2015)or at the?ling level to examine investors’interest in a given SEC?ling(e.g.,Lee et al.2015).For example,Drake et al.(2015)examine the?rm characteristics and events that drive information acquisition via EDGAR.They?nd that user activity on EDGAR is targeted at speci?c?rms(i.e.,large?rms with strong information environments)and events(i.e.,earnings announcements).They further provide evidence that EDGAR search is associated with price formation around earnings announcements,which suggests that equity investors use the database.Drake et al.(2015)do not,however,explore speci?c attributes of the users themselves.In this study,we focus on a different unit of observation; namely,we focus on the geographic region,as measured by U.S.postal ZIP codes,where the user resides.

We begin data assembly with the raw EDGAR server log?les acquired directly from the SEC for academic research purposes.These?les provide a record of all the activity that occurs on the servers that host the EDGAR data.They contain an IP address associated with each click on an EDGAR?ling,and this IP address allows us to estimate the geographic location of the user.For privacy reasons,the SEC has anonymized the last octet of the IP address,but the?rst three octets of the IP address make it possible to estimate the IP address location for many IP addresses.

We match this partial IP address to a proprietary database of IP address characteristics from MaxMind.The MaxMind database provides location-based variables such as country,state,ZIP code,and longitude/latitude,with varying degrees of Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR57

Accounting Horizons Volume31,Number3,

2017

coverage for each IP address.For our main analysis,we retain only EDGAR downloads that originate from IP addresses in the United States.Within the set of U.S.IP addresses,a large portion (over 80percent)is missing ZIP code data,but has information on longitude/latitude coordinates.Thus,to improve the coverage of our sample,we employ a geodesic search algorithm to map each coordinate pair to the nearest and most likely ZIP code or foreign country.With this algorithm,we are able to estimate the ZIP code or country for 99.9percent of EDGAR hits.MaxMind assigns U.S.downloads of unknown geographic origin to Potwin,KS,which is the approximate center of the contiguous United States.Approximately 3percent of all downloads are assigned to Potwin,and we retain these downloads for our country comparisons in Table 1,Panel B,but delete them for all other tables.

The matching procedure of the MaxMind IP address data to the IP addresses in the EDGAR database merits additional discussion.As noted above,the SEC only reports the ?rst three octets of the IP address,with the last octet anonymized with random letters (e.g.,10.7.88.kat).This convention is not particularly problematic for our purposes because the MaxMind database often reports locational characteristics for a range of IP addresses that are usually within the ?rst three octets (e.g.,10.7.88.000through 10.7.121.255).Thus,despite the anonymization of a particular IP address from the EDGAR database,a large majority are still identi?able because they fall within a range of identi?ed IP addresses from the MaxMind database.Finally,when the MaxMind range of IP addresses fails to cover the entire fourth octet,we use the ZIP code that occupies the largest fraction of the IP address range.

We also require that the data meet certain requirements.First,we are interested in identifying retail users of ?nancial statement information and,as such,we apply ?lters to remove institutions and automated web crawlers that download thousands of SEC ?lings.7Additionally,although it is possible that individual users could falsify or hide their source IP address to obscure their identity (i.e.,IP address spoo?ng),we believe the extent of IP address spoo?ng in our sample is limited.8Finally,because we are inherently interested in the usage of accounting ?nancial statements,we retain only search activity for 10-Ks or 10-Qs,excluding search activity for other types of SEC ?ling types (e.g.,Registration Statements or Form 4s).

From these combined data,we create the two variables that serve as the primary dependent variables in this study.First,we create a measure of ?nancial statement usage in the ZIP code,labeled FINANCIAL USAGE,which is the natural logarithm of 1plus the total number of requests on EDGAR for 10-K or 10-Q (for any company or ?scal year)that originates from a particular ZIP code.9FINANCIAL USAGE serves as a proxy for ?nancial statement usage in a geographic region.Second,we create FINANCIAL USERS,which is the natural logarithm of 1plus the total number of unique IP addresses from a given ZIP code that acquired a 10-K or 10-Q for any company or ?scal year.FINANCIAL USAGE measures the construct of the depth of ?nancial statement usage (i.e.,one user gathering many ?nancial statements),and FINANCIAL USERS measures the construct of the breadth of ?nancial statement usage.Demographic Data and Variables

To measure the demographic characteristics of ?nancial statement users,we gather demographic data from two https://www.sodocs.net/doc/db3306518.html,ernment sources:the U.S.Census Bureau and the IRS.10We focus on a single year,2011,because of several data limitations that we detail below.We focus on four broad demographic categories:income,household characteristics,education,and local conditions,which we now discuss.

We rely on IRS data for all measures of income from a given ZIP code.11We calculate all of our income variables by taking the total amount (in thousands)of the particular income type for the ZIP code divided by the total number of returns for the ZIP code,as reported by the IRS in 2011.We measure SALARIES/WAGES using total Adjusted Gross Income,INTEREST INCOME as the total amount of taxable interest for those ?lers that reported non-negative interest income,DIVIDEND INCOME as the sum of ordinary and quali?ed dividends for those ?lers that reported non-negative dividend income,CAPITAL GAINS as the sum of the Schedule D capital gains,RETIREMENT INCOME as the sum of taxable social security bene?ts and

7

Speci?cally,in the spirit of Drake et al.(2015)and Lee et al.(2015),we exclude searches made by high-frequency downloaders (i.e.,IP addresses that search more than ?ve times a minute or more than 1,000times a day).The remaining observations,therefore,are more likely to be individuals searching for ?nancial statements.

8Indeed,others using these same data have some evidence of thousands of downloads from major corporations and major government entities,suggesting that even among sophisticated stakeholders,IP address spoo?ng is limited (Bozanic,Hoopes,Thornock,and Williams 2017).

9We follow prior literature (e.g.,Drake et al.2016)and take the natural logarithm of the EDGAR ?nancial statement usage variables to reduce the in?uence of skewness.

10

Census and IRS data are self-reported data,which make them susceptible to concerns about misreporting or nonreporting.In contrast to other survey data,however,failing to report data to the Census or the IRS is illegal and can lead to legal penalties.For example,U.S.Code Section 221allows the government to levy ?nes of up to $100for refusing to report data to the Census and ?nes up to $500for willfully providing false information.Failing to ?le taxes,as well as providing false information to the IRS,may also result in signi?cant ?nes.

11

We acquired the data from the following IRS website:https://https://www.sodocs.net/doc/db3306518.html,/uac/SOI-Tax-Stats-Individual-Income-Tax-Statistics-2011-ZIP-Code-Data-(SOI )

58Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

taxable pensions and annuities,BUSINESS INCOME as the amount of sole-proprietor business income,and UNEMPLOYMENT INCOME as the total amount of unemployment income.

We acquire household demographics from various sources.We focus on homeownership,number of children,and veteran status.From the IRS database,we create the following variables:HOME OWNERS,which is the percent of tax returns in the ZIP code that report a deduction for mortgage interest;and CHILDREN,which is the total number of dependents in the ZIP

TABLE 1

Most Active EDGAR Access Locations

Panel A:Top Ten ZIP Codes,by EDGAR Downloads

Rank Location ZIP Financial Usage (Raw)

1Dallas,TX 75244271,0092New York,NY 10007174,0103New York,NY 10022124,7364New York,NY 1001391,2465Chicago,IL 6061683,6716New York,NY 1001179,0917New York,NY 1001978,4268New York,NY 1001677,2339San Francisco,CA 9410477,13610

Dallas,TX

75202

68,885

Panel B:Top 30Countries,by EDGAR Downloads

Rank Location Financial Usage (Raw)

Proportion of Total

1United States 4,920,52667.67%2India 538,5247.41%3Canada

340,300 4.68%4United Kingdom 250,775 3.45%5China

190,356 2.62%6Hong Kong 101,922 1.40%7Japan 99,427 1.37%8Germany 89,571 1.23%9Taiwan 69,3370.95%10Singapore 68,1880.94%11Australia 63,6980.88%12Mexico

54,4410.75%13The Netherlands 50,7430.70%14France 44,0740.61%15Philippines 43,7680.60%16Switzerland 42,6580.59%17Israel

37,6950.52%18Republic of Korea 36,8290.51%19Russian Federation 35,6200.49%20Egypt 24,0630.33%21Malaysia 23,8130.33%22Argentina 23,4780.32%23Italy 22,5210.31%24Belgium 19,9590.27%25Spain 18,2560.25%26Brazil 17,7210.24%27Sweden 13,0700.18%28Ireland 11,6950.16%29Indonesia 9,0910.13%30

Turkey

8,822

0.12%

Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR 59

Accounting Horizons

Volume 31,Number 3,

2017

code,scaled by the total number of tax returns in the ZIP code.Finally,to measure VETERAN,we obtain Census data to measure the percentage of the civilian population over 25years old in the ZIP code who are veterans.

Next,we obtain data on the average education in the ZIP code,as reported by the Census in its American Community Survey (ACS).12The ACS provides researchers with one-,three-,and ?ve-year estimates of demographic characteristics.ZIP-code-level data are only available in the ?ve-year estimates and,thus,we use the 2011?ve-year estimates.ACS bases its ?ve-year estimates,which represent the average characteristics of a ZIP code,on the survey information it collects on a near-daily basis over the ?ve years.While the multi-year estimates,such as the ?ve-year estimates,are less current than the single-year estimates,the multi-year estimates also come from a much larger sample than the single-year estimates.The much larger sample size means the multi-year estimates are the most reliable estimates.We examine three levels of education:some college,graduated with a bachelor’s degree,and graduated with a graduate degree (‘‘no college ’’is included in the intercept of our empirical models).The variables are labeled as SOME COLLEGE,BACHELORS,and GRAD SCHOOL .Each of these education variables represents the percentage of individuals who are over 25years of age in the ZIP code with that level of education.

In our empirical models,we also include several variables that measure the general conditions in the ZIP code.We obtain the number of ?nance and accounting jobs in the ZIP code (#FINANCIAL JOBS )from the Census.Following Dyreng,Mayew,and Williams (2012),we compute MAJOR METRO as an indicator variable equal to 1if the ZIP code falls in one of the top ten largest metropolitan areas in the United States:Baltimore,Boston,Chicago,Dallas,Detroit,Houston,Los Angeles,New York,Philadelphia,and San Francisco.We measure TOP 100BUSINESS SCHOOL as an indicator for whether the ZIP code contains an M.B.A.program ranked in the top 100by the 2015U.S.News &World Report .

We are also concerned that our EDGAR usage variables are,in part,capturing differences in internet availability or population across ZIP codes.A report from the U.S.Census Bureau on internet use across the country ?nds that internet usage is higher for individuals with higher incomes and education levels,which are attributes we also consider in this study (U.S.Census Bureau 2013).To control for cross-sectional differences in internet availability,we obtain data on the percentage of households with high-speed internet access by county from the Department of Commerce.We label this variable INTERNET ACCESS .We also include the log of population (POPULATION )as reported by the Census in 2010,as well as state ?xed effects.Sample Period

Our analyses focus on a single year—2011.We make this research design choice for three important reasons.First,ACS data for ZIP codes are simply unavailable before 2010.While data are available starting in 2006for larger geographic regions (i.e.,those with a population exceeding 65,000people),ZIP-code-level data are unavailable before 2010.Moreover,the Census reports that the data before 2010are based on smaller survey sizes and are less reliable to researchers than the ?ve-year ACS estimates that ACS provides starting in 2010,which we employ (Census 2011).Second,as a result of budget cuts to the United States Postal Service,the number of ZIP codes consistently declines over time.Thus,each year ZIP code boundaries are changing as they are consolidated,making it dif?cult to draw inference from time-series analyses.Third,the IP address data from MaxMind provide the locational information based on the most updated data—a historical time-series of IP address ranges is unavailable.Because of these data challenges,we limit analyses to cross-sectional tests for a single year in time,2011.Given that variation in most of our demographic variables is largely cross sectional in nature (rather than time series in nature),we do not feel that this design choice unduly limits our ability to draw inferences.Geographic Portrait

In Figure 1,we present a heat map of EDGAR usage across the United States.The ?gure displays the number of EDGAR searches in 2011by ZIP code,scaled by the population of the ZIP code to reduce the in?uence of population size on the visual depiction.In Table 1,Panel A we tabulate the top ten ZIP codes in terms of EDGAR downloads.These ZIP codes come from four major ?nancial centers including New York,Dallas,Chicago,and San Francisco.Although data limitations restrict our focus to EDGAR usage within the U.S.,we tabulate the top 30countries in Table 1,Panel B.We ?nd that individuals in the

12

The ACS is different from the decennial census in its residence rules.The ACS collects information on a near-daily basis using a current residence concept.The current resident concept counts an individual as a resident of wherever that individual is as of the time of survey,as long as the time of the residence exceeds two months.The current resident concept allows the ACS to estimate annual averages of the characteristics of all areas.In contrast,the decennial census uses a usual residence concept.For the 2000census,the survey asked individuals where they lived as of April 1,2000.Surveys are subject to various types of nonsampling errors,and the U.S.Census Bureau makes a number of corrections to mitigate nonsampling errors.Nonsampling errors include coverage errors,errors arising from complete or partial nonresponse,response errors,and processing errors.

60Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

U.S.account for two-thirds of all non-machine EDGAR downloads globally.The U.S.is followed by India (7.4percent),Canada (4.7percent),the U.K.(3.5percent),and China (2.6percent).Descriptive Statistics and Correlations

Table 2,Panel A reports the descriptive statistics for the variables included in the analyses.To limit the in?uence of extreme values on our inferences,we winsorize our dependent and independent variables,other than our indicator variables,at the 1st and 99th percentiles.We ?rst focus on FINANCIAL USAGE,for which there is an average of 439downloads of ?nancial statements per ZIP code in 2011.The mean FINANCIAL USERS is 53,suggesting that,on average 53unique users/households per ZIP code acquired ?nancial statements in 2011.The standard deviation of these variables suggest that there is considerable cross-sectional variation in EDGAR downloads.

Our income variables are from the IRS and are reported in thousands of dollars for that ZIP code,divided by the total number of returns for that ZIP code.An examination of our income variables reveals that the mean interest income is $650,the mean dividend income is $1,660,the mean business income is $1,870,the mean capital gains is $1,410,the mean retirement income is $5,530,and the mean unemployment income is $620.Across the income variables,the standard deviations and variable ranges also reveals that the types of income vary considerably across ZIP codes.

In terms of household attributes and educational attainment,Table 2reveals,on average,23.0percent of households in a ZIP code are homeowners.The average number of dependents,which are most commonly children,per household is 0.66,and the mean value for VETERAN is approximately 11percent.Moving now to educational attainment,Table 2reports that 29percent of heads of households have some college education,16percent graduated with a bachelor’s degree,and 9percent have a graduate degree.Finally,46percent of heads of household in our sample have no college education.

Finally,the ZIP codes in our sample have an average of 7.67?nancial jobs,and 2.8percent of households fall within the identi?ed major metropolitan areas (MAJOR METRO ).We further note that very few of the ZIP codes in our sample have a top 100business school (mean TOP 100BUSINESS SCHOOL ?0.003when taken out to three decimal places)and that on average 85percent of the individuals in ZIP codes have access to broadband internet.The average population size is 9,060.

FIGURE 1

Heat Map of the EDGAR Downloads,Scaled by

Population

Source:EDGAR server log ?les and MaxMind database.

Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR 61

Accounting Horizons

Volume 31,Number 3,

2017

T A B L E 2

D e s c r i p t i v e S t a t i s t i c s a n d C o r r e l a t i o n s

P a n e l A :D e s c r i p t i v e S t a t i s t i c s

V a r i a b l e

n

M e a n S t d .D e v .M i n Q 1M e d i a n Q 3M a x

R A W D O W N L O A D S 15,591439.311141.7905403097,781F I N A N C I A L U S A G E 15,5913.812.3601.793.715.748.96R A W U N I Q U E I P 15,59152.5892.76031361554F I N A N C I A L U S E R S 15,5912.741.6701.392.644.136.32S A L A R I E S /W A G E S 15,59139.2417.1617.7228.9334.1743.74117.68I N T E R E S T I N C O M E 15,5910.650.770.040.290.450.715.56D I V I D E N D I N C O M E 15,5911.662.890.020.380.781.5819.91C A P I T A L G A I N S 15,5911.413.48à0.050.140.441.0925.42R E T I R E M E N T I N C O M E 15,5915.532.480.913.955.166.7314.5B U S I N E S S I N C O M E 15,5911.871.560.10.981.412.149.68U N E M P L O Y M E N T I N C O M E 15,5910.620.300.40.570.791.53H O M E O W N E R S 15,5910.230.1200.140.220.320.51C H I L D R E N 15,5910.660.190.220.550.640.751.33V E T E R A N 15,5910.110.040.020.080.10.130.23S O M E C O L L E G E 15,5910.290.070.120.240.290.330.46B A C H E L O R S 15,5910.160.090.030.090.140.210.41G R A D S C H O O L 15,5910.090.0800.040.070.120.38#F I N A N C I A L J O B S 15,5917.672.153.535.887.539.4511.98M A J O R M E T R O 15,5910.030.1600001T O P 100B U S I N E S S S C H O O L 15,59100.0500001I N T E R N E T A C C E S S 15,5910.850.150.240.790.910.961P O P U L A T I O N

15,5919.061.26.018.199.2110.0611.11

P a n e l B :C o r r e l a t i o n s

(1)

(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)

(12)

(13)(14)(15)(16)(17)(18)(19)(20)

(1)F I N A N C I A L U S A G E 0.95

0.360.200.380.310.000.350.070.30à0.04à0.34à0.100.480.480.630.190.070.420.75(2)F I N A N C I A L U S E R S 0.950.350.200.380.310.010.350.080.30à0.03à0.33à0.080.470.470.620.190.070.420.79(3)S A L A R I E S /W A G E S 0.380.360.450.560.500.380.530.000.80à0.14à0.140.000.720.650.390.010.010.270.20(4)I N T E R E S T I N C O M E 0.260.240.630.830.780.550.59à0.150.40à0.400.06à0.050.550.550.08à0.030.010.070.08(5)D I V I D E N D I N C O M E 0.310.290.670.910.770.540.65à0.100.52à0.43à0.01à0.070.690.720.280.020.030.180.24(6)C A P I T A L G A I N S 0.280.260.640.880.890.360.64à0.220.38à0.26à0.03à0.070.590.570.140.020.030.110.21(7)R E T I R E M E N T I N C O M E à0.010.000.340.440.420.290.310.030.55à0.420.390.150.370.45à0.04à0.15à0.03à0.02à0.07(8)B U S I N E S S I N C O M E 0.330.310.750.750.780.740.37à0.080.43à0.17à0.20à0.190.600.580.310.110.020.230.25(9)U N E M P L O Y M E N T I N C O M E 0.07

0.07à0.03à0.14à0.13à0.14à0.01à0.08

0.17à0.05à0.080.02à0.09à0.070.190.02à0.040.180.11

(c o n t i n u e d o n n e x t p a g e )

62

Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

T A B L E 2(c o n t i n u e d )

(1)

(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)

(10)H O M E O W N E R S 0.290.290.710.300.340.260.510.440.16à0.140.000.150.660.610.35à0.09à0.040.280.20(11)C H I L D R E N à0.05à0.04à0.12à0.28à0.27à0.19à0.44à0.170.00à0.14à0.26à0.02à0.31à0.35à0.010.00à0.05à0.040.19(12)V E T E R A N à0.33à0.32à0.20à0.04à0.08à0.110.41à0.19à0.090.00à0.260.37à0.15à0.13à0.37à0.22à0.07à0.27à0.30(13)S O M E C O L L E G E à0.13à0.11à0.22à0.23à0.28à0.260.12à0.310.010.10à0.040.39à0.01à0.09à0.09à0.15à0.04à0.11à0.01(14)B A C H E L O R S 0.490.480.760.520.570.490.340.64à0.100.65à0.31à0.17à0.140.810.460.050.040.340.29(15)G R A D S C H O O L 0.460.450.760.600.670.580.420.72à0.110.55à0.32à0.19à0.300.810.420.050.060.310.29(16)#F I N A N C I A L J O B S 0.630.620.400.220.280.24à0.040.330.190.340.02à0.36à0.110.460.420.260.050.570.55(17)M A J O R M E T R O 0.210.200.090.110.130.16à0.140.130.01à0.090.01à0.22à0.170.080.110.290.080.160.16(18)T O P 100B U S I N E S S S C H O O L 0.080.090.020.030.040.06à0.030.03à0.04à0.04à0.06à0.07à0.050.050.090.050.080.040.03(19)I N T E R N E T A C C E S S 0.360.360.240.110.160.13à0.010.190.140.27à0.02à0.19à0.070.300.270.490.110.030.34(20)P O P U L A T I O N 0.720.760.190.080.100.09à0.060.150.130.220.20à0.29

à0.020.260.220.540.150.030.29

P a n e l A p r e s e n t s t h e d e s c r i p t i v e s t a t i s t i c s f o r t h e v a r i a b l e s u s e d i n o u r a n a l y s e s .T h e u n i t o f a n a l y s i s i s t h e Z I P c o d e a n d t h e s a m p l e y e a r i s 2011.P a n e l B p r e s e n t s t h e c o r r e l a t i o n s a m o n g t h e v a r i a b l e s ,

w i t h P e a r s o n a b o v e a n d S p e a r m a n b e l o w t h e d i a g o n a l .

V a r i a b l e D e ?n i t i o n s :F I N A N C I A L U S A G E (R A W D O W N L O A D S )?t h e l o g g e d (r a w )n u m b e r o f 1p l u s 10-K a n d 10-Q d o w n l o a d s o r i g i n a t i n g f r o m a g i v e n Z I P c o d e ;F I N A N C I A L U S E R S (R A W U N I Q U E I P )?t h e l o g g e d (r a w )n u m b e r o f 1p l u s u n i q u e I P a d d r e s s e s o r i g i n a t i n g f r o m a p a r t i c u l a r Z I P c o d e ;S A L A R I E S /W A G E S ,I N T E R E S T I N C O M E ,D I V I D E N D I N C O M E ,C A P I T A L G A I N S ,R E T I R E M E N T I N C O M E ,B U S I N E S S I N C O M E ,a n d U N E M P L O Y M E N T I N C O M E ?m e a s u r e s o f d i f f e r e n t f o r m s f o r h o u s e h o l d i n c o m e f r o m t h e I R S ,c o m p u t e d a s t h e m e a n (i n t h o u s a n d s o f d o l l a r s )p e r t a x r e t u r n i n t h a t Z I P c o d e ;S A L A R I E S /W A G E S ?t h e t o t a l s a l a r i e s a n d w a g e s ;I N T E R E S T I N C O M E ?t o t a l i n t e r e s t i n c o m e ;D I V I D E N D I N C O M E ?t o t a l o r d i n a r y a n d q u a l i ?e d d i v i d e n d i n c o m e ;C A P I T A L G A I N S ?t o t a l n e t c a p i t a l g a i n s ;R E T I R E M E N T I N C O M E ?t h e s u m o f t a x a b l e s o c i a l s e c u r i t y b e n e ?t s a n d t a x a b l e p e n s i o n s a n d a n n u i t i e s ;B U S I N E S S I N C O M E ?t h e a m o u n t o f s o l e -p r o p r i e t o r b u s i n e s s i n c o m e ;U N E M P L O Y M E N T I N C O M E ?t o t a l u n e m p l o y m e n t c o m p e n s a t i o n ;H O M E O W N E R S ?t h e n u m b e r o f r e t u r n s w i t h m o r t g a g e i n t e r e s t d i v i d e d b y t h e n u m b e r o f r e t u r n s f r o m I R S ;C H I L D R E N ?t h e n u m b e r o f d e p e n d e n t s d i v i d e d b y n u m b e r o f r e t u r n s f r o m t h e I R S ;V E T E R A N ?t h e p e r c e n t o f c i v i l i a n s o v e r 25i n t h e Z I P c o d e w h o a r e v e t e r a n s ;S O M E C O L L E G E ?t h e p e r c e n t a g e o f a d u l t s 25o r o l d e r w i t h s o m e c o l l e g e e d u c a t i o n ,b u t n o b a c h e l o r ’s d e g r e e f r o m t h e 2011A C S ;B A C H E L O R S ?t h e p e r c e n t a g e o f a d u l t s 25o r o l d e r w i t h a b a c h e l o r ’s d e g r e e ,b u t n o t a g r a d u a t e d e g r e e f r o m t h e 2011A C S ;G R A D S C H O O L ?t h e p e r c e n t a g e o f a d u l t s 25o r o l d e r w i t h a g r a d u a t e d e g r e e f r o m t h e 2011A C S ;#F I N A N C I A L J O B S ?t h e n u m b e r o f F i n a n c e a n d I n s u r a n c e j o b s ,f r o m t h e C e n s u s ’s d e t a i l e d Z I P C o d e B u s i n e s s P a t t e r n s ,f o l l o w i n g t h e f u l l s i x -d i g i t N o r t h A m e r i c a n I n d u s t r y C l a s s i ?c a t i o n S y s t e m ;M A J O R M E T R O ?a n i n d i c a t o r v a r i a b l e e q u a l t o 1f o r o b s e r v a t i o n s i n o n e o f t h e t e n l a r g e s t m e t r o p o l i t a n a r e a s (H o u s t o n ,D a l l a s ,D e t r o i t ,B o s t o n ,P h i l a d e l p h i a ,S a n F r a n c i s c o ,B a l t i m o r e ,C h i c a g o ,L o s A n g e l e s ,a n d N e w Y o r k ),0o t h e r w i s e ;T O P 100B U S I N E S S S C H O O L ?a n i n d i c a t o r v a r i a b l e e q u a l t o 1i f t h e Z I P c o d e c o n t a i n s a U .S .N e w s &W o r l d R e p o r t T o p 100M .B .A .S c h o o l ,a n d 0o t h e r w i s e ;I N T E R N E T A C C E S S ?t h e p e r c e n t a g e o f h o u s e h o l d s i n a p a r t i c u l a r Z I P c o d e w i t h D S L a c c e s s a s o f J u n e 30,2011f r o m t h e U .S .D e p a r t m e n t o f C o m m e r c e ;a n d P O P U L A T I O N ?t h e n a t u r a l l o g a r i t h m o f t h e p o p u l a t i o n f r o m t h e 2010U .S .c e n s u s .

Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR

63

Accounting Horizons

Volume 31,Number 3,

2017

Table 2,Panel B provides the Pearson (below)and Spearman (above)correlations.For the sake of brevity,we focus our discussion here primarily on the correlations of test variables with FINANCIAL USAGE .We see that FINANCIAL USAGE is strongly positively associated with FINANCIAL USERS,with a correlation coef?cient of 0.95.FINANCIAL USAGE and FINANCIAL USERS are most strongly associated with POPULATION,which has a correlation coef?cient above 72percent,followed by #FINANCIAL JOBS with a correlation of 63percent.FINANCIAL USAGE and FINANCIAL USERS are also highly positively correlated with BACHELORS,GRAD SCHOOL,and SALARIES/WAGES .We note that many of the demographic variables are highly correlated.As such,in our multiple-variable analyses,we are mindful of the potential for multicollinearity in our models,and we present related diagnostic statistics.

TESTS AND RESULTS

We now present the tests and results of our demographic analysis of ?nancial statement usage.We begin by examining the different demographic categories in isolation—that is,we separately estimate the association between ?nancial statement usage and income,household characteristics,and education.In these analyses,our focus is on the relative explanatory power of each category.We recognize that by focusing on a speci?c set of variables,we induce a correlated omitted variable problem that confounds the interpretation of individual coef?cient estimates.Thus,we do not report coef?cient estimates and instead focus on the relative explanatory power of each category.We believe that comparing the explanatory power of each category sheds light on the factors associated with ?nancial statement usage.In the ?nal analysis,we examine all of the demographic variables together in a comprehensive regression.We draw our primary inferences with respect to associations of particular variables from the estimation results of that comprehensive model.

Our ?rst model focuses on ?nancial statement usage and household income.We examine the association using seven categories of income.Speci?cally,we estimate the following two cross-sectional OLS regressions (i indexes ZIP codes):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1SALARIES =WAGES i tb 2INTEREST INCOME i

tb 3DIVIDEND INCOME i tb 4CAPITAL GAINS i

tb 5RETIREMENT INCOME i tb 6BUSINESS INCOME i tb 7UNEMPLOYMENT INCOME i te i :

e1T

In the above model,and in all subsequent models,state ?xed effects (b STATE )are included in the model.13We also cluster the standard errors by state.

Our second model examines associations between household characteristics and ?nancial statement usage.We estimate a model similar to Equation (1),but replace the income variables with the household characteristics as follows (i indexes ZIP codes):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1HOME OWNERS i tb 2CHILDREN i tb 3VETERAN i te i :

e2T

In the third model,we examine household education levels by estimating the following cross-sectional OLS regression (i indexes ZIP codes):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1SOME COLLEGE i tb 2BACHELORS i

tb 3GRAD SCHOOL i te i :

e3T

Note that NO COLLEGE (i.e.,the percentage of households with up to a high school diploma)is omitted from the regression and is captured in the intercept.Thus,the coef?cients on the three education variables (b 1–b 3)should be interpreted relative to the omitted group.

Finally,we examine a model that includes variables that measure different elements of local conditions within the ZIP code.We use the following cross-sectional OLS regression (i indexes ZIP codes):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1#FINANCIAL JOBS i tb 2MAJOR METRO i

tb 3TOP 100BUSINESS SCHOOL i tb 4INTERNET ACCESS tb 5POPULATION i te i :

e4T

We estimate Models (1)through (4)and present the results in Table 3in their respective columns.The dependent variables in Panels A and B of Table 3are FINANCIAL USAGE and FINANCIAL USERS,respectively.We focus our discussion on the

13

The results of,and inferences from,Table 3are largely consistent if we omit the state ?xed effects.

64Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

FINANCIAL USAGE results here and throughout the paper,because the results are generally consistent across both variables.In Table 3,our primary focus is on the explanatory power of the different categories.At the bottom on each panel,we present Vuong (1989)test statistics to investigate whether the explanatory power is signi?cantly different across our models.We ?nd that income explains 27.6percent of the variation in EDGAR usage.Similarly,household characteristics explain 28.1percent of the variation.The Vuong (1989)test reveals that the explanatory power of these two groups is statistically different.We also ?nd that education explains 32.8percent of the variation in EDGAR usage,which is statistically greater than that explained by household characteristics (and income).Finally,in Column (4)we ?nd that local conditions explain a far greater percentage of EDGAR usage than do the other three demographics.14Approximately 62.3percent of the variation in EDGAR usage is

TABLE 3

The Relative Explanatory Power of Demographic Attributes for Financial Statement Usage

Panel A:FINANCIAL USAGE as the Dependent Variable

Model (1)

Model (2)Model (3)Model (4)

Adjusted R 2

0.276

0.281

0.328

0.623

Vuong test for difference in R 2Model (1)versus Model (2)

Model (2)versus Model (3)

Model (3)versus Model (4)

Z-stat à5.6à8.27à40.94p-value

0.00

0.00

0.00

Panel B:FINANCIAL USERS as the Dependent Variable

Model (1)

Model (2)

Model (3)Model (4)

Adjusted R 2

0.261

0.28

0.323

0.662

Vuong test for difference in R 2Model (1)versus Model (2)

Model (2)versus Model (3)

Model (3)versus Model (4)

Z-stat à7.41à7.94à46.66p-value

0.00

0.00

0.00

This table presents the relative explanatory power from regressing ?nancial statement usage on attributes grouped into demographic categories.The dependent variable in Panel A,FINANCIAL USAGE,is the log of 1plus the number of 10-K and 10-Q downloads originating from the ZIP code.The dependent variable in Panel B,FINANCIAL USERS,is the log of 1plus the number of unique IP addresses originating from a particular ZIP code.The models have 15,591observations.State ?xed effects are included,and p-values are based on two-tailed tests with standard errors clustered by state.All other variables are de?ned in Table 2.

The speci?c models represented in each column are as follows:Model (1):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1SALARIES =WAGES i tb 2INTEREST INCOME i tb 3DIVIDEND INCOME i

tb 4CAPITAL GAINS i tb 5RETIREMENT INCOME i tb 6BUSINESS INCOME i tb 7UNEMPLOYMENT INCOME i te i :Model (2):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1HOME OWNERS i tb 2CHILDREN i tb 3VETERAN i te i :Model (3):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1SOME COLLEGE i tb 2BACHELORS i tb 3GRAD SCHOOL i te i :Model (4):

FINANCIAL USAGE i or FINANCIAL USERS i ?b STATE tb 1#FINANCIAL JOBS i tb 2MAJOR METRO i tb 3TOP 100BUSINESS SCHOOL i

tb 4INTERNET ACCESS tb 5POPULATION i te i :

14

ZIP code population (POPULATION )explains more variation in EDGAR ?nancial statement usage than does any other variable,which is consistent with high-population ZIP codes having more people who can download ?nancial statements.Thus,our local conditions variables can also be interpreted as control variables.

Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR 65

Accounting Horizons

Volume 31,Number 3,

2017

explained by the number of ?nancial jobs,whether the ZIP code is in a major metropolitan area,whether the ZIP code is home to a top 100business school,the availability of high-speed internet access,and the ZIP code population.

In our ?nal analyses,we combine Models (1)through (4)into a single model.We use this model to investigate the association between the speci?c components of each category and ?nancial statement usage.We present the results in Table 4.In Column (1),we present the results using FINANCIAL USAGE,and in Column (2),we present the results using FINANCIAL USERS .Given the large number of correlated variables in the model,we ?rst examine whether multicollinearity is likely to be in?uencing the estimation results by calculating variance in?ation factors (untabulated).We ?nd that the variance in?ation factors are identical across the two models and that they range from 1.04(TOP 100BUSINESS SCHOOL )to 9.13(DIVIDEND INCOME ).The average factor is 3.55.Given that the average factor is low and that none of the factors are greater than the common 10.0threshold,we conclude that multicollinearity is unlikely to be signi?cantly impacting the results.

We focus our discussion of speci?c associations using the results in Column (1).With respect to income,we ?nd that DIVIDEND INCOME and CAPITAL GAINS are both positively associated with FINANCIAL USAGE,and that INTEREST INCOME,RETIREMENT INCOME,BUSINESS INCOME,and UNEMPLOYMENT INCOME all exhibit negative associations.We ?nd no evidence that SALARIES/WAGES are associated with ?nancial statement usage in the model that includes all of the demographic variables.These results suggest that individuals invested in equities are more likely to use ?nancial statements and that individuals invested in ?xed-income securities are less likely to use them.Relatedly,retirees,who are most likely to be living on ?xed retirement incomes,also appear to use ?nancial information less,as do business owners.

TABLE 4

The Association of Financial Statement Usage with Cross-Sectional Demographic Attributes

Model (1)Model (2)Coeff.

t-stat Coeff.t-stat Income

SALARIES/WAGES 0.003(0.91)à0.002(à0.63)INTEREST INCOME à0.163***(à2.74)à0.150***(à4.56)DIVIDEND INCOME 0.075***(4.88)0.044***(3.96)CAPITAL GAINS

0.028***(2.97)0.030***(5.49)RETIREMENT INCOME à0.116***(à8.82)à0.076***(à8.93)BUSINESS INCOME

à0.091***(à4.29)à0.058***(à4.28)UNEMPLOYMENT INCOME à0.673***(à7.87)à0.486***(à7.78)Household

HOME OWNERS à1.102***(à2.72)à0.332(à1.25)CHILDREN à1.784***(à14.15)à1.225***(à14.39)VETERAN à1.213*(à1.87)à0.466(à1.10)Education

SOME COLLEGE 0.182(0.46)0.172(0.59)BACHELORS 3.408***(10.27) 2.364***(10.47)GRAD SCHOOL 3.676***(6.22) 2.762***(7.25)Local Conditions

#FINANCIAL JOBS 0.177***(17.58)0.092***(14.15)MAJOR METRO

0.228(1.62)0.168*(1.83)TOP 100BUSINESS SCHOOL 0.616**(2.18)0.598***(2.82)INTERNET ACCESS 0.252***(2.69)0.240***(3.30)POPULATION 1.147***(55.27)

0.891***(49.87)

State Fixed Effects Yes Yes S.E.Clustered by State State Observations 15,591

15,591

Adjusted R 2

0.694

0.726

*,**,***Indicate coef?cient estimates that are statistically signi?cant in a two-tailed test at the 10percent,5percent,and 1percent levels,respectively.The dependent variable in Model (1),FINANCIAL USAGE,is the log of 1plus the number of 10-K and 10-Q downloads originating from the ZIP code.The dependent variable in Model (2),FINANCIAL USERS,is the log of 1plus the number of unique IP addresses originating from a particular ZIP code.State ?xed effects are included,but unreported.Robust t-statistics,based on standard errors clustered by state,are reported in parentheses.All other variables are de?ned in Table 2.

66Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

The household characteristics variables reveal that homeownership and the presence of children in the home are both negatively associated with?nancial statement usage—the coef?cients on HOME OWNERS and CHILDREN are negative and signi?cant.This result suggests that homeowners and households with less discretionary time and money are less likely to access?nancial statements.We also?nd modest evidence that households with veterans are less likely to use?nancial statements.

Next,we examine education.We?nd a clear positive association between higher levels of education and?nancial statement usage.More speci?cally,although SOME COLLEGE is not signi?cantly associated with FINANCIAL USAGE,both BACHELORS and GRAD SCHOOL have positive and signi?cant coef?cients.These results may suggest that individuals with increasing levels of education are able to extract greater bene?ts from?nancial reports.

Finally,we?nd that?nancial statement usage is higher in areas with a top100business school,15areas with more accounting and?nance jobs,areas with greater access to high-speed internet,and areas with a higher population.

CA VEATS AND CONCLUSION

In summary,we address the question of who uses the?nancial statements from a demographic https://www.sodocs.net/doc/db3306518.html,ing ZIP-code-level data on both?nancial statement usage and demographics,we can make approximate inferences about household characteristics of EDGAR’s?nancial statement users.We acknowledge several shortcomings of the study and provide the following caveats.First,we cannot link?nancial statement usage directly to a particular user or that user’s demographics.Data constraints restrict our examination to the general demographic characteristics of the area where the user accesses the?nancial statements.This limitation prevents us from making statements about speci?c individuals or demographics,such as,‘‘Households with children use?nancial statements less.’’Instead,we can only infer that?nancial statement usage is lower in areas where households are more likely to have children.Thus,we acknowledge this limitation of our data and urge caution in generalizing our results to speci?c individuals or households.

Second,we do not link?nancial statement usage to any particular?rm’s?nancial statements.We make this choice intentionally,acknowledging the costs of doing so,because we are speci?cally focused on the human demographics of?nancial statement usage,rather than the?rm characteristics of?nancial statement usage(which,as we note in the‘‘Introduction’’section,have been examined in great detail in previous research).

Finally,we have?nancial statement usage only as proxied by downloads from one particular source of?nancial reports—https://www.sodocs.net/doc/db3306518.html,ers can acquire the?nancial statements from a number of different locations online,including investor relations websites,free data aggregators(e.g.,Yahoo!Finance),and fee-based data aggregators(e.g.,Bloomberg).We recognize that certain types of?nancial statement users will gravitate toward certain?nancial information providers online.Because our analyses are based on?nancial statement acquisition from just one source,caution should be exercised when interpreting the results and drawing policy implications.

Notwithstanding these limitations,this study is the?rst to present a demographic pro?le of who uses the?nancial statements,and thus we argue that these?ndings are an important advancement of the literature.Decades of accounting research has examined the questions of what to disclose,when to disclose,how to disclose,and how much to disclose,but this study is the?rst to provide large-scale empirical evidence on the demographic pro?le of who uses the?nancial statements.

REFERENCES

Ball,R.2013.Accounting informs investors and earnings management is rife:Two questionable beliefs.Accounting Horizons27(4): 847–853.doi:10.2308/acch-10366

Beaver,W.1998.Financial Reporting:An Accounting Revolution.Englewood Cliffs,NJ:Prentice Hall.

Bozanic,Z.,J.L.Hoopes,J.R.Thornock,and B.Williams.2017.IRS attention.Journal of Accounting Research55(1):79–114. Drake,M.S.,D.T.Roulstone,and J.R.Thornock.2015.The determinants and consequences of information acquisition via EDGAR.

Contemporary Accounting Research32(3):1128–1161.doi:10.1111/1911-3846.12119

Drake,M.S.,D.T.Roulstone,and J.R.Thornock.2016.The usefulness of historical accounting reports.Journal of Accounting& Economics61(2/3):448–464.doi:10.1016/j.jacceco.2015.12.001

Dyreng,S.D.,W.J.Mayew,and C.D.Williams.2012.Religious social norms and corporate?nancial reporting.Journal of Business Finance&Accounting39(7/8):845–875.doi:10.1111/j.1468-5957.2012.02295.x

15It is also likely that this variable measures EDGAR use by business school professors for research purposes.Our data-cleaning procedures,however, remove downloads from all automated web crawlers,including business school professors that use web-scraping programs to download large volumes of EDGAR?lings.

Who Uses Financial Statements?A Demographic Analysis of Financial Statement Downloads from EDGAR67

Accounting Horizons Volume31,Number3,

2017

Financial Accounting Standards Board (FASB).2010.Statement of Financial Accounting Concepts No.8.Available at:http://www.fasb.

org/jsp/FASB/Document_C/DocumentPage?cid=1176157498129

Gonedes,N.J.,and N.Dopuch.1974.Capital market equilibrium,information production,and selecting accounting techniques:

Theoretical framework and review of empirical work.Journal of Accounting Research 12(1):48–129.doi:10.2307/2490499Healy,P.M.,and https://www.sodocs.net/doc/db3306518.html,rmation asymmetry,corporate disclosure,and the capital markets:A review of the empirical

disclosure literature.Journal of Accounting &Economics 31(1/3):405–440.doi:10.1016/S0165-4101(01)00018-0

Lawrence,A.,J.Ryans,E.Sun,and https://www.sodocs.net/doc/db3306518.html,ptev.2017.Yahoo Finance Search and Earnings Announcements .Working paper,London

Business School.

Lee,C.,P.Ma,and C.Wang.2015.Search-based peer ?rms:Aggregating investor perceptions through internet co-searches.Journal of

Financial Economics 116(2):410–431.doi:10.1016/j.j?neco.2015.02.003

Lev,B.1989.On the usefulness of earnings and earnings research:Lessons and directions from two decades of empirical research.

Journal of Accounting Research 27:153–192.doi:10.2307/2491070

U.S.Census https://www.sodocs.net/doc/db3306518.html,puter and Internet Use in the United States .May 2013.Available at:https://https://www.sodocs.net/doc/db3306518.html,/prod/

2013pubs/p20-569.pdf

Vuong,Q.H.1989.Likelihood ratio tests for model selection and non-nested hypotheses.Econometrica 57(2):307–333.doi:10.2307/

1912557

Watts,R.L.,and J.L.Zimmerman.1979.The demand for and supply of accounting theories:The market for excuses.The Accounting

Review 54(2):273–305.

Watts,R.L.,and J.L.Zimmerman.1986.Positive Accounting Theory .Englewood Cliffs,NJ:Prentice Hall.

68Drake,Quinn,and

Thornock

Accounting Horizons

Volume 31,Number 3,2017

Copyright of Accounting Horizons is the property of American Accounting Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.However,users may print,download,or email articles for individual use.