Stockownership in the United States: …...Trends in institutional stockowner-value of stock owned bv U'S' indi- the information from the 1960 special ship 24 viduals (and by individuals

By MARSHALL E. BLUME, JEAN CROCKETT, AND IRWIN FRIEND

Stockownership in the United States:Characteristics and Trends

PagePart 1: Introduction and Summary 16

Summary o f main results 1 7 T k j ^ ¥ j i J » i nPart 1: Introduction and SummaryStockownership 18

Importance of institutions 18Importance of upper income __

groups 19 XXELATIVELY little is known about The purpose of this article is to fillPart 3: Distribution of Dividends and the patterns of Stockownership or m some of these deficiencies, mainly

Stockholdings Among Broad c},&Ilges in these patterns over time, on the basis of data on individualr o u p s a l t h o u g h stockholdings a r e a highly dividend receipts and t h e income,The dividend gap 19 , , , £ , , i 1 ,1 occupation, location, and broad ageTT .. A . . *v , . rt, important component of total wealth, F. ' . '. fe

nUnlisted domestic stock 21 • n * • j- • i i , grouping of the recipients, as reportedAll domestic stock 23 especially for individuals at upper - , f , .-^ j i iAH domestic SLOCK Z6 F J FF m two \^^Q stratified random samples

Part 4: Trends in Concentration of ™com.e levels' Even the available of individual income tax returns (formsStockownership Since Late 1950 's___ 24 historical series on the total market 1940) for 1960 and 1971. Although

Trends in institutional stockowner- value of stock owned bv U'S' indi- the information from the 1960 specialship 24 viduals (and by individuals and nonin- sample was analyzed in earlier papers,

Trends in individuals' stockowner- dividuals combined) are subject to a this article represents the first use of

ATshil' -"""-:------":- 25 substantial margin of error. More the 1971 data.1

.New data on distribution of stock- r™ -./^^ i , i iownership 27 deficient still is the information on the The 1971 results are based on a spe-

" """• value and characteristics of individual cial random sample of 17,056 returns,Part 5: Distribution and Performance stratified so as to oversample greatly

of Stockholdings by Types of In- lssues and stock portfolios held by . £ & J

, ,, %, e i: , ^ . . n ,, . n the upper income groups. Ine actualvestors and by Types of Stock 29 various income and other sociodemo- ^ i i i i T i, . I T - returns were sampled b y t h e Internal

Employment status 29 graphic groups and on the investment Reyenue Service (IRg)> For each re_Types of stock held...... 30 experience of these groups. Such infor- tum in the } the data on the

"er̂ T _ . '" 31 mationis val«able f^ analy«es of a amount of individual dividend receipts,wide range of economic issues, including the names of the payer corporations,

Appendix to Part 3: Estimation of problems associated with the inequality and the income and other sociodemo-Aggregate Value and Distribution of : , . , . . , nDividends and Stockholdings 33 m ™le distribution of income and graphic characteristics ot the taxpayers

wealth, the magnitude and timing of (but not their names) were transmitteddomestic 33 asset effects on consumption and sav- to the Census Bureau. The authors

stock ' . _ _ 34 ing, and the riskiness and performance Provided to Census the information onMarket value of all domestic of stock investments held by different the dividend yield> market rates of

stock 34 return, industrv, size, and risk charac-groups. ^ . .' " „ '

Appendix to Part 4: Estimation of tenstics for each of the payer corpora-Distribution of Dividends and Stock- XT rp, ,, „ - . tions listed in the sample returns;, , ,. . T .. . . NOTE.-—The authors are Professors of ^ , , . i •holdings of Individuals by Family T,. , . . . T^ , TT71 . Census then prepared tapes matchingTT,™™ f a i 4. ^ v o- finance and members of the Rodney L. White . . „ . . , -,Income for Selected Years 3o Centcr for Finan j } R k h t th Wh ^e corPorate information with the

Appendix to Part 5: The 1960 and 1971 School of the University of Pennsylvania-Samples of Individual Income Tax Thc research on which this article is based was L J' c™ckett and I. Friend, "Characteristics of stockForms 1040 36 ^ j u T Ownership," Proceedings of the Business and Economic

hnanced by a grant from the Rational Science statistics Section of the American Statistical Association, 1963,

The first stage 37 Foundation and was greatly assisted by the and I. Friend and J. de Cani, "Stock Market Experience of

The second stage"~~~~" 39 cooperation of the Bureau of Economic ™flerent IIwcstor G™.ps'" ^oceedingsof the Business andr n u . . 1

& L Economic Statistics Section of the American StatistThe third stage 40 Analysis (BEA). «a«cm,i966.

16

Digitized for FRASER http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis

November 1974

November 1974 SURVEY OF CURRENT BUSINESS 17

data on the individual returns. Thesetapes, which were designed to preservethe anonymity of individual returns,were used by BEA to carry out thetabulations necessary for this study.Only IRS had access to the actualreturns.

The 1960 and 1971 special samplesare unique in that, by permitting thematching of characteristics of individualstockholders with those of the stock-issuing corporations, they make it pos-sible to estimate the market value ofstock owned by different sociodemo-graphic groups. Although IRS publishesannually the distribution of dividendsby income class of recipient, it is notpossible to estimate satisfactorily thedistribution of market value directlyfrom these data, since price-dividendratios may vary substantially by incomeclass. Using dividend receipts from in-dividual payer corporations and appli-cable price-dividend ratios, the 1960and 1971 special samples provide thebasis for estimating average price-dividend ratios for stock held by dif-ferent groups of individuals. While themarket value of stock held by thesegroups can be estimated directly fromthe sample data, somewhat more reli-able estimates of the distribution ofmarket value by income class areobtained by applying the estimatedprice-dividend ratio for each incomeclass to the aggregate IRS figure fordividend receipts by that class. Thedistributions of market value by othersociodemographic characteristics esti-mated from the sample data are madeto conform to the distribution by in-come class obtained in this way. (Adetailed description of the proceduresfollowed, including the adjustmentsmade for nondividend-paying stock, isprovided in the appendix to part 5.)

From the 1960 and 1971 data, it ispossible not only to obtain fairly reliableestimates of the distribution amongsociodemographic groups of the marketvalue of all stock held by individualsbut also to determine other character-istics of the stock held by these groups.The data can further be used to analyzeportfolio performance and risk charac-teristics and to improve the accuracyof estimates of the total market value

of outstanding stock in the UnitedStates.

Some information—specifically, esti-mates of the distribution of dividendincome and market value of all stockby income class—will be presented for1958, 1964, 1969, and 1970, as well asfor 1960 and 1971. However, themarket value estimates for the first 4years are not as reliable as for thelast 2.

Summary of main result

The main results and implications ofthe analysis are:

1. The concentration of dividend in-come and market value of stock amongupper income groups continued todecline from 1958 to 1969, but notfrom 1969 to 1971. The share in stock-ownership of the wealthiest 1 percentof the population changed very littleover the entire period, in contrast toan appreciable decline from 1958 to1969 in the share of the other upperincome groups. Other data suggest thatthe 1958-71 period was characterizedby stability, or a slight decline, in theconcentration of total family incomeand net worth, although these esti-mates—especially those for net worth—are subject to substantial error.

2. Although data on the distributionof income and net worth after 1971are not available, the sharp drop instock prices since then, relative toprices of other assets, implies a sig-nificant decline in the concentration ofnet worth, inasmuch as stock consti-tutes a major part of the assets of theupper, but not of the lower, incomegroups. However, no similar effect 011the distribution of total income betweenthe two groups would be expected,since dividends, unlike stock prices,have not been depressed.

3. Although the distributions ofboth total income and dividend incomebecame considerably less concentratedfrom the 1920's to the end of WorldWar II, only the latter continued toshow a significant trend toward lessconcentration in the following years,and even that trend seems to haveabated substantially in recent years.

4. Despite the fairly substantialmovement in the postwar period, and

probably earlier, toward a more egali-tarian distribution of stockownership,the 1971 distribution among differentincome classes remained quite con-centrated. Thus, the 1 percent of U.S.families (including single individuals)with the largest personal income ac-counted for 47 percent of dividendincome received and 51 percent of themarket value of stock owned by allfamilies, while the 10 percent of familieswith the largest income accountedfor 71 percent of dividend income and74 percent of market value. (Foreignas well as domestic stock and beneficialownership of stock held by fiduciariesand agents are reflected in thesefigures.) The 1 percent and 10 percentgroups in 1960 owned 50 percent and79 percent, respectively, of the marketvalue of families' shareholdings. The1971 and 1960 figures, each of whichis based on a single year's income,probably understate the concentrationof stockownership that would be indi-cated for upper income groups if familieswere classified by their normal lifetimeincome or their average income overa period of years.

5. As of mid-1971, U.S. individualsowned an estimated $780 billion instock. (This is moderately higher thanthe corresponding Securities and Ex-change Commission (SEC) and Fed-eral Reserve Board (FRB) estimatesand may be compared with $335 billionfor mid-1960.) Of the $780 billion,$460 billion was held in domestic NewYork Stock Exchange (NYSE) andother listed issues, $50 billion in mutualfund stock, $35 billion in unlisted bankand insurance company stock, and$190 billion in direct holdings of othertraded and privately held unlistedstock.

6. The two employment statusgroups with the largest stockownershipin 1971 were the managerial and theretired. The relative share of stockowned by families headed by retiredpersons was appreciably higher thanin 1960.

7. In 1971, a surprisingly high pro-portion of the portfolios held by in-dividuals was dominated by a verysmall number of issues; thus, the port-folios were not well diversified. This

562-873 O - 74 - 3


November 1974

18 SURVEY OF CURRENT BUSINESS November 1974

finding applies to all income groups.Since there is ample evidence thatinvestors are risk-averse, the lack ofeffective diversification strongly sug-gests that two of the basic assumptionstypically made in capital asset pricingtheory cannot both be valid: namely,that investors measure risk by the vol-atility of the rate of return on theentire portfolio, and that investorshold homogeneous expectations aboutrates of return and risk. The lack ofeffective diversification also has im-portant social implications since, in amajor downturn in the stock market,a high proportion of investors willdo very much worse than the market.Thus, since early last year, when themarket value of NYSE stock as awhole dropped nearly 40 percent fromits high point, millions of investors—including many with moderate means—must have experienced catastrophiclosses.

8. The lower income groups tendedto hold somewhat less risky stock thandid the upper income groups. Althoughthe latter owned substantially morestock on the average, as high a pro-portion of their portfolios were aspoorly diversified as those of the lowerincome groups. Mutual funds were amuch more, and NYSE stock asomewhat more, important part oflower income portfolios. Among theNYSE stock, the lower income groupswere relatively more likely to holdtelephone and electric and gas utilitystock than the upper income groups,but the differences for telephone stockwere smaller in 1971 than they hadbeen in 1960. Electric and gas utilitystock constituted a much smallerproportion of holdings of all incomegroups in 1971 than in 1960.

9. Among employment statusgroups, managers tended to hold theriskier stock and retired and other notgainfully employed persons the lessrisky stock.

10. Investors in the upper incomegroups tended to hold stock with higherprice-dividend ratios than other in-vestors did. This tendency is consistentwith the greater tax advantages tohigh-income individuals of stock withlow dividend payout, that is, a high

earnings retention ratio. The sametendency was observed in 1960, butbecame more pronounced by 1971.

11. The rates of return realized onaverage in 1970-72 on stock held by thelower income groups in 1971 were notsignificantly different from thoserealized by the middle and upper in-come groups in these periods. Thisresult is quite similar to that found forthe years immediately preceding andfollowing 1960.

12. There were no noteworthy dif-ferences in 1971 investment perfor-mance among occupational or regionalgroups holding a substantial amount ofstock. This article provides the firstcomprehensive data on this subject.

13. While the total market value ofstock owned by U.S. families and thenumber of individuals owning stockincreased greatly from the late 1950's

to 1971 (and still remained much higherthan in the earlier period), the per-centage of stock owned by individualinvestors declined appreciably. Thisdecline reflects both the rapid rise inassets of financial institutions and theincreased proportion of these assetschanneled into stock investment. Manyindividual holdings of all sizes havebeen replaced by a much smaller num-ber of large institutional holdings, anda large number of new and generallyrather small stockholders have acquiredshares through the reduction in holdingsof more substantial individual investors.As a result, since institutions have notplayed an active role in corporateaffairs, and small individual investorshave tended to be less active than largeinvestors, managerial control of U.S.corporations may have been enhancedover this period.

Part 2: Earlier Studies of Trends in StockownershipEarlier studies have provided his-

torical insights into a number of differ-ent facets of stockownership, thoughmuch of the information provided bythese studies was based on fairlytenuous data. There are reasonablyuseful, but rough, long-term estimatesof the: (1) total market value of stockoutstanding in the United States,(2) aggregate amounts owned by thetwo major groups of investors—finan-cial institutions and families or house-holds, (3) number of individuals own-ing stock, and (4) amounts of dividendsand of total income received by groupsof families classified by total income.2

Historically, the market value ofstock has increased considerably morethan that of total net worth either ofthe economy as a whole or of the house-hold sector.3 For many years, stock has

2. There are no long-term series available on the numberof families owning stock.

3. R. W. Goldsmith, R. E. Lipsey, and M. Mendelson,Studies in the National Balance Sheet of the United States,National Bureau of Economic Research, 1963, provideshistorical estimates of the value and composition of assetsand liabilities of households and financial institutions. Morerecent, though less comprehensive, estimates can be foundin the Securities and Exchange Commission (SEC) StatisticalBulletins and the Federal Reserve Board (FRB) Flow ofFunds publications.

been by far the largest of the financialassets held by families and has consti-tuted one of the two major componentsof household net worth.

Importance of institutions

Excluding personal trusts, most ofwhich are administered by commercialbanks, stockholdings and stock tradingby financial institutions became impor-tant only after World War II. In 1940,such holdings accounted for less than5 percent of the market value of alloutstanding stock in the United States;even by 1950 this percentage was lessthan 8, in contrast with over 24 percentcurrently. Stock held in personal trustfunds experienced little change in rela-tive importance over the past half-century, accounting for about 10 per-cent of all outstanding stock owned bynoncorporate entities. A relatively smallnumber of institutions now hold close to35 percent of all outstanding stock; theremainder is owned by somewhat under32 million individual stockholders.4

4. New York Stock Exchange (NYSE) 1973 Fact Book.The NYSE shareownership series started in 1059.


November 1974


Despite the marked decline in theshare of the market value of all stockowned by individuals, the number ofsuch stockholders has increased greatlysince the turn of the century. Earlierstudies have indicated that the numberof individual stockholders in the firstthree decades of this century may haverisen from about 1 million to 10 million.5

In the next two decades, the numberactually declined, but the decline wasreversed in the 1950's. By the end ofthe decade, the number had increasedto about 12.5 million, and by early 1972a peak of 32.5 million was recorded.

Information on the number of stock-holders, or the ratio of that number tothe total population, obviously providesa completely inadequate picture of thediffusion of ownership among differentsectors of the population. It does noteven provide an altogether satisfactorypicture of the growth in the numberof basic consumer units (families orhouseholds) owning stock, since severalmembers of the same basic unit mayhold stock in their own names and thenumber doing so may vary over timeas a result of changes in tax laws.

The two major sources of informationon historical trends in the distributionof stockownership among differentgroups are the dividends reported byincome class on income tax returns(forms 1040) and the asset data onestate tax returns.6 Of the two, theestate tax data are less useful informa-tion sources because they cover aconsiderably smaller range of incomes,and, more importantly, because theyrequire a number of questionable as-sumptions to estimate the assets ofwealthy survivors from those reportedfor wealthy decedents (see part 4).

5. See E. B. Cox, Trends in the Distribution of Stock Owner-ship, University of Pennsylvania Press, 1960, for a summaryof these studies.

6. The income tax data have been analyzed in S. Kuznets,Shares of the Upper Income Groups in Income and Savings,National Bureau of Economic Research, 1953, and Cox,Trends. The estate tax data have been analyzed in R. Lamp-man, The Share of Top Wealth-holders in National Wealth,National Bureau of Economic Research, 1962.

Importance of upper income groupsThe analyses of trends in the dis-

tribution of dividend income based onincome tax data point to a substantialdecrease in the proportion of dividendincome received by the highest incomeclasses over the 1919-57 period. Onthe other hand, over this period, esti-mates derived from estate tax datapoint to a moderate increase in theconcentration of the market value ofstockholdings in the top wealth group.The discrepancy seems too large tobe explained wholly by differences thatmay exist between the concentrationof dividend income by income class andthe concentration of value of stock bywealth group as a result either ofdifferential movements in price-divi-dend ratios of stock held by upper andlower income families or of differentialmovements in the relation of incometo wealth for these two groups. As notedpreviously, the findings from the incometax data seem more reliable and appearto suggest some decrease in the pro-portion of stock held by the upperincome and probably also the upperwealth families. Those findings alsoseem more plausible in light of thefairly broad range of evidence that the

concentration of total income in theupper income groups diminished duringmost of this period.7

Data on the distribution of dividendincome, based on income tax returns,and on the distribution of the marketvalue of stock, based on estate taxreturns, are available for a number ofyears after the late 1950's. These willbe discussed in part 4 of this articlein conjunction with the data for 1971.

Probably the most comprehensiveand reliable data previously availableon the distribution of stockownershipby income class and by other socio-demographic characteristics are con-tained in the 1960 study, which is theprecursor of the present analysis.8

The 1960 and 1971 studies make pos-sible the first reliable estimates of themarket value and of the ownershiptrends of stock held by different groupsof families over this period. In additionto giving information on the distribu-tion of stockownership, the two studiesalso make possible improved estimatesof the market value of outstandingstock in the United States and providenew information in the risk, rate ofreturn, and other characteristics ofthe stock held by different groups.9

Part 3: Distribution of Dividends and StockholdingsAmong Broad Groups

A basic input in estimating the ag-gregate value and distribution by in-come class of the shareholdings ofindividuals is the information on divi-dends reported on Individual IncomeTax Forms 1040. Such information,based on a very large sample of returns,is developed each year by the InternalRevenue Service (IRS) and publishedin Statistics of Income: Individual In-come Tax Returns. However, the Sta-tistics of Income (SOI) data omit twocomponents of dividends allocable toindividuals: (1) dividends retained byestates and trusts on individuals' behalfas beneficiaries, and (2) dividends re-ceived by individuals, but not reportedon individual tax returns, either be-cause recipients were not legally re-

quired to report them or because re-cipients illegally underreported them.

The dividend gapThe aggregate magnitudes of the

two omitted components were esti-mated by the following procedure. Thefirst aggregate was derived from totaldividend receipts of estates and trustsas reported on fiduciary income taxreturns, after allowance for distribu-

7. Kuznets, Shares, and D. B. Radner and J. C. Hinrichs,"Size Distribution of Income in 1964,1970, and 1971," SURVEYOF CURRENT BUSINESS, October 1974.

8. The earlier results are presented in Crockett and Friend,"Characteristics," and Friend and de Cani, "Stock MarketExperience."

9. The 1960 figure on the market value of outstandingstock was used as a new benchmark by the SEC.


November 1974

20 SUEVEY OF CURRENT BUSINESS November 1974

tions of fiduciary income to individualsand other categories of beneficiaries.The income tax data, which are avail-able for 1970, were updated by usingthe market value of stock held by bank-administered trusts and estates in 1971(see appendix to part 3). The secondaggregate was derived by comparingdomestic corporations' total cash dis-tributions to stockholders, as reportedon corporation income tax returns,with total dividend receipts as reportedon forms 1040, after allowance fordividend receipts of other stockowner-ship groups and a number of reconcil-iation items (see table I).10

Total cash distributions of domesticcorporations exceed the receipts ofdomestic individuals by the dividendspaid to domestic corporations, non-profit institutions, and foreigners andby the dividends paid to fiduciaries,but retained by them or used to paytaxes or defray expenses. Such divi-dends therefore had to be subtractedin arriving at the cash distributionspaid to individuals.11 On the other hand,cash distributions paid by foreign cor-porations to domestic individuals hadto be added. These adjustments pro-duce a figure of $20.5 billion for 1971cash distributions by domestic andforeign corporations to domestic in-dividuals (see table 1).

Some portion of this total is notreportable as dividend income onindividual income tax returns: (1) dis-tributions of small business corporationselecting to be taxed as partnerships,(2) distributions taxable as capitalgains, and (3) nontaxable distributions.For comparability with dividends ac-

Table 1.—Estimation of Dividend Receipts by Individuals Not Reported on IndividualIncome Tax Returns, 1971

[Millions of dollars]

10. A detailed explanation of the sources and proceduresutilized in deriving the items in this table is given in theappendix to part 3. A comparable table for 1960 appears inCrockett and Friend, "Characteristics."

11. For some ownership groups, dividend receipts had tobe inferred from the market value data provided by Govern-ment sources. This required that market value be multipliedby a ratio of dividend-paying stock to total stock appropriateto the types of stock held, to obtain the value of dividend-paying stock only. This figure then must be multiplied by adividend yield (dividend-price ratio) appropriate to theportfolio held, to obtain dividends. For estates, trusts,nonprofit institutions, and foreigners, the proportion of stockpaying dividends and the dividend yield utilized are thosecharacteristic of listed stock and large unlisted issues tradedover the counter (OTC). For simplicity, the two steps de-scribed were combined, and market value was multiplied bythe ratio of dividends to total market value for the broadclass of stock appropriate to the portfolio of a particularownership group.

1. Distributions (other than own stock) of domestic corporations.

2.3.

Less: Domestic dividends (other than those paid by Federal Reserve Banks) received by domestic corporations..Plus: Distributions (other than own stock) by foreign corporations to domestic individuals, fiduciaries and tax-exempt institutionsLess: Domestic dividends paid to foreigners

Equals: Distributions (other than own stock) by domestic and foreign corporations to domestic individuals, fiduci-aries and tax-exempt institutions J

Less: Dividends received by corporate pension fundsDividends received by State and local government retirement fundsDividends received by other tax-exempt institutions (including those distributed through fiduciaries)..Dividends retained by estates and trusts or utilized to pay taxes or administrative costs

10. Equals: Distributions (other than own stock) by domestic and foreign corporations to domestic individuals 2

Less: Distributions of small business corporations taxed as partnerships.Nontaxable distributionsDistributions taxable as capital gains

11.12.13.

14. Equals: Dividends reportable on individual income tax returns.

Less: Dividends reported on individual income tax returns.Equals: Dividend gap

32,580

5,460

110840

26,390

2,460330

1,4401,660

20,500

1,290560880

17,770

16,790980

1. Includes a small amount of nondividend cash distributions paid to domestic corporations and foreigners.2. Includes a small amount of nondividend cash distributions paid to other ownership groups.

Source: See appendix to part 3.

tually reported on forms 1040 in 1971,these distributions had to be sub-tracted; this procedure yields a figureof $17.8 billion for dividends reportableon individual income tax returns. Com-pared with the $16.8 billion reported in1971, there is a dividend gap of about$1 billion.

This dividend gap is presumed toconsist of three components: (1) thesmall amount of illegal underreportingof dividends revealed by audit checks,(2) dividends received by nonfilers—either those with gross income so lowthat they were not legally required tofile or those who escaped audit checks,and (3) dividends below the exclusion,which the recipients neglected to indi-cate on their tax forms and which werenot found on audit.12

Since different procedures should beused in distributing the three compo-nents by income class, rough estimatesof their relative magnitudes were made.An estimate of illegal underreportingat 2 percent of reported dividends givesa figure of $340 million. This per-centage is considerably less than the 5percent figure assumed in the 1960study. The 5 percent figure, based on1959 IRS estimates published by Hol-land, was derived by checking corporate

12. In 1971, there was no requirement that dividends belisted on schedule B if total dividend receipts fell below $100.While such dividends should have been indicated on the firstpage of the return (and thus caught by the SOI sample,though not by the 1971 special sample), it is probable thatsome filers may have neglected to do so since no tax liabilitywas involved.

information reports against stock-holders' income tax returns.13 No cur-rent estimates on this basis have beenpublished, but unpublished IRS studiesshow a substantial reduction in under-reporting since 1959. This reduction ispartially attributable to increased en-forcement effort by the IRS and par-tially to the policy of making availableto the individual stockholder a state-ment of the dividends ascribed to himin corporate information reports toIRS. A lower limit to current under-reporting is probably represented bythe iy2 percent implied by the IRS1963 Taxpayer Compliance Measure-ment Program data, which do notattempt to match individual reportswith corporate information reports.

The dividends attributable to non-filers are estimated at $430 million, ortwo-thirds of the remaining gap. Thisfigure is considerably above the 1960estimate, in part because the grossincome requirement for filing was subse-quently raised from $600 to $1,700($2,300 on joint returns and higher forretired persons). In addition, New YorkStock Exchange (NYSE) figures in-dicate a very large increase (of almost1 million from 1965 to 1970) in thenumber of minors owning stock,14

a high proportion of whom are likely

13. D. M. Holland, Dividends Under the Income Tax,Princeton University Press, 1962, p. 90.


November 1974


to have gross income below the currentrequirement for filing.15

The remaining $210 million of thedividend gap is attributed to theomission of dividend receipts from taxreturns in cases where receipts werewithin the legal exclusion. Althoughabout 4% million filers in 1971 listeddividends totaling less than the ex-clusion to which they were entitled,the NYSE stockholder census indicatesthat there were 12}£ million holderswith portfolios under $5,000 at thebeginning of 1970.16 Receipts of a largeproportion of these stockholders wouldbe expected to fall below the $100exclusion, so that the total number ofindividuals receiving dividends in thisamount may substantially exceed the4}£ million filers who reported dividendsbelow the exclusion. The average divi-dend received in such cases would, ofcourse, be very small.

Unlisted domestic stock

The information in table 1, aug-mented by data drawn from Govern-ment or industry sources and from the1971 special sample of individual in-come tax returns described in theappendix to part 5, can be used togenerate estimates of the aggregatemarket value of unlisted domestic stockand of its distribution among ownershipgroups. Such stock is a very substantialcomponent of the total financial wealthof households, but existing estimates ofits total value are subject to widemargins of error. While the InvestmentCompany Institute (ICI) provides reli-able figures on the market value andbusiness and institutional holdings ofmutual funds, and the Securities andExchange Commission (SEC) estimatesthe market value of unlisted stock ofbanks and insurance companies, nosimilarly reliable estimates are availablefor other unlisted stock. This residualgroup is largely nonfinancial; and asignificant proportion is not traded over

the counter (OTC), in which case, pricequotations are unavailable.17

Two basic approaches that havecustomarily been used to estimate thevalue of the residual group of unlistedstock are followed here. A third pro-cedure, depending in part on the 1971special sample of individual income taxreturns, is also presented.

The first approach is based on ag-gregate cash distributions on all cate-gories of stock, which can be determinedwith a high degree of accuracy fromcorporate income tax data. From this,dividends on listed stock, mutual funds,and unlisted stock of banks and in-surance companies, which can be esti-mated with varying degrees of accurac}^from industry and Government sources,are removed.18 Next, nondividend dis-tributions are removed, leaving divi-dends on other unlisted stock as aresidual. (These computations areshown in the appendix to part 3.) Anestimate of the aggregate value ofdividend-paying stock in the residualcategory is obtained from aggregatedividends by dividing by an appropriatedividend yield, based on a large marketvalue-weighted sample of stock in thecategory under consideration.

This method, however, provides nofirm basis for estimating the value ofnondividend-paying stock. Evidenceindicates that a far higher proportionof unlisted than of listed stock pays nodividends. It is possible to estimatethis proportion on a sample basis forthe category of stock under considera-tion; and the aggregate previouslyobtained for dividend-paying stockcan then be correspondingly augmented.However, little confidence can beplaced in such an estimate becausesamples are necessarily drawn from an

15. Furthermore, the 2 percent estimate used for illegalunderreporting in 1971 may not fully correct for nonfilers whowere legally required to file. If so, a small but undeterminedamount of dividends received by such nonfilers, who are as-sumed to fall predominantly in the adjusted gross income(A GI) class under $5,000, may be included here.

16. NYSE, Sharcownership, 1970, p. 9.

17. Unlisted stock not traded OTC (that is, stock in whichtransactions involving a dealer or broker-dealer do not occur)either is closely held for control purposes, as in a familycorporation, or has a strictly local market, as in the case ofa smalltown bank or retail enterprise. When the return onsuch stock is taxed as partnership income, the market valueis excluded from the total. This is consistent with nationalincome accounts procedure, which excludes such return fromdividend income.

18. Where the sources supply market value rather thandividend data, it is necessary to estimate both the averagedividend yield and the proportion of stock paying dividendson a sample basis. Dividend figures are highly accurate forNYSE stock and for mutual funds, less so for other listedstock and unlisted stock of banks and insurance companies.

incomplete listing that consists onlyof issues for which price quotationsare available, and because the largesample that is available from theRodney L. White Center files almostcertainly overrepresents large firms to avery substantial, but unknown, degree.19

Since it is clear, from classifying thissample by market value of stock, thatthe proportion of nondividend-payingstock increases sharply as firm sizedecreases, the overrepresentation oflarge firms is a considerable dis-advantage.

The second approach deals directlywith market values, but on a samplebasis. Data on number of shares out-standing are collected for individualfirms for which price quotations canbe found. The NYSE, in connectionwith its most recent census, Share-ownership, 1970, contacted 7,450 un-listed firms (other than mutual funds)early in 1970 and determined their mar-ket value to be $366 billion. Such asample aggregate, since it is not ex-haustive, necessarily understates theuniverse total. At a minimum, theNYSE figure must be adjusted up-ward to account for unlisted stock(other than mutual funds) not tradedOTC. From the adjusted figure, it isthen necessary to eliminate the marketvalue of unlisted stock of banks andinsurance companies to arrive at theaggregate that is being measured.

Apart from the mutual fund com-ponent, any estimate of the marketvalue of unlisted stock not traded OTCis subject to a wide margin of error.The procedure in this study follows thatof Tri in basing the estimate on 1965estate tax data, which distinguishprivately held stock20 from the holdingsof traded stock reported in the 97,000Federal estate tax returns filed in that

19. Similar sampling limitations apply to the estimate ofaverage dividend yield utilized in obtaining the aggregatevalue of dividend-paying stock, but the consequences areless serious since the sample of dividend-paying stock prob-ably covers a large fraction of total market value for theuniverse sampled. No such presumption can be made for thesample of nondividend-paying stock.

20. Stock that was not identified by executors as tradedwas considered to be privately held if no price quotationswere readily available.


November 1974

22 SUEVEY OF CUERENT BUSINESS November 1974

year.21 Such stock amounted topercent of other stockholdings, asreported in these returns.

In the 1971 special sample of indi-vidual income tax returns, a basisexists for approximating, for that year,the aggregate holdings that correspondto the category of traded stock rec-ognized in the breakdown of stock-holdings from the 1965 estate taxreturns.22 An estimate is then derivedfor individuals' ownership of privatelyheld stock in 1971 by taking 15%percent of traded holdings. This pro-cedure assumes that the relationshipof privately held to traded stock forall individuals in 1971 is similar tothat for the decedents represented inthe 1965 estate tax returns. To obtainthe figure for total market value ofprivately held stock, a small allowancemust be made for holdings of otherownership groups (which may be ex-pected to constitute a rather smallproportion of such stock), and thestock of small corporations electingto be taxed as partnerships must bededucted. (This last category of stockis apparently included in the privatelyheld category in the estate tax data,although it is excluded here.)

Both approaches to estimating un-listed stock, other than that of mutualfunds and banks and insurance com-panies, can be seen to involve question-able steps. The first approach en-

21. Statistics of Income, 1965; Fiduciary, Gift, and Estate TaxReturns, table 1. L. M. Tri, "The Market Value of CorporateStock in the U.S.," SEC Office of Policy Research, June 1971,pp. 20-21.

22. Sample holdings that can be identified as listed stock,mutual funds, unlisted stock of banks or insurance com-panies, or other unlisted stock traded OTC are presumed tofall in this category, as is stock held in agency or custodialaccounts or in street name—that is, stock held as nominee bya bank or brokerage house, for the interest of the beneficialowner. In all but the last case, the dividend data can beconverted to market values with some confidence on acompany-by-company basis. While the conversion is lessprecise for stock held in agency or custodial accounts or instreet name, the overall figure for market value of individuals'holdings of the group of stock in question is a reliable one.(See part 5 for further details of the conversion procedures.)

Individuals' beneficial ownership of stock through fidu-ciaries is excluded here, in part because such stock will notnecessarily appear as part of the beneficial owner's estateand in part because a significant proportion of the stock innonbank-administered trusts may be privately held. Thetotal obtained for individuals' holdings of traded stock prob-ably falls short of the figure that would correspond preciselyto the traded stock category as utilized in analysis of theestate tax returns—to the extent that traded stock held intrusts does appear in the estates of beneficial owners and tothe extent that stock of unidentified paying corporations isin fact traded.

counters particular problems in theestimation of the nondividend-payingcomponent and the second in the esti-mation of the privately held com-ponent. In addition, inaccuracies arecertain to be introduced in any processthat converts dividends to marketvalue, or vice versa, on the basis ofsample estimates of the ratio of oneto the other for a particular class ofstock.

The third procedure depends, asdoes the first, on an estimate of thetotal dividends paid on stock of therequisite type, but it uses the 1971special sample of income tax returnsin determining these dividends. Thedividends received by individuals ondirect holdings of unlisted stock otherthan mutual funds are immediatelyavailable from the sample. This isa fairly reliable figure, but it mustbe augmented by estimates of thedividends from unlisted stock heldby individuals in agency and custodialaccounts and in street name and byfiduciaries and other ownership groups.23

Total dividend receipts for stockheld in agency and custodial accountsand in street name are obtained fromthe 1971 special sample; for fiduciariesand other ownership groups, dividendreceipts have already been estimatedfor the purposes of table 1. (See ap-pendix to part 3 for details.) If plau-sible assumptions are made as to theproportion of dividend income derivedfrom unlisted stock, an estimate canbe obtained of dividends on all un-listed stock not held directly by in-dividuals. The assumptions as to port-folio composition for the various groupsmust meet one constraint: the totaldividends allocated to listed stock(including individuals' direct holdingsas determined from the 1971 specialsample) must be consistent with thehighly accurate external figure fortotal market value of listed stock,

taking into account the average div-idend yield and the proportion ofstock paying dividends that charac-terize listed stock.24

To this estimate of the dividendson unlisted stock not held directly byindividuals, the sample-based estimateof dividends on individuals' directholdings of unlisted stock other thanmutual funds must be added. Aftersubtracting the small amount of mutualfund dividends received by groupsother than individuals and the aggre-gate dividends on unlisted stock ofbanks and insurance companies, anestimate is obtained—alternative tothat developed by the first approach—of dividends on the category of stockfor which the market value is beingdetermined. The market value ofdividend-paying stock is then derivedby multiplying dividends by theestimated dividend yield.

As with any approach based ondividend information, the problem re-mains of obtaining a satisfactory es-timate of the value of nondividend-paying stock. However, the 1971 specialsample provides some assistance herealso. To derive a figure for nondividend-paying stock from the estimated ag-gregate of dividend-paying stock, it isnecessary to estimate the overall ratioof nondividend-paying to dividend-paying issues for the class of stockunder consideration. However, it isnot feasible to obtain a large randomsample from the relevant universe onwhich to base such an overall ratio.The available sample is believed to bestrongly biased in favor of large firms,but it should provide a relativelyunbiased estimate of the requiredratio within each size class. If appro-priate weights were available (ideally,the population aggregate of dividend-paying stock within each size class),

23. There is room for some difference of opinion as to howmuch, if any, of the dividends for which the paying corpora-tion could not be identified represent listed stock incorrectlyspecified by the filer. In view of the care taken to identifycorporate payers, at least as to listing status, the proportioncannot be large. The 10 percent assumed here is probablyan upper limit. There is also an element of arbitrariness indetermining how much of the dividend receipts attributedto banks represents dividends on bank stock and how muchrepresents return on stock held in bank-administered truststhat has been distributed to the individual as beneficiary.

24. Since domestic corporations are known to invest heavilyin unlisted as well as listed subsidiaries, the assumption ismade that the proportion of intercorporate dividend receiptscoming from unlisted stock is as high as for individuals'direct holdings, that is, 27 percent. The portfolios for estatesand trusts and for agency and custodial accounts are assumedto be similar to those held directly by individuals, but alittle more conservative than those held directly by individ-uals, so that a somewhat smaller proportion of dividendreceipts is assigned to unlisted stock. For nonprofit institu-tions, individuals' holdings in street name, and foreigners,a very small proportion of dividend receipts is assumed tocome from unlisted stock.


November 1974


a weighted average of the ratios forindividual size classes would provide asuitable estimate of the overall ratio.The 1971 special sample data on therelative importance of each size class inindividuals' holdings of dividend-pay-ing stock within the relevant categoryis used to indicate population weights.25

This use of sample information onindividual holdings of dividend-payingstock to approximate populationweights is equivalent to assuming that,for each dollar of dividend-payingstock held in a given size class, anamount of nondividend-paying stock isheld equal to the ratio of nondividend-paying to dividend-paying stock forthat size class. When this weightingscheme is used for averaging over sizeclasses, the average ratio obtained istermed "sample-weighted ratio."

As a check on the sample-weightedratio of nondividend-paying todividend-paying stock, a randomsample of 130 unlisted stock (notstratified by size) was drawn from theBank and Quotation Record, a listingsubject to somewhat less size bias thanthe large sample available from theRodney L. White Center files. Thesmall random sample provided an esti-mate almost identical to the sample-weighted ratio just described.

The estimates obtained by these threeapproaches are in fairly close agree-ment. The first approach yields adividend figure of $5.2 billion and,utilizing sample-weighted averages forthe dividend yield and for the propor-tion of nondividend-paying stock,implies a market value of $318 billion.The second approach yields a figure of$358 billion. This figure is derived bytaking the $366 billion figure obtainedby the NYSE in early 1970 for 7,450unlisted firms that were traded OTC,26

adding $33 billion for privately heldstock, other than that of corporationselecting to be taxed as partnerships,and subtracting $41 billion of unlistedstock of banks and insurance corn-

Table 2.—Market Value of All Domestic Issues, by Market Type and Ownership Group,June 30, 1960 and 1971

[Billions of dollars]

Type of stock

ListedNYSE, domestic and foreign issues.Other, domestic and foreign issues..Less: Listed foreign issues

UnlistedMutual fundsBanks and insurance companiesOther

All domestic stock

All holders

1960

1636

108

326

160

486

1971

7315425

5941

358

760

458

1,218

Individuals, 1971

Directholdings 1

5133

189

317

273

590

Bene-ficial

owner-ship 2

144

43

187

Non-profit

institu-tions,19713

135

18

153

Domesticcorpora-tions,1971

138

121

259

For-eigners,

1971

26

3

29

1. Includes some stock held in street name. The 1971 special sample did not always permit the segregation of such stock.2. Stock held by fiduciaries, in agency and custodial accounts and in street name, for the beneficial interest of individuals.3. Includes pension funds and other nonprofit organizations. See text for complete coverage of item.

Sources: See text and appendix to part 3.

25. Even on this basis, some bias probably still existstoward overrepresentation of large firms, leading to an under-estimate of nondividend-paying stock.

26. In view of the unavailability of a broadly based priceindex for unlisted stock other than mutual funds, no adjust-ment is attempted to reflect the general price rise thatoccurred in the first half of 1971, after a very slight declineduring 1970.

panies. The third approach yields adividend estimate of $5.7 billion and,utilizing the same dividend yield andproportion of nondividend-paying stockas in the first approach, a maiket valueof $350 billion—intermediate betweenthe first two estimates, but close to thesecond. Thus the second and thirdapproaches tend to confirm each other,and this provides some support for theassumptions as to portfolio compositionthat are utilized in the third approach.

All domestic stock

Market value figures for domesticlisted issues, mutual funds, and unlistedstock of banks and insurance companies,as obtained from industry and Govern-ment sources are combined with thesecond estimate for other unlisted stockto obtain total market value of domes-tic issues (table 2).27 The second esti-mate, the largest of the three, is chosenpartly because it utilizes a directattempt to measure market value,rather than an indirect approach viadividends, and thus avoids the difficultproblem of evaluating nondividend-paying stock by inference, and partlybecause its conceptual shortcomings liein the direction of understatementrather than overstatement. This under-statement arises because the NYSEsample cannot have completely exhaus-ted the universe of unlisted traded stockother than mutual funds and becausesome price rise almost certainly occur-red between early 1970 and mid-1971.

27. A detailed explanation of the sources and proceduresused in deriving table 2 appears in the appendix to part 3.

Total holdings of individuals (directholdings plus beneficial ownership ofstock held by fiduciaries or in agencyor custodial accounts or in street name)are derived from the 1971 special sam-ple of income tax returns, after adjust-ment to exclude holdings of foreignstock (see table 2). Those of foreignersand nonprofit institutions (corporatepension funds, State and local govern-ment retirement funds, foundations,and educational endowments) are de-rived from Government sources andadjusted as shown in the appendix topart 3. The stockholdings of fiduciarieshave been allocated between individualsand charitable organizations in thesame proportion as the distributions byfiduciaries shown in that appendix.While total receipts of domestic divi-dends by domestic corporations areknown from corporate income tax data,the market value of the correspondingdomestic stockholdings is not known,and so it is computed as a residual (seetable 2).

Individuals' direct holdings of listedstock can also be obtained from the1971 special sample. Information onother holdings of listed stock dependson the assumptions mentioned earlieras to portfolio composition. Specifi*cally, the assumptions are that, (1) forestates and trusts and agency andcustodial accounts, 25 percent of themarket value (and hence a smaller per-centage of the dividends) is assignableto unlisted stock, and (2) for nonprofitinstitutions, foreigners, and the stockof individuals held in street name, 10


November 1974

24 SURVEY OF CUKRENT BUSINESS November 1974

percent of market value (and hence asmaller percentage of dividends) isassignable to unlisted stock.

Corporate holdings of listed stockare again determined as a residual.When this value is compared with theamount of intercorporate dividendspreviously assumed to arise from listeddomestic issues (that is, 27 percent ofthe $5.5 billion aggregate obtained fromcorporate income tax returns), theresulting ratio of dividends to marketvalue 2s is that characteristic of listedstock as a whole. This tends to con-firm the reasonableness of the assump-tions as to portfolio composition.

Since the stock of mutual funds andunlisted stock of banks and insurancecompanies is to a very large extentheld directly by individuals, and sincethere are good external estimates ofthe total market value of such stock,individuals' direct holdings are obtainedby adjusting total market value for theholdings of fiduciaries and other owner-ship groups. The market value of indi-viduals' direct holdings of other un-listed stock is then obtained by remov-ing, from the sample-derived dividendson all direct holdings, the dividendsalready accounted for by the estimateddirect holdings of listed stock, stock ofmutual funds, and unlisted stock ofbanks and insurance companies. Theresidual dividends are then convertedto a market value figure.29

The value of unlisted holdings offiduciaries, nonprofit institutions, andforeigners is already determined by theportfolio composition assumptions,

28. This ratio is the product of the proportion of stockpaying dividends and the dividend yield.

29. The ratio of dividends to total market value used issomewhat higher than the sample-weighted ratio for non-financial firms traded OTC. This is done in the belief thatindividuals probably would not be inclined to hold the veryhigh proportion of nondividend-paying stock that charac-terizes the small unlisted firms (market value under $15million) for which there is dividend information.

30. In comparing this residual market value with inter-corporate dividends previously assigned to unlisted domesticissues, the ratio of dividends to market value is found to besomewhat lower than the sample-weighted ratio. This is aconsequence of the decision to use a somewhat higher ratioin converting dividends on individuals' direct holdings to amarket value figure, since the dividends on individuals'direct holdings and corporate holdings combined bear arelationship to the combined market value that is very closeto the sample-weighted ratio. If the holdings of each groupwere made to conform precisely to the overall ratio for theresidual category of unlisted stock, the effect would be toi ncrcase the total holdings of individuals by about $20 billionand to decrease the holdings of domestic corporationscorrespondingly.

given the data on total stockholdings.The holding of corporations are againdetermined as a residual.30

The total market value for domesticissues was $1,220 billion in mid-1971(table 2). This is 2% times the corres-ponding estimate for 1960. (The totalincludes intercorporate holdings—finan-cial and nonfinancial—unlike the SECfigures that are discussed in part 4.)The value for listed stock increased ata slightly lower rate, unlisted nonfinan-cial stock at a somewhat more rapidrate, and mutual funds, of course, at amuch more rapid rate, than the total.31

In view of the substantial trend duringthe intervening years toward the listingof bank holding company stock, it isperhaps not surprising that the marketvalue of unlisted stock of banks andinsurance companies increased verylittle.

In 1971, individuals' direct holdingsaccounted for over 40 percent of listed

stock, somewhat over 50 percent ofunlisted stock other than that of mutualfunds and banks and insurance com-panies, and about 60 percent of allunlisted stock. Total stock of indivi-duals, including beneficial ownershipof stock held by fiduciaries and inagency and custodial accounts andstreet name, amounted to about 60percent of listed stock and 70 percentof unlisted stock. Nonprofit institutionsaccounted for 18 percent of listed stockand, under the assumptions here, forvery little unlisted stock. Intercor-porate holdings accounted for 18 per-cent of listed stock and over one-fourthof unlisted stock. The latter resultdepends to some extent on the assump-tion that corporations are consid-erably more likely than individuals tohold substantial amounts of nondivi-dend-paying stock in small unlistedfirms other than mutual funds andbanks and insurance companies.

Part 4: Trends in Concentration of StockownershipSince Late 1950's

The most widely publicized structuraldevelopments in the securities marketsover the past two decades have been thevery substantial growth in the relativeimportance of financial institutions inthe ownership of corporate stock andthe even more rapid rise in theirstock-trading activity. These develop-ments, associated with a correspondingdecline in the relative importance ofindividual investors, have been citedas having seriously adverse effects onmarket liquidity and, indirectly, onthe ability of most corporations toraise equity capital. Thus, it has beenargued that institutions tend to buyand sell large blocks of stock and toconcentrate their activity on a rela-tively small number of large issues.Also, it has been asserted that, sincethey are subject to the same influences,have access to the same information,and closely follow each other's assess-ments and actions, institutions are

more often than not on the same sideof the market. The result is said to bemuch greater price volatility in thestock in which institutions trade thanwould exist in a market dominated byindividual investors.32 Price volatility,except to the extent it can be offsetthrough diversification, increases therisk of stock investment and hencethe cost of equity capital. Moreover,it has been claimed that, to the extentinstitutions divert funds that wouldotherwise have been invested in smalland risky issues, they tend to depressthe prices of such issues and, as aresult, penalize new ventures.

Trends in institutional stockowner-ship

Pension funds accounted for the larg-est growth in institutional stockowner-

31. Since 1971, the growth rate of mutual funds has nolonger exceeded that of the market as a whole.

32. There is no convincing evidence that institutionaltrading is in fact associated with greater price volatility.The Securities and Exchange Commission (SEC) Institu-tional Investor Study (1971) provides some contrary butgenerally inconclusive evidence. However, institutions havebecome much more important in the stock market since theperiod covered by that study.


November 1974


ship. Mutual funds, which were anot-too-close second for the period as awhole, were of diminishing relativeimportance in recent years. Until thisstudy, there had been no systematicexamination of the types of individualswho accounted for the decline in theindividuals' share of stockownershipand trading. It has frequently beenasserted, however, that it is the smallinvestor who has left the market as aresult of a loss of market liquidityand unfavorable investment experience.Before presenting the new data ontrends since the 1950's in the distri-bution of stockownership among differ-ent family income classes, it is usefulto review the available information onthe changing relative importance ofaggregate institutional and familystockholdings.

In 1950, stockholdings of financialinstitutions, other than stock in bank-administered personal trusts, wereabout 7.6 percent of the market valueof all noninvestment company stockoutstanding in the United States ownedby domestic individuals, institutions,and foreigners.33 This figure increasedto 16.5 percent in 1960, 19.8 percent in1969, 22.5 percent in 1971, and 24.0percent in 1973. The share of the trustsremained relatively constant at 10 per-cent of all such stock during this period.The share of domestic individuals, in-clusive of trusts, declined from 89.1percent in 1950 to 72.3 percent in 1973.Institutions' relative importance instockownership is greater for publiclytraded corporations and especially sofor corporations traded on the NewYork Stock Exchange (NYSE).

The changes in the proportion of themarket value of stock held by institu-tions reflect the magnitude of their netpurchases of stock compared with thesize of net corporate stock issues and,presumably to a lesser extent, the priceperformance of the stock they held

33. Intercorporate holdings, other than investment com-pany holdings of noninvestment company stock, are excludedfrom the total; foreign issues outstanding in the UnitedStates are included. The source of the estimated holdings ofinstitutions, which includes nonprofit organizations, is theS KG Statistical Bulletin. Estimates of the total market valueof outstanding stock were also obtained from the SEC for1950 and 1960, and from the procedures outlined in this arti-cle for 1971. Rough approximations were obtained for 1969and 1973 by extrapolating the 1971 figure on the basis of thetrends shown by the corresponding SEC series. All figuresare yearend.

compared with the performance of themarket as a whole.34 For 1950-73,institutional net stock purchases of$153 billion substantially exceeded netcorporate stock issues of $77 billion.(Net stock issues are defined as sales ofstock issues less stock repurchases byU.S. corporations other than mutualfunds.) Net stock issues moderatelyexceeded institutional net purchasesuntil the late 1950's; since then, institu-tional net purchases have greatly ex-ceeded net stock issues. This excess ofinstitutional net purchases over corpor-ate net sales of stock in recent years,averaging more than $7 billion annuallysince 1965, represented almost exclu-sively net stock sales by domesticindividuals.

Trends in individuals' stockowner-ship

Some insights into the characteristicsof the individuals who sold these sub-stantial amounts of stock to institutionscan be obtained from data availablebefore this study. Thus, it is known thatodd-lot balances (purchases less sales)on the NYSE and American StockExchange (AMEX), which are rela-tively more important for small thanfor large investors, turned negative inthe late 1950's. The rate of odd-lotnet sales, which amounted to $5.0billion for 1950-73, increased over theperiod and reached a level of about $2.0billion annually after 1970.35 Moreover,since 1971, these odd-lot sales balanceshave been in excess of net purchases ofmutual fund shares, which are generallybought by small investors, and since1972, more mutual fund shares havebeen sold than purchased. The rate ofodd-lot net sales over the past twodecades was only a small fraction of thetotal net sales by domestic individuals

34. A number of studies document that the investmentperformance of institutional investors (that is, rate of returnfor a given risk) has not differed significantly from that ofthe market as a whole and that the risk characteristics ofstock held by individuals and institutions differ markedlyonly in the much higher proportion of non-NYSE stockowned by individuals. Therefore, the only noteworthy im-pact of differences in price performance on the relativeimportance of institutional holdings of stock would reflectdifferences in the price trends of NYSE and other stock.There is evidence to suggest that NYSE stock did not fareas well as other stock for much of the 1960's (SEC Institu-tional Investor Study}, but the reverse was probably true insubsequent years.

35. SEC Statistical Bulletins for monthly 1973 data; NYSE1973 Fact Book and AMEX 1973 Data Book for annual datafor other years.

to financial institutions. There is thussome reason to believe that, over thisperiod, larger individual investors werealso selling stock on balance, that is, thedollar value of their sales was greaterthan their purchases.

This belief is further supported bythe extremely rapid rate of increase inthe number of stockholders after early1959. This rate of increase was verymuch larger than the rate of growthin the value of all stock owned by indi-viduals that is attributable to netpurchases of stock rather than tochanges in stock prices.36 Thus theaverage stockholder owned a smallerproportion of all stock at the end ofthe period than at the beginning.These results seem to suggest an in-crease in the diffusion of stockowner-ship among small investors.

However, none of this informationprovides very much insight into theextent of changes in the distributionof stockownership among differentgroups of families since the 1950'sand, in particular, among the more andless affluent sectors of the population.Before the availability of the dataprovided in this article, there were twosources of data for investigating suchchanges.

The first consists of Smith's andFranklin's estimates, based on estatetax returns, of the share of corporatestock (and other major components ofnet worth) held by the richest 0.5percent and 1.0 percent of the popula-tion in 1953, 1958, 1962, 1965, and1969.37 The second consists of the morecomprehensive data on the incomedistribution of dividends by adjustedgross income (AGI) class availableannually (currently through 1971) fromthe Internal Revenue Service (IRS)publication Statistics of Income—In-dividual Income Tax Returns.™

36. See part 2 of this article for historical and recent dataon number of stockholders; R. W. Goldsmith, A Study ofSavings in the United States, Princeton University Press,1955, for historical data on net stock purchases by individuals;and the SEC Statistical Bulletins for recent data on net stockpurchases.

37. J. D. Smith and S. D. Franklin, "The Concentrationof Personal Wealth, 1922-69," American Economic Review,May 1974.

38. Both the estate tax and income tax data reflect owner-ship in the shares of investment companies, including mutualfunds, as well as those of other corporations.


November 1974


Smith's and Franklin's estimatespoint to a substantial decline in theshare of the richest 0.5 percent and 1.0percent of U.S. individuals in corporateshareownership over the 1953-69 period.This decline is associated with relativelylittle change in the share of such in-dividuals in total net worth. There issome evidence of a decline of the shareof these upper wealth groups in totalnet worth from 1965 to 1969; but giventhe margin of error associated with esti-mates based on estate tax data, littleconfidence can be placed on thisevidence since it could be changed bya small revision in either the 1965 or1969 figures. For corporate stock, theestate tax estimates indicate a declinein the share of the richest 1 percent ofindividuals, from 86.3 percent of themarket value of all stock in 1953 to74.4 percent in 1958, 62.0 percent in1962, 61.2 percent in 1965, and 50.8percent in 1969.

There are, however, a number ofpotentially serious inadequacies in theestimates derived from estate tax data.These include (1) possibly substantialbiases involved in the assumption thatthe assets and liabilities of decedentsare representative of the assets andliabilities of living individuals in thetop wealth groups, (2) deficiencies inthe mortality rates used to characterizespecific groups in the population,39 (3)systematic understatement in the estatetax estimates of the values of certainassets held by the top wealth groups(including closely held stock and largeblocks of publicly traded issues) evenafter the reported values are adjustedon the basis of sample audits, and (4)the treatment of individuals rather thanfamilies or households as the basicconsumer units. Moreover, Smith's andFranklin's estimates of the ratio of theholdings of the upper income groupsto the total market value of stockowned by all individuals appear to in-clude the shares and certificates ofsavings and loan associations as partof stockholdings, and they use earlierestimates of total market value, which

Table 3.—Percentage Distribution of Families,1 Dividend Income, and Valueof Stock by Family Income Level, 1958-71

Family income 2

tinder $5 000$5 000-$9 999$10 000-$14 999 -$15 000-$24 999$25 000-49 999$5o'oOO-$99 999 - - - - -$100 000 and over - - - - - - - - -

Total - - - - - - - -

Under $5,000 - . - . . -$5 000-$9 999$10 000-$14 999$15 000-$24 999 - - - - - - . -$25 000-$49 999$5o'oOO-$99 999$100 000 and over - - -

Total

Under $5,000$ 5 000-$9,999 - - - - - - - -$10 000-$14,999. . --$15 000-$24 999$25'oOO-$49,999$50 000-$99,999- - - ... -$100 000 and over

Total -

1958 1960 1964 1969 1970 1971

Number of families

48.7537.98.53.51.1.2

.05

100.0

43.939.410.64.61.2.25.05

100.0

37.238.616.06.01.7.4.1

100.0

26.932.721.815.22.3.7.2

100.0

23.931.923.115.94.3.7.2

100.0

22.031.423.517.34.8.8.2

100.0

Aggregate dividend income

4.610.512.917.420.715.518.4

100.0

5.010.711.718.221.813.519.1

100.0

4.010.611.015.120.517.221.6

100.0

3.09.99.4

14.620.219.823.1

100.0

2.98.69.4

14.119.720.125.2

100.0

2.88.29.3

13.818.920.026.9

100.0

Aggregate market value of stock

4.410.212.617.220.615.819.2

100.0

4.810.311.217.621.914.020.2

100.0

3.910.310.715.020.417.422.3

100.0

2.68.69.0

13.719.220.726.2

100.0

2.57.48.4

13.218.821 22S.5

100.0

2.47.08.9

12.817.820.930.2

100.0

1. Definition of families includes unattached individuals.2. Family personal income before income taxes.

Sources: BEA estimates on income distribution by family income class, IRS data on distribution of dividends by AGI,and results from two special samples of IRS returns for 1960 and 1971. See appendix to part 4 for details.

39. These deficiencies and other problems of estate taxdata, including the need to adjust for lifetime transfers, havebeen discussed most recently in J. D. Smith, The Concen-tration of Personal Wealth in America, Pennsylvania StateUniversity, 1973.

are less reliable than the revised figurespresented in this article.

The second published source of datafor analyzing changes in the distribu-tion of stockownership by differentincome groups—the Statistics of In-come (SOI) data on the income distri-bution of dividends—is subject to fewerdeficiencies than the estate tax data.It also has the great advantage thatboth the total of dividends reportedby all individual taxpayers (on forms1040) and the specific amounts reportedon each return are subject to checkagainst external sources. These checksinclude the total of dividends reportedpaid by U.S. corporations on corporatetax returns, adjusted in the mannerdescribed in part 3 of this article, andthe IRS audits of man}^ individualreturns, also mentioned in part 3. Thecheck results provide a reasonable de-gree of confidence in these data as anindication of the AGI distribution ofdividends received by individuals whoare required to file tax returns, whereAGI is defined as in the tax laws.

Even the income tax data, however,have three significant deficiencies for

the purposes of this study. First, AGIper return is not a satisfactory econom-ic measure of income for a householdunit. It does not conform very closelyto the concept of income used in thenational income accounts or to thefamily unit used for distributionalanalysis in those accounts. The taxmeasure of income is deficient perhapsmost notably because wealthy familieshave a tax incentive to distributedividend income among different mem-bers of the family, each of whom wouldfile a separate return, and becausecertain forms of income are fully orpartially tax-exempt and therefore notproperly reflected in AGI. Second,families or individuals with AGI belowspecified limits do not have to submitincome tax returns. Third, the distri-bution of dividend income by incomeclass may differ appreciably from thedistribution of the market value ofstock owned, since in view of the taxstructure, high income families mightbe expected to hold stock with a rela-tively low dividend payout, a highgrowth rate of earnings, and, hence, ahigh price-dividend ratio.


November 1974

November 1974 SURVEY OF CUERENT BUSINESS

Table 4.—Trends in the Distribution of Stockownership by Selected Total IncomePercentiles, 1958-71

1958I960. .1964

1969 .19701971

Percentage of total incomereceived by highest

1%

7.57.28.0

n.a.7.67.5

5%

19.919.420.0

n.a.19.219.1

10%

29.429.030.0

n.a.29.228.9

50%

76.776.877.6

n.a.77.176.7

Percentage of dividend incomereceived by highest

1%

50.648.448.5

45.946.946.9

5%

72.869.869.3

63.964.863.8

10%

82.678.375.9

72.172.171.6

50%

95.293.593.1

91.391.190.5

Percentage of stock value ownedby highest

1%

51.750.549.1

50.451.551.1

5%

73.771.370.5

66.668.067.1

10%

83.279.577.1

74.575.475.1

50%

95.594.093.3

92.592.492.0

N.a. Not available.NOTE.—The percentages 1, 5, 10, and 50 refer to the specified percentage of families with highest total income.

Source: See appendix to part 4 for details.

Despite these deficiencies, the incometax data might be expected to providea reasonably good indication of thetrend in the income distribution ofdividend receipts, from which the trendin market value can be estimated, inperiods when there were only smallchanges in the relevant tax laws.Thus, in 1958-69, when there were nomajor changes in the definition of AGIor in the minimum income classesrequired to submit tax returns, thereis again evidence of a reduction inconcentration of dividend income bytotal income class.40 The Lorenz curvesfor these years, with the cumulativepercentage of returns on one axis andthe cumulative percentage of dividendson the other, indicate a continued shiftin dividend income (in percentageterms) away from the upper incomegroups. A further small movement inthe same direction occurred in 1970,but in view of the very substantialupward revision in the minimum incomeclasses required to submit tax returns,not too much reliance can be placed onthis finding. No further change in theincome distribution of dividends oc-curred in 1971.

Thus, the income tax, like the estatetax, data point to some tendency to-ward a further reduction in the con-centration of stockownership amongthe upper income groups after 1958.However, the reduction implied by theincome tax data on dividends seemsless than that indicated by the estate

40. In 1966, dividend income on form 1040A had to bereported separately for the first time and, hence, could beincluded in the SOI data. A special tabulation for thatyear, however, indicates that the amount of dividendsinvolved was negligible, and the estimated income distri-bution of dividends in 1966 (as measured by a Lorenz curve)was quite close to that in 1965.

tax data on market value of stock held,unless the differential changes in price-dividend ratios for the upper and lowerincome groups are much larger than seemsplausible. According to the income taxdata, the 1 percent of returns withhighest income received 52 percent ofall dividends reported on tax returns in

27

1958, 49 percent in 1960, 43 percent in1969, and 42 percent in 1971. Thistrend implies a much smaller decline inthe concentration of stockownershipthan the estate tax estimates mentionedearlier.

New data on distribution of stock"ownership

More satisfactory estimates of therecent trends in the distribution ofstockownership by income class can beobtained by extrapolating the BEAestimates of the distribution of dividendincome by family income class. Theseestimates can be extrapolated from theone year for which they are available toother years on the basis of the IRSdata on dividend income by AGIclass. The resulting time series can thenbe converted to a series on the distribu-tion of market value on the basis of

CHART 6

Trends in the Distribution of StockownershipLorenz Curves, 1958-71

o

100

90

80

70

60

50

40

30

20

10

1958

1960 —19691971

.

10 20 30 40 50 60Number of Families (Percent)

70 80 90 100

U.S. Department of Commerce, Bureau of Economic Analysis


November 1974


appropriate price-dividend ratios de-rived from the two special samples ofindividual tax returns for 1960 and1971 discussed in the appendix to part 5.

The BEA estimates used for thispurpose consist of the distribution offamilies and income by family incomeclass for 1958, 1960, 1964, 1970, and1971 and the distribution of dividendincome by family income class for1964.41 The SOI data used are those onthe distribution by AGI class of thenumber of income tax returns, AGI,and dividends for 1958-71. The meth-odology followed in combining thesedifferent sources utilized the SOI dataon changes in the distribution of returnsand dividends by AGI class in 1958,1960, 1969, and 1971, relative to a 1964base, to estimate the correspondingchanges in the BEA distribution ofdividends by family income class.Appropriate price-dividend ratios werethen applied to obtain estimates of thedistribution of the market value ofstock held by different family incomeclasses (see appendix to part 4 fordetails). The distribution of dividendincome by BEA family income class,which was obtained as an intermediatestep, shows a smaller shift in Lorenzcurves from 1968 to 1971 and in theconcentration of dividend incomeamong the top income recipients thanthe income tax data described pre-viously.42

41. The 1964, 1970, and 1971 figures on the income distribu-tion of family income were obtained from Radner and Hin-richs, "Size Distribution," the 1958 and 1960 figures werederived from the SURVEY OF CURRENT BUSINESS, April 1964,and the 1964 figures on the distribution of dividends wereobtained from Size Distribution of Family Personal Income:Methodology and Estimates for 1964, BEA Staff Paper No. 21,June 1973. The 1964 estimates are the most reliable; the 1958and 1960 estimates used a somewhat less satisfactory method-ology than those for 1964, 1970, and 1971, and figures for thelast 2 years do not incorporate as much information as thosefor 1964. The main conceptual differences between the pre-and post-1964 income estimates are the inclusion of income(including dividends) retained by fiduciaries and privatepension and annuity benefits in the more recent, but not inthe earlier, series, while the reverse change occurred forbenefits received from health and welfare funds and employercontributions to pension funds. The conceptual differenceswill affect somewhat the comparability of the measures oftotal, but not dividend, income presented in this article,since the 1964 procedures for dividend income have beenapplied to the other years.

42. The BEA family income estimates differ from AGIreported on tax returns (1) by using a family (or unrelatedindividual) instead of the tax return as the basic economicunit, (2) by covering money income that does not have to bereported or is underreported on tax returns, and (3) by in-cluding nonmoney income and (4) by excluding all capitalgains and personal contributions for social security.

The results of this analysis show acontinued downward movement in theshare of dividends received and stockheld by upper income groups for theperiod 1958-69, with little change for1969-71 (see tables 3 and 4 and chart 6).The share in stockownership of therichest 1 percent of the populationchanged very little over the entireperiod, in contrast to an appreciabledecline from 1958 to 1969 in the shareof the other upper income groups. Theabsence of any clear decline in theconcentration of total family income(see table 4) may reflect the factthat the 1958 and 1960 income dis-tributions tend to overstate somewhatthe share of the bottom quintile intotal income as compared with the1964, 1970, and 1971 incomedistributions.43

Thus, for this period, there does notseem to be any support for the beliefthat small individual investors havebeen switching out of stocks to a greaterextent than large individual investors.On the other hand, it is true that thesubstantial rate of decline in the con-centration of stockownership amongupper income groups, which charac-terized the period preceding 1958,seems to have slowed. To some extent,the slowing in the historical trendtoward a more equal distribution in thedirect ownership of stock among dif-ferent income groups might be con-sidered to reflect the rise in indirectownership by the lower and middleincome groups as a result of theirgrowing beneficial ownership of stockthrough financial institutions that donot issue their own stock. However,such beneficial ownership largely re-flects the growing importance of cor-porate pension funds, where, as aresult of contractual obligations, thecorporations are more likely than theemployee beneficiaries to gain (or lose)by the composition of the funds' port-folios. As a result, there is little reasonfor families to take into account theirindirect interest in stock held by suchfunds in determining the proportion oftheir own assets to invest directly instock. While families may well treatequity in a pension fund as a partial

substitute for other forms of saving asa whole, any effect of an increase in afamily's pension equity on a single formof saving, such as investment in stock,is likely to be Mnall.

A question that naturally arises is,How do these trends in the income dis-tribution of stockownership comparewith trends in the income distributionitself? Though the estimates on thedistribution of total income by incomeclass are subject to a considerablemargin of error, they probably aresufficiently accurate to depict signifi-cant changes over time. The estimatesshow very little change in the con-centration of total income by incomeclass in the entire period after WorldWar II. There is some evidence of adecline in the share of total incomesreceived by the top income brackets(the highest five or so percentiles) ,44

However, the decline in concentrationof income among the top five per-centiles after the war was rather small,and the Census Bureau's CurrentPopulation surveys suggest that theshare of the top percentile in totalmoney income may have been risingsince 1967.45

It would appear, therefore, that giventhe margin of error in these estimates,the most impressive finding is therelative constancy of income shares bydifferent income groups. This contraststo the substantial movement toward amore egalitarian distribution of incomefrom the 1920's to the postwar period—a movement that would be even morepronounced on an after-tax basis.46

Thus, while the distribution of bothtotal and dividend income became muchless concentrated from the 1920's to theend of World War II, only dividendincome continued to show a significanttrend toward less concentration in the

44. E. C. Budd, "Postwar Changes in the Distribution ofIncome in the U.S.," American Economic Review, May 1970,and Radner and Hinrichs, "Size Distribution."

45. The more comprehensive BEA series are not availablefor the years between 1964 and 1970.

46. Kuznets, Shares. See also U.S. Income and Output,U.S. Department of Commerce, 1958, which presents esti-mates by S. F. Goldsmith for 1929 and 1956.

43. Radner and Hinrichs, "Size Distribution."


November 1974

November 1974 SUEVEY OF CURRENT BUSINESS 29

following years, and even that trendseemed to have been muted consider-ably in recent years.

Another question that can be raisedis, How do the trends in the incomedistribution of stockownership comparewith those in total wealth or net worth(that is, the market value of assets lessliabilities)? While the data available foranswering this question are rather weak,they again point to a decline in theshare of wealth owned by the topincome groups (highest 1 percent) fromthe 1920's to 1945, with no definitetrend thereafter.47

The finding that a clear trend towarda more egalitarian distribution of indi-vidual income and stockownership per-sisted after 1945, unlike the behavior ofnet worth or income, may reflect thefact that the ownership of corporatestock was (and to a lesser degree stillis) much more concentrated amongupper income groups than is true ofwealth generally. Thus, the observedtrend is consistent with a greater di-versification of asset structure by bothupper and lower income groups. It mayalso reflect (1) the increased use bywealthy investors of other forms of

investment (such as municipal bondsand real estate holdings) to minimizetaxes, in view of the marked rise in taxrates from the prewar period, (2) thepublicity given to the high stockmarketreturns realized over the postwar perioduntil recent years, and (3) the extensiveefforts made by the Wall Street com-munity to attract small investors intothe market.

Finally, the reduction in concentra-tion of stockownership among upperincome groups that has taken place overthe past half-century does not neces-sarily imply any reduction in theconcentration of corporate control.What has occurred is that many indi-vidual holdings of all sizes have beenreplaced by a small number of verylarge institutional holdings, and anextremely large number of new andgenerally rather small stockholders haveacquired shares through the reductionin holdings of a comparatively smallnumber of much more substantialindividual investors.48 Both develop-ments would appear to facilitate mana-gerial control of U.S. corporations, atleast until institutions play a moreactive role in corporate affairs.

Part 5: Distribution and Performance of Stockholdingsby Types of Investors and by Types of Stock

Besides providing an estimate of themarket value of stock held by in-dividuals and permitting an analysisof the trends in the concentration ofholdings the 1971 special sample ofIndividual Income Tax Forms 1040 col-lected for this study can be used togain insight into the distribution andperformance of stockholdings by typesof investors and by types of stock.49

47. Smith and Franklin, "Concentration"; J. B. Lansingand J. Sonquist, "A Cohort Analysis of Changes in theDistribution of Wealth," Six Papers on the Size Distributionof Wealth and Income, National Bureau of Economic Re-search, 1969; and Lampman, Share.

48. This is reflected both in the much more rapid increasein the number of individual stockholders than the growthin the value of outstanding stock attributable to new issues,and in the substantial reduction in the proportion of themarket value of stock held by the upper income groups.

49. The appendix to part 5 describes the 1971 special samplein detail.

Employment status

The 1971 special sample of individualincome tax forms reveals that employed

persons, including (for this article)the self-employed, accounted for 60.3percent of the forms 1040 filed in 1971,but only 49.0 percent of the marketvalue of stock held by individuals(see table 5). As a group, therefore,employed persons accounted for asmaller percentage of stock held thanof forms filed. Within this group,however, a more detailed breakdownshows that managers were responsiblefor only 10.2 percent of the forms filed,but accounted for 19.0 percent of thestock held by individuals.

In 1971, retired persons filed only16.5 percent of the foims, but owned19.3 percent of individual stockholdings.Like the retired, the other twobroadly defined employment statusgroups, not gainfully employed andunknown, owned larger percentages ofstock than the percentages of formsfiled. The not gainfully employed un-doubtedly included some unemployed,some housewives, some wealthy in-dividuals who had no need to work,and some minors who filed formsseparately from those of the economichead of the household. The unknowncategory represents forms for which theoccupation box was left blank. Thesefilers could have had any employmentstatus, but data to be presented latersuggest that most of these forms werefiled by retired and not gainfullyemployed persons.

A more detailed analysis of theoccupational data suggests that thelarger percentage of stock held bymanagers relative to the percentage offorms filed, and the correspondingly

Table 5.—Distribution of Individuals' Stockholdings by Employment Status, 1960 and 1971

Employment status

EmployedManagersProfessionalClericalSalesFarmersOther

RetiredNot gainfully employedUnknown

Total

Percentage offorms, 1971

10 214.44 67 02 1

22.0

60 3

16.54.5

18 7

100 0

Percentage of market value

1971

19.010.91 43.91.4

12.4

49.0

19.36.5

25.2

100.0

1960

55.2

13.66.1

25.1

100.0

Change1960-71

-6.2

5.7.4.1

.0

NOTE.—Employment categories were defined by the Bureau of the Census. Self-employed persons are included in the,employed category.

Sources: 1971 special sample and Crockett and Friend, "Characteristics."


November 1974

30 SUEVEY OF CUERENT BUSINESS November 1974

Table 6.—Percentage Distribution of Market Value of Individuals' Stockholdings in Various AGI Classes by Market Type of IssuingFirm, 1971

AGI class

Under $5,000$5 000-$9 999$10 000-$14 999

$15 000-$24 999$25 000-$49 999$50 000-$99,999 -- - --$100,000-$199,999$200 000-$499 999$500 000 and over

N YSE by market value of outstandingshares (millions of dollars)

500 ormore

30.524.932.0

29.928.324.024.126.010.9

100 to499

7.64.69.5

9.110.98.06.46.2

12.1

Under100

3.93.72.0

2.22.72.51.62.12.7

Total

42.033.243.5

41.241.934.532.134.325.7

AMEX

0.83.11.4

2.02.22.63.22.33.1

OTC

2.06.53.3

5.46.14.98.07.37.4

Unidentified stocks

Banks andinsurancecompanies

7.91.34.6

4.74.16.24.02.82.8

Miscel-laneous

24.112.816.0

18.917.223.526.821.320.3

Agency,custodial,and street

name

2.54.57.4

4.96.17.27.54.5

12.9

Mutualfunds

15.223.59.6

11.25.03.01.70.50.0

Trustsand

estates

5.415.214.2

11.717.517.916.626.927.7

Total

100.0100.0100.0

100.0100.0100.0100.0100.0100.0

Source: See text.

smaller holdings of other employedpersons, stem not from any greaterpredilection of managers, as managers,to hold stock, but rather from the factthat managers have higher incomes thanother employed persons. If managerswere to have a greater predilection forstock, one would expect that at anylevel of income, the ratio of the pro-portion of stock owned to the propor-tion of forms filed would be larger formanagers than for other employedpersons. However, an examination ofsuch ratios for each of several incomeclasses50 reveals no such tendency.Thus, for any class of employed per-sons, the percentage of market valueheld by filers in any adjusted grossincome (AGI) class of less than $50,000is smaller than the percentage of formsfiled, and greater for those in any AGIclass of $50,000 or over.51

For each of the three remainingcategories—retired, not gainfully em-ployed, and unknown—filers in anyAGI class in excess of $25,000 accountedfor more stock than their numberswould have implied, while the reverseoccurred for those in lower AGIclasses. Since individuals in the firsttwo categories would be receiving little,if any, wage income, it might be ex-pected that more of their AGI wouldcome from dividend income than foremployed persons. Therefore, the levelsof AGI at which the percentage of stockheld exceeded the percentage of formsfiled would be expected to be lower for

these two groups than for the employedgroups. A comparison of the percentageof stock owned with the percentageof forms filed in the unknown categoryreveals a pattern more like that of theretired and not gainfully employed thanof the employed. This fact suggeststhat most of the filers in the unknowncategory were not employed.

Compared with the 1960 results, theshare of the market value of individualholdings attributable to the employedfilers fell by 6.2 percentage points.52

Over the same period, the retired in-creased their share 5.7 percentagepoints. Since the proportion of retiredin the population of persons over 21increased by only 1.0 percentage point,this abolute increase in stockownershipalso represents a relative increase.Because the breakdown of the employedin 1960 appears to be based uponslightly different definitions, a satis-factory comparison with the new resultsis not possible.53

Types of stock held

To analyze the kinds of stock heldby AGI class, the total value of eachissue held by filers within each AGIclass was estimated. Each issue wasthen classified into one of severalbroadly defined stock categories, andthe total market value within eachcategory was calculated. Table 6 liststhese categories and the market valuesexpressed as a percentage of the total

50. This analysis is based upon the income classes given intable 6.

51. As the previous part pointed out, there are distinctlimitations of the use of AGI as a measure of economic earn-ings. Nonetheless, for lack of a better measure, this part usesAGI as a surrogate for such earnings.

52. Crockett and Friend, "Characteristics."53. That the changes in the not gainfully employed an

unknown categories—two categories that were presumablydefined identically in 1960 and 1971—were small suggests thatthe identified breakdowns in both years were consistentlydenned.

stock held within each AGI class.With the exception of the unidentifiedstock, the descriptions are self-explan-atory. The unidentified banks andinsurance companies consist of thecompanies whose names are clearlythose of a bank or an insurancecompany, but for which additionalfinancial data are unavailable. For themost part, the stock in the unidentifiedmiscellaneous category represents closelyheld over-the-counter (OTC) stockwith limited markets or OTC stockwith a small number of shares out-standing.

The proportion of stock invested inNew York Stock Exchange (NYSE)issues and held in an individual's ownname tends to decrease as income in-creases. The rank order correlation is— 0.67, which is significant at the 10percent level. Within the NYSE, thisnegative relationship is apparent forissues larger than $500 million andsmaller than $100 million. For themiddle-sized issues, $100 to $500 mil-lion, the relationship is positive but notsignificant (rank order correlation of0.23). OTC, agency and street name,and estates and trusts are stronglypositively related to AGI, with rankorder correlations of 0.73, 0.60, and 0.88,respectively. (Street name stock isstock held as nominee by a brokeragehouse for the interest of the beneficialowner.) If not a statistical aberration,the large percentage of assets in agencyand street name for those with AGIin excess of $500,000 may stem from thedesirability for individuals with ex-tremely large portfolios to delegate thecustodial function. For the unidentifiedstock, the relationships between the


November 1974

November 1974

percentage of stock held and AGI classare very weak.54

A percentage distribution for eachAGI class by industry group instead ofby broad market type was also pre-pared. An analysis of this distributionreveals a remarkable similarity in thepercentages of each industry heldacross AGI classes. The only majordifferences across AGI classes occurredin the telephone and communicationindustry and in the utilities. Both ofthese industries tended to be a muchmore important part of the portfoliosof lower income filers than of upperincome filers.

For filers in AGI classes of less than$25,000, the percentages in utilitiesranged from 4.7 to 6.5; for incomes of$200,000 and above, the percentageswere less than 1.0. While the 1960study found a similar pattern by AGI,it may be noted that the percentages ofindividual portfolios held in utilitystock at all levels of AGI were largerin 1960 than in 1971.

For filers with incomes of less than$25,000, the percentages invested inthe telephone and communication in-dustry ranged from 5.0 to 10.5; forincomes of $200,000 and above, thepercentages ranged from 0.6 to 3.6.In 1960, the comparative importanceof holdings in this industry in port-folios of persons in the lower, relativeto the upper, AGI classes was evenmore pronounced than in 1971.

Diversification and return charac-teristics

To measure the diversification andreturn characteristics of the portfoliosof individuals, several statistics foreach portfolio were calculated. Table 7presents averages of these statistics byAGI class and in total. Before examin-ing these averages, however, it may beuseful to review some of the funda-mental tenets of portfolio theory.

Under several alternative assump-tions, it can be shown55 that an in-

54. The large percentage for unidentified banks and insur-ance companies for the lowest AGI class may result from themisreporting, as dividends, of interest from privately ownedbanks and thrift institutions and "dividends" from partici-pating policies of stock companies. As explained in theappendix to part 5, there was substantial evidence of suchmisreported dividends from mutual companies in the 1971special sample of individual tax forms.

55. H. Markowitz, Portfolio Selection: Efficient Diversifica-tion of Investments, John Wiley and Sons, 1958.

SURVEY OF CURRENT BUSINESS 31

vestor, whether he be risk-averse ornot, can evaluate a portfolio in termsof the prospective expected return andstandard deviation of the return, wherereturn includes all dividends and capi-tal gains or losses.56 Further, a risk-averse investor would always want tominimize the standard deviation of thereturn for any given level of expectedreturn. In this theoretical framework,the risk of a portfolio might be equatedwith the standard deviation of returns.As long as returns on individual secu-rities are not perfectly positively corre-lated, diversification will always pay.57

The 1971 special sample does notprovide an ideal basis for estimating theextend to which individuals have diver-sified their portfolios of common stockbecause the sample contains informa-tion only on dividend-paying items.Yet an analysis of just these items doesgive a great deal of insight into theamount of diversification in individualportfolios of common stock.58 The

56. In theory, such a portfolio should include all assetsheld by an individual, including human wealth. In practice,the risk of a portfolio of common stock is typically evaluatedin isolation from other assets because of data limitations.The empirical work based on the 1971 special sample canonly, and therefore will only, evaluate the characteristics ofthe common stock portion of an individual's assets.

57. P. A. Samuelson, "General Proof That DiversificationPays," Journal of Financial and Quantitative Analysis,March 1967.

58. The Federal Reserve Board's Survey of FinancialCharacteristics of Consumers in 1962 would seem to be anideal survey to analyze diversification. The Rodney L. WhiteCenter is currently analyzing this file to provide confirmationof the results derived from the 1971 special sample.

results are so strong that it is doubtfulthat the inclusion of issues with nodividends would substantially alter thequalitative nature of the conclusions.

One theoretically appealing index ofdiversification would be a function ofthe potential reduction in the variabilityof the returns on a portfolio throughfurther diversification, holding expectedreturn constant. Since the data neededto construct such an index are unavail-able, other less satisfactory measuresmust be used. One measure of diversi-fication that has been used in otherother studies is the number of issues in aportfolio. The underlying assumptionis that the greater the number of issues,the greater the potential for diversifica-tion. On average, this statistic rangesfrom 3.2 for filers with AGI of lessthan $5,000 to 18.7 for filers with AGI of$500,000 and over (table 7). It is notuntil an AGI of $100,000 is reachedthat the average number of items perform exceeds 10.0.

In 1963, the Internal Revenue Serv-ice (IRS) collected information on thenumber of payer corporations perreturn by AGI class.59 Because ofchanges in the levels of income anddefinition of AGI, it is difficult tocompare the 1971 results with those for1963. Nonetheless, it does not appearthat there have been marked changesin the number of issues held per port-

59. SOI, 1963: Individual Income Tax Returns.

Table 7.—Measures of Risk, Diversification, and Realized Returns by AGI Class, 1971

AGI class

Under $5,000 .. . _$5,000-$9,999. . . . . . . . .$10,000-$14,999$15,000-$24,999$25,000-$49,999$50 000-$99 999$100,000-$199,999.$200,000-$499,999. . ..$500,000 and over.. ...

Total .

No. ofitems perportfolio

3.23.84.04.36.79.2

13.216.818.7

4.5

Diversi-ficationmeasure

0.59.55.47.48.47.52.56.55.64

.52

Realized returns (percent)

NYSE only

1/70-12/70

2344002

-32

1

7/71-6/72

5-1

5656793

5

All items

7/71-6/72

1089

111112121210

11

NOTE.—The measures are weighted averages of the measures for the individual portfolios. The weight given to a specificportfolio is proportional to the product of the market value of the sample portfolio and the appropriate blowup factor given inthe appendix to part 5.

Source: See text.


November 1974


folio at comparable levels of AGI.Below an AGI of $50,000, the number ofdividend-paying issues held per port-folio was less than 10 in 1963; abovethis AGI, the number was greater than10. If an AGI of $50,000 in 1963 isroughly comparable to an AGI of$100,000 in 1971, the 1963 and 1971results are strikingly similar.

With any reasonable estimate of thenumber of nondividend-paying items,the portfolios in 1971 or 1963 wouldnot be considered highly diversified,even at the higher levels of AGI.60

At the lower levels of AGI, diversifica-tion is extremely limited.

To achieve the full potential ofdiversification within a fixed number ofissues, not too much of one's assetsshould be concentrated in any one ortwo securities. A better measure thannumber of items held of the extend towhich the value of a portfolio is con-centrated in a few issues can be con-structed by summing the squares ofthe proportions invested in eachsecurity. Thus, a portfolio of two secu-rities with 90 percent in one and 10percent in the other would have adiversification measure of 0.82, the sumof the squares of 0.9 and 0.1, while anequally weighted portfolio of two secu-rities would have a diversificationmeasure of 0.5. In general, this diversi-fication measure will be between 1.0and the reciprocal of the number ofitems in the portfolio. The lower thediversification measure, the morediversfied the portfolio.

The average values of these measures,given in table 7 by AGI class, rangefrom 0.47 to 0.64. This range is roughlyconsistent with the level of diversifica-tion achieved in an equally weightedportfolio of two securities. Thus, at leaston average, individuals tend to concen-trate their holdings in a limited numberof issues, probably taking on consider-ably more risk than necessary.

The inherent danger in reporting onlyan average of some statistic is that thereis always a tendency to attribute toeach component the average value and

not to recognize that the values for thecomponents can vary quite widely.Consider, for instance, an averagediversification measure of 0.46 for twoportfolios, each of which contains 10securities. This figure of 0.46 could beobtained from two poorly diversifiedportfolios in which 48 percent is in-vested in each of two securities andthe remaining 4 percent spread equallyover the remaining eight. The sameaverage could be obtained from onewell-diversified portfolio with 10 per-cent invested in each security and avirtually undiversified portfolio with 90percent in one security and the re-mainder spread equally over the othernine securities.

For an examination of the disper-sion in the diversification measures,the data underlying table 7 were fur-ther analyzed. This analysis showsthat there is much variability in theextent of diversification of individualportfolios. It is estimated that 13 per-cent of filers reporting dividends andholding 24 percent of stock had a diver-sification measure of 0.23 or less, whilemore than 40 percent of filers holding22 percent of stock had a diversificationmeasure of 0.88 or larger.61

One reason why a person might holdan undiversified portfolio is to be ableto realize the potential returns fromsuperior security analysis. (In thisconnection, it might be noted that thereis no evidence that any substantialgroup of investors, except for ex-change specialists and, to some extent,corporate insiders, has outperformedthe market consistently over longperiods of time.) A second reason isthat an individual may have a largeholding in a particular security in orderto maintain effective control over thecompany. A third reason is that, overtime, the one or two securities with thehighest returns will tend to dominatea portfolio if, because of tax considera-tions or other reasons, no adjustmentsare made. A fourth reason is that someinvestors do not understand the prin-ciples of diversification; therefore, thestandard deviation of returns on a

portfolio is not the appropriate measureof risk in explaining their behavior.The explanation for such poorly diver-sified portfolios must await furtherresearch.

Though these two measures of di-versification suggest that some investorsmay be assuming greater risks thannecessary through improper diversi-fication, the measures are deficient inthat they do not distinguish amongstock with different degrees of non-diversifiable risk. A preliminary analy-sis using the so-called beta coefficient—a standard measure of nondiversifiablerisk—shows that filers with largerAGI tended to hold stock with greaternondiversifiable risk.62 This analysisalso shows that managers tended tohold the riskiest, and retired and notgainfully employed the least risky,portfolios.

The final characteristic to be meas-ured in this part is the rate of return,including dividends and capital gains,that individuals realized on their stockportfolios. Returns have been cal-culated for NYSE issues for 1970 andfor July 1971 through June 1972.Returns were also calculated for allitems in the latter period.63 Since thecomposition of individuals' portfoliosis estimated from the dividends re-ceived over all of 1971, the estimatedcomposition would be expected to beclosest to the actual composition onJune 30, 1971—the midpoint of theyear. Thus, the returns from July 1971through June 1972 can be interpreted asthose that would have been realized onthe portfolios attributed to individualsin mid-1971 if there were no changes inthese portfolios over the subsequentyear. The rates of return for 1970 aremore suspect, since they are based uponthe composition of the portfolio as esti-mated from dividends in 1971, eventhough the 1970 composition would beexpected to be somewhat different.However, the turnover rate of theaggregate of stock held by individualsis not great, so that these returns

60. The empirical evidence in Lawrence Fisher and JamesII. Lorie, "Some Studies of Variability of Returns on Invest-ment in Common Stocks," Journal of Business, April 1970,shows that equally weighted portfolios of 128 securities areconsiderably better diversified than equally weighted port-folios of only 8 or 16 securities.

61. To determine whether trusts, custodial, or agencyaccounts might have biased the average values for the diver-sification measures, the measures were recalculated excludingany form with this kind of item. The averages were not sub-stantially changed and, in some cases, even increased.

62. Marshall E. Blume, ''On the Assessment of Risk,"Journal of Finance, April 1971, contains a summary of therationale underlying this measure and the procedures forcalculating it.

63. Any item for which the return was unknown wasassigned a default value, as explained in the appendix topart 5.


November 1974


probably approximate quite closely thereturns realized by individuals in 1970.

In 1970, individuals on averagegained 1 percent on their NYSEdividend-paying investments. Fromthe files of the Rodney L. WhiteCenter, it was determined that thevalue-weighted return on all dividend-paying stock was 0.7 percent; thus,individuals fared as well as the market.64

On average, filers with AGI less than$25,000 realized somewhat greater re-turns than those with higher AGI.

From July 1971 through June 1972,individuals on average realized 5 per-

cent on their NYSE stock and 11percent on all items. The larger returnson all items resulted from the sub-stantially better performance of OTCissues in this period. From the Center'sfiles, it was found that the value-weighted return on all NYSE dividend-paying stock was 8.8 percent.65 Indi-viduals thus fared somewhat worsethat the market, at least on their NYSEstock.66 In contrast to the 1970 results,individuals with higher AGI averagedmarginally higher returns than thosewith lower AGI.

Appendix to Part 3: Estimation of Aggregate Value and Distributionof Dividends and Stockholdings

The dividend gap (table 1)

Items 1, 2, and 11: These items wereobtained from SOI, Preliminary 1971:Corporation Income Tax Returns, pp.4 and 18. Item 2 was adjusted to ex-clude dividends paid by Federal Re-serve banks, which did not enter intoitem 1. Item 11 was slightly reducedon the basis of later information.

Item 3: Market value figure wasderived from R. B. Scholl, "TheInternational Investiment Position of

64. The NYSE Composite Index in 1970 fell 2.5 percentbefore adjustment for dividends. After a 3.1 percent adjust-ment for dividends, the Center's files and the NYSE indexgive virtually the same results.

65. In the same period, the NYSE Composite Indeximplies a return of 7.7 percent before adjustment for divi-dends and 10.8 percent after adjustment. It is not knownwhat the actual reasons are for the difference of 2.0 percentbetween the Center's estimate and the NYSE's estimate.There are, however, several conceptual differences betweenthe two indexes. First, the Center's return includes preferredstock, and NYSE-preferred stock returned only 1.5 percentin this period. Second, in determining market weights, theCenter uses as the number of shares the number authorizedto be issued and issued, less Treasury shares; the NYSE basesits index on the number of shares authorized to be listed andlisted. The most significant difference from this source is theweights given to foreign companies traded on the NYSE.Third, the Center's returns include only dividend-payingstock. Although nondividend-paying stock performed betterin this period, adjusting for them would change the Center'sreturn by only 0.1 percent. Since the returns in table 7 werecalculated from the Center's files, the Center's return of 8.8percent is the most reliable benchmark for comparison.

66. That individuals performed less well in this periodmeans that nonindividuals, primarily some groups of insti-tutions, must have performed better. While mutual funds didnot perform better than the market, there is some evidencethat banks performed considerably better. (William G.Burns and Richard H. Klemm, "Performance of BankManagers of Trust Funds," Rodney L. White Center forFinancial Research, University of Pennsylvania Press,August 1973.)

the United States: Developments in1972," SURVEY OF CURRENT BUSINESS,August 1973, p. 18. Dividends on the$7 billion of foreign portfolio stockheld by domestic ownership groupswere estimated by multiplying marketvalue by the ratio of aggregate divi-dends to aggregate market value forNYSE, American Stock Exchange(AMEX), and large OTC issues com-bined as of mid-1971. The resultingfigure was slightly increased to allowfor cash distributions other than divi-dends, and $90 million was allocatedto holding and investment companieson the basis of SOI information onthe foreign dividends received by suchcompanies. The remainder was assignedto individuals, fiduciaries, and tax-exempt institutions.

Items 4 and 6-8: Market value datawere derived from SEC StatisticalBulletin, May 30, 1973, p. 520. Yearendvalues were adjusted to midyear onthe basis of the NYSE index of stockprices; they were then multiplied bythe ratio of dividends to market valueutilized for item 3. For item 8, thisestimate of dividend receipts was aug-mented by 8 percent of the dividendreceipts of estates and trusts, to allowfor dividends retained by fiduciaries onbehalf of charitable organizations asbeneficiaries. The estimate was furtheraugmented by $150 million, estimatedto be received by church and hospitalendowments not covered by the SECfigure for foundations. The dividendreceipts of corporate pension funds andof State and local government retire-ment funds, as derived from SECmarket value figures, were increased by$150 million and $50 million, respec-tively, to account for stockholdings ofunion pension funds, corporate profit-sharing funds, and understatement ofmunicipal retirement funds due toincomplete coverage.

Item 9: Market value of stock-holdings of bank-administered trustsand estates were obtained from TrustAssets of Insured Commercial Banks-1971, Board of Governors of the FederalReserve System, Federal Deposit In-surance Corporation, Office of theComptroller of the Currency. Divi-dends were derived by multiplyingmarket value by the ratio utilized foritem 3. This dividend estimate wasthen expanded to cover dividend re-ceipts of all estates and trusts bymultiplying by the ratio of the 1970

Table A.—Estimation of Dividend Income of Fiduciaries Distributed to Individuals, toCharitable Organizations, and Not Distributed, 1971

Distributions to individuals. .Distributions to charitable organizationsRetained income_.Administrative costsTaxes paid _ _

Total uses

Percentage allocation ofgross income less businessdeductions and distribu-tions to other fiduciaries

1965

Taxablefiduciaries

29.2.8

52.24.2

13.6

100.0

Nontaxablefiduciaries

73.312.43.8

10.4

100.0

Estimated allocations of dividend receipts(billions of dollars)

1971

Taxablefiduciaries

0.53.01.95.08.25

1.82

Nontaxablefiduciaries

1.99.34.10.28

2.71

Allfiduciaries

2.52.35

1.05.36.25

4.53

Source: See text.


November 1974

34 SUEVEY OF CUEEENT BUSINESS November 1974

SOI figure for dividend receipts for allestates and trusts (SOI, 1970: Fidu-ciary Income Tax Returns, p. 14) toreceipts of bank-administered trustsand estates estimated, in the mannerdescribed previously, from the 1970stockholdings reported to bank regula-tory agencies by these fiduciaries. (Theratio of 1.5 thus obtained is somewhatbelow the ratio implied by 1962 SOIdata, which segregate bank-admin-istered from other trusts and estates(SOI, 1962: Fiduciary, Gift, and EstateTax Returns, pp. 16, 22, and 26).)

The proportion of fiduciaries' divi-dend receipts not distributed to benefi-ciaries was estimated from the 1965breakdown of the uses of fiduciaryincome from all sources (SOI, 1965:Fiduciary, Gift, and Estate Tax Returns,p. 25). In table A, the percentageallocation, among uses, of gross incomeless business deductions and distri-butions to other fiduciaries is developedfrom the SOI data and applied to the1971 dividend total. (It is assumed thatno business expense is incurred in thegeneration of dividend income and thatadministrative costs represent the sameproportion of net income for dividendreceipts as for all income.) Distributionsto charitable organizations are includedas part of item 8 in table 1. Distribu-tions to individuals, augmented by aproportional share of undistributeddividend income and reconverted to amarket value figure, provide a controltotal of $138 billion for individuals'beneficial ownership of stock throughfiduciaries in the analysis of the 1971sample.

Items 12 and 15: These items werederived from SOI, 1971: IndividualIncome Tax Returns, p. 62. Item 12

was adjusted upward by $50 millionfor estimated underreporting and fornontaxable distributions to ownershipgroups other than individuals. To theextent that liquidating dividends aresuccessfully excluded from item 1, butare included in nontaxable distributionsreported on individual income taxreturns, this figure may represent anoveradjustment. Item 15 was adjustedto delete $88.5 million (based onfindings from the 1971 sample) for themisreporting, as dividends, of incomereceived from such sources as creditunions, mutual savings and loan assocrations, mutual life insurance companies,and mutual savings banks.

Item 13: Net realized capital gainsof mutual funds were obtained fromMutual Fund Fact Book, 1971, p. 54.This item was adjusted by adding anestimated $100 million for capitalgains distributions of closed-end fundsand of mutual funds not members ofICI. Item 13 substantially exceeds the$662 million reported on forms 1040as distributions taxable as capital gains(SOI, 1971: Individual Income Tax Re-turns, p. 62), but the $662 million figureexcludes capital gains distributions toownership groups other than indi-viduals.

Dividends on unlisted domestic stock

Aggregate dividends on unlisted do-mestic stock other than that of mutualfunds and banks and insurance com-panies were derived from total cashdistributions of domestic corporations,as shown in table B.

Market value of all domestic stock

The 1960 data, which were obtainedfrom Crockett and Friend, "Character-

Table B.—Estimation of Dividends on Unlisted Nonfinancial Stock, 1971

[Millions of dollars]

1. Distributions (other than own stock) of domestic corporations2. Less: Distributions that are nontaxable or taxable as capital gains3. Dividends taxable as partnership income4. Cash dividends on domestic N YSE issues5. Cash dividends on other listed domestic issues6. Equals: Dividends on unlisted domestic stock (excluding small corporations taxable as partnerships)7. Less: Mutual fund dividends8. Dividends on stock of unlisted banks and insurance companies9. Equals: Dividends on unlisted domestic nonfinancial stock (excluding small corporations taxable as partnerships).

32, 5801,4401,29021,2501,0807,5201,460880

5,180

Sources: For items 1, 2, and 3, see table 1, items 1, 11, 12. and 13. Item 4: Total cash dividends equal $21.616 million,NYSE 1973 Fad Book, p. 79. Dividends on foreign issues were estimated at $366 million (based on a market value of $12.4billion for listed foreign stock, NYSE 1972 Fact Book, p. 34). Item 5: This was derived by dividing the $37 billion market valueof domestic AMEX issues, plus an estimated $5 billio'i for stock listed on regional exchanges, by a ratio of dividends to marketvalue characteristic of AMEX issues. Item 7: Mutual Fund Fact Book, 1«<72, p. 54. The published figure was increased by 10percent for dividends of nonmembers oi TCI. Item 8: SEC estimates of traded unlisted stock of banks and insurance companieswere adjusted to midyear, increased by $2 billion to allow for privately held issues, and multiplied by a sample-weighted ratioof dividends to total market value for identified OTC financial firms (6.0214).

istics," p. 163, were adjusted to removeforeign stock.

NYSE listed stock was calculatedby summing data for individual firms.Foreign stock listed on NYSE ($12.4billion) was obtained from the NYSE1972 Fact Book. Total stock and for-eign stock listed on AMEX ($49 billionand $12.3 billion, respectively) wereobtained from the exchange. Domesticstock listed on regional exchanges wasestimated at $5 billion. Stock ofmutual funds was obtained by increas-ing the figure given in the Mutual FundFact Book, 1972, by 10 percent to allowfor nonmembers of the ICI. Unlistedstock of banks and insurance companieswas based on SEC figures, increasedby $2 billion to allow for privately heldissues.

The estimate of unlisted stock otherthan that of mutual funds and banksand insurance companies was based onthe NYSE figure of $366 billion for un-listed traded stock other than that ofinvestment companies in early 1970.This figure was adjusted by subtractingthe estimate for unlisted stock of banksand insurance companies and adding anestimate for stock of closely held com-panies derived by the following method.Based on 1965 estate tax data, indi-viduals' holdings of such stock weretaken to be 15.5 percent of their hold-ings of traded stock, as determined fromthe 1971 special sample. This figure,$75 billion, was increased by 25 percentto allow for holdings of other ownershipgroups, giving a total of $94 billion.However, much of this presumablyrepresents the stock of small corpora-tions taxed as partnerships, virtuallyall of which must fall in the presentcategory. Based on dividends of $1.3billion for such stock, an assumed divi-dend yield of 3.5 percent (relativelyhigh to reflect low prices due to lack ofmarketability), and the average ratioof total to dividend-paying marketvalue for nonfinancial firms tradedOTC, the value of such corporationswas estimated at about $61 billion, andthis amount was subtracted from the$94 billion total.

Individuals' direct holdings of listedstock were based on the market valueof identified NYSE and AMEX hold-ings in the 1971 special sample, with


November 1974


minor adjustments to incorporate asmall fraction of the unidentified stockincluded in the sample and to removeestimated holdings of listed foreignstock. Individuals' direct holdings ofmutual funds and unlisted stock ofbanks and insurance companies wereobtained by removing, from the totaloutstanding market value in thesecategories, the relatively small holdings(13 percent and 20 percent, respec-tively) of other groups, including fidu-ciaries. Other direct holdings of un-listed stock by individuals were deter-mined from the residual remainingafter dividends already accounted forby the assigned amounts of listed stock,mutual funds, and unlisted stock ofbanks and insurance companies hadbeen removed from total sample divi-dends for all direct holdings. Theratio of dividends to total marketvalue used in converting this residualto a market value figure was the sample-weighted ratio for medium-sized non-financial firms traded OTC (marketvalue, $15 million to $100 million).The figure for medium-sized, ratherthan total, OTC firms was chosen be-cause it seems unrealistic to assumethat individuals would be inclinedto hold nondividend-paying stock ofsmall corporations (market value under$15 million) in the proportions in whichsuch stock is represented in the sampleof firms in this size class.

Twenty-five percent of the stock heldby fiduciaries or in agency accountsand 10 percent of stock held in streetname was assumed to be unlisted. Theseproportions are consistent with thesample estimate of total dividends onbeneficial holdings of individuals, whensample-weighed ratios of dividends tototal market value for listed and un-listed stock, respectively, are applied.

Ten percent of the stock held in theportfolios of nonprofit institutions orforeigners was assumed to be unlisted.Again, this is roughly consistent withthe dividends assigned previously tononprofit institutions and foreigners,given ratios of dividends to totalmarket value appropriate to the twoclasses of stock. The figure of $135billion for holdings of listed stock by

nonprofit institutions is reasonablyconsistent with an estimate by theNYSE of $124 billion of NYSE issuesheld by such institutions at the end of1971 (NYSE press release, March 12,1973).

Intercorporate holdings of listed andof unlisted stock were determined asresiduals. As a rough check of reason-ableness, the ratios of dividends tomarket value implicit in these estimatesmay be examined. If, as assumedearlier, unlisted stock accounts forabout 27 percent of the $5.504 billion

of domestic dividends received, theimplicit ratios are 0.029 for listed and0.012 for unlisted stock, equal to thesample-weighted ratio in the case oflisted stock and somewhat lower thanthe sample-weighted ratio (0.016) thatcharacterizes traded unlisted stock offirms other than mutual funds andbanks and insurance companies. Thelatter finding results from the previousdecision to apply a ratio somewhathigher than 0.016 in converting in-dividuals' dividends on direct holdingsof such stock to a market value figure.

Appendix to Part 4: Estimation of Distribution of Dividends andStockholdings of Individuals by Family Income for Selected Years

The basic source of recent informa-tion on the distribution of dividendincome by family income class is BEAStaff Paper No. 21, which presentssuch estimates for 1964. To derivecomparable distributions for otheryears, average dividend receipts perfamily by income class were determinedfrom the 1964 BEA estimates and ad-justed to other years by the change inaverage dividends per return forroughly equivalent AGI classes, as ob-tained from SOI individual income taxdata for those years. The adjustedaverage receipts were then combinedwith BEA esimates on number of fam-ilies by income class for those years toyield aggregate dividends by familyincome class.

The first step in integrating BEAestimates on family income with theIRS data on AGI was to determine theapproximate range of AGI correspond-ing to each of several fairly broadfamily income classes. The upper limitof the AGI range was established by(1) subtracting, from the upper limit ofthe family income class, an amountbased on the average proportion ofincome due to transfer payments andto imputed income and (2) adding anamount based on the average propor-tion represented by personal contribu-tions for social insurance, within thatclass, as determined from the 1964 BEAstudy. In addition, the average dividendexclusion claimed in 1964 and the aver-

age adjustment required to convertgross income to AGI for the mostnearly corresponding AGI class wereremoved and the average net capitalgain was added.

The equivalences thus established arevery rough. It is not certain that therelative importance of transfers, im-puted income, and other reconciliationitems for 1964 are equally applicablefor other years. More importantly,multiple returns may be filed by mem-bers of the same consumer unit; there-fore, a return with relatively low AGImay relate to a member of a highincome family. Thus, at low incomes,the returns in the equivalent AGIrange, while reflecting the dividendreceipts of consumer units in thecorresponding family income class, willbe somewhat distorted by the presenceof other returns representing indi-viduals from higher family incomeclasses.

In particular, the number of returnsin the AGI range corresponding tofamily income of $2,000 to $5,999 farexceeds the number of consumer unitsin that family income class. The sameis true for family income under $2,000(roughly corresponding to AGI under$600) if allowance is made for the factthat a substantial fraction of consumerunits in this range may well be nonfilers.On the other hand, for families withincomes of $15,000-$49,999, and es-pecially $15,000-$19,999, the number


November 1974


of consumer units somewhat exceedsthe number of returns in the corre-sponding AGI range. For family in-comes of $6,000-$14,999, results arevariable from year to year, but thegeneral tendency is for the number ofreturns in the corresponding AGI rangeto exceed slightly the number of con-sumer units.

The second step was to estimate aver-age dividends per consumer unit byfamily income class for years other than1964. This was done by adjusting the1964 value based on BEA estimates bythe sometimes considerable change,from 1964 to the desired year, in averagedividends per return for the correspond-ing AGI range. To the extent that thismovement fails to reproduce move-ments in average dividends per con-sumer unit, errors will be introduced.Since underreporting of dividend in-come declined somewhat over the1958-71 period and since this under-reporting was somewhat more prevalentamong the lower income families, theestimated concentration of dividend in-come among the upper income groupsin the years after 1964 may be slightlyunderstated relative to the earlier years.Finally, the average dividend thusobtained was multiplied by the numberof consumer units in the appropriateincome class in the given year, asdetermined in Radner and Hinrichs,"Size Distribution." The distributionof consumer units by family incomeclass is not directly available for 1965-69; thus, the 1969 distribution wasobtained by interpolation, utilizing the1964, 1970, and 1971 distributions.

A check of the results thus obtained isavailable for 1960 and 1971. The sum-mation over income classes of dividendsderived as mentioned was comparedwith the total dividend receipts of indi-viduals obtained by augmenting SOI-reported dividends by estimates of (1)illegal underreporting and (2) dividendsreceived by nonfilers and by filers whofail to report dividends totaling lessthan the legal dividend exclusion. Thetwo alternative estimates are veryclose for 1960 and within 4 percent for1971, with the approach based onSOI aggregates yielding the higherfigure.

The third step was to use the BEA

distribution of dividend receipts toconstruct distributions of market valuesof holdings. Since the ratio of marketvalue to dividends tends to increasewith income, as demonstrated for 1960by Crockett and Friend, "Character-istics," and for 1971 by the resultspresented in part 5 of this article, thedistribution of market value should besomewhat more concentrated than thatof dividend receipts. To make thisadjustment, the logarithms of theratios of total market value to dividendsby AGI class were regressed upon thelogarithm of (100—_p), where p is theaverage of the two percentiles fromthe distribution of all filers correspond-ing, respectively, to the lower boundand upper bound of an AGI class.Such a regression was fitted using the1960 data (Crockett and Friend, "Char-acteristics") and the results from the1971 special sample given in table Gof this appendix to part 5.

Using the same definition of p, butcalculated from the BEA distributionof income, the regressions were usedto estimate price-dividend ratios ap-plicable to each of the BEA incomegroups. The 1960 regression was usedin 1958, 1960, and 1964; the 1971regression, in 1969, 1970, and 1971.These estimated price-dividend ratioswere interpreted as those applicableto the BEA classes up to a multiplica-tive constant varying from year toyear. Multiplying the BEA dividends

by the corresponding estimate fromone of these regressions gives thedistribution of market value up to amultiplicative constant. Expressing theresulting values as percentage distribu-tions gives the required distributionsof market value.

A final step was necessary to inter-polate these distributions of dividendincome and market value of stock byincome class in order to obtain thepercentage of each accounted for byspecified percentiles of families withhighest total income. For 1964, there isno significant problem of interpolation,since the BEA dividend distributionshows information for 22 income classesand since both linear and curvilinearinterpolations give almost identicalresults. However, this is not true for theother years for which data, on dividendincome and market value, are availableonly for seven broader total incomegroups. For these years, the method ofinterpolation used assumed that thedistribution of families and dividendsamong the several narrower incomeclasses corresponding to each of theseven broader income groups was identi-cal to that in 1964. While the results ofcurvilinear and linear interpolationsapplied to the narrower income classesare fairly close, the curvilinear inter-polation seemed preferable and wasused. Curvilinear interpolation of datafor the broader income groups givessimilar results.

Appendix to Part 5: The 1960 and 1971 Samples of IndividualIncome Tax Forms 1040

This appendix presents detaileddescriptions of the sampling proceduresfollowed in selecting the 1971 specialsample of individual Income Tax Forms1040 and the adjustments made to thesample in deriving the various estimatespresented in the text.67 To preserveconfidentiality, the IRS was the onlygroup that had access to the actualforms.

The appendix is organized in threestages, according to the three stages in

67. Crockett and Friend, "Characteristics," contains asimilar description for the 1960 sample.

which the sample was selected andprocessed. The first stage describes thesampling design and analyzes the extentand magnitude of potential biases inthe special sample relative to the pop-ulation of forms 1040 filed in 1971. Thesecond stage presents the proceduresthat the Census Bureau followed inpreparing a tape for subsequent pro-cessing at BEA and indicates the stepstaken to preserve complete confident-iality of the original returns. Thethird stage discusses the adjustmentsmade to the sample and then derivesestimates of the dividends received and


November 1974

November 1974 SUEVEY OF CURRENT BUSINESS

Table C.—The SOI Sample and the 1971 Special Sample by Sample Strata

37

Stratum

11121314151617

21

22

23

24

25

26

27

28

30

Description

Total

Nonbusiness total

Absolute size of largest income item-

Under $10 000$10 000-$14 999$15 000-$ 19 999$20000-$49 999- . -$50,000-$99,999 . _ . . _. . - .$100 000-$199 999$200 000 and over

Business, total -- - - --

Absolute size oflargest income item— and business receipts—

Under $10,000 __ . Under $20,000 _

$10 000-$ 14 999, . Under $50,000Under $10,000. . $20,000-$49,999__.

$15 000-$19 999 Under $100,000Under $15 000 $50,000-$99,999

$20 000-$29 999 Under $250,000Under $20 000 $100,000-$249,99~9

$30,000-$49,999. ... . . Under $500,000Under $30,000 $250,000-$499,999

$50 000-$99 999 Under $750,000Under $50 000 $500,000-$749,999

$100,000-$199,999 . . - . . Under $1,000,000.Under $100 000 $750,000-$999 999

$200,000 and over. Any amountUnder $200 000 $1 000,000 and over

Tax preference:Size of minimum tax $17,000 and over

Number of forms

Population

74,841,993

65, 759, 059

43, 027, 78912, 935, 6225, 795, 8853, 660, 142

273, 84852, 04213, 731

9, 082, 725

3, 996, 188

} 2,364,823

| 1,217,378

| 880, 725

| 403, 630

| 168, 565

| 34, 608

| 16, 808

209

SOIsample

269,421

133, 605

21, 52919, 47517, 16421, 72421, 95218, 03013,731

135, 607

14, 117

16, 636

18, 345

17,480

18, 035

16, 919

17, 267

16, 808

209

Min.numberexpectedin 1971specialsample

16, 912

10, 978

2,1531,9481,7162,1722,195

451343

5,929

706

832

917

874

902

846

432

420

5

Actualnumberin 1971specialsample

17,056

10,893

2,0951,8961,6722,1142,139

582395

6,136

707

833

919

876

903

847

502

549

27

Actualnumberin 1971specialsamplew. sch.

B, part 1

6,444

3,951

129180319823

1,654494352

2,470

66

141

222

316

504

550

343

328

23

Finalblowupfactors

20, 5386,8233,4671,731

1288935

5,652

2,838

1,324

1,005

447

199

69

31

8

Sources: Population and SOI sample figures were obtained from 801, 1971: Individual Income Tax Returns, p. 310. Actual number in 1971 special sample figures were calculated bydividing blowup factors into population. Final blowup factors were supplied by IRS.

the value of stock owned by individualinvestors by AGI classes.

The first stage

In the first stage, IRS designated asubsample of the 1971 SOI sample forfurther processing. The SOI sampleitself is a sample of forms 1040 stratifiedby: (1) the presence or absence of busi-ness receipts and (2) the absolute sizeof the largest income item and, if abusiness return, (3) the value of re-ceipts. In addition, one small stratumincludes all forms with a tax in excessof $17,000 on tax preference items ex-clusive of those in sample strata whereall forms were sampled. Within eitherthe business or nonbusiness groups, thesampling rates increased with the ab-solute size of the largest income itemor, where appropriate, receipts. TableC presents the criteria for the strata,the number of forms for each stratumin the population, and the numberdrawn in the SOI sample.

To be sure that, at the lower incomelevels, there would be sufficient num-bers of forms with dividends for laterstatistical analysis, the 1971 specialsample was selected in such a way as toreduce the magnitude of the over-sampling of upper income forms in theSOI sample. To this end, the IRSselected a subsample of the forms ineach of the SOI strata according to aprocedure that should have yielded apredetermined minimum number ofrandomly selected forms from eachstratum. This predetermined minimumnumber varied from stratum to stra-tum.68

A comparison of these minimumnumbers with the actual numbers sub-sampled from the SOI sample showsthat the actual numbers by samplestrata are in excess of the minimum

numbers, as they should be, except fornonbusiness forms with AGI under$100,000 (table C). IRS personnelcould provide no plausible explanationof why the numbers subsampled forthese nonbusiness forms were less thanthe predetermined minimum under thesampling design.69 If it can be assumedthat there was nothing unique aboutthe forms that presumably should havebeen in the subsample, but were not,the ratios of the population numberof forms to the actual number sampledin each stratum provide the appropri-ate blowup factors for subsequentlyestimating the market value and othercharacteristics of stock held by indi-viduals (see table C).

As the forms were selected from theSOI sample, IRS personnel photocopied

68. Specifically, the procedure would have been expectedfor each of the strata 11-15 to yield a minimum of 1 out of 10of the SOI forms, for strata 21-26 a minimum of 1 out of 20,and for the remaining strata a minimum of 1 out of 40.

69. Due to a clerical error at the IRS, an undeterminedbut, according to the IRS, small number of forms withattachments to schedule B's was not included. While theeffect should be minor in any case, the subsequent adjust-ments should minimize the potential impact of this error.


November 1974


all those with completed schedule B,part 1, for later processing by theCensus Bureau. This photocopying wasdone in such a way as to exclude thenames, addresses, and social securitynumbers of the filers. Table C showsthe number of forms with schedule B's,part 1, in the 1971 special sample.

Schedule B, part 1, contains a listof the sources and correspondingamounts of any dividend income orcapital gain distributions. The sum ofthese amounts less capital gain distri-butions is entered on the front of form1040 in box 13a. After deducting theexclusion, which may range up to $200for a joint return, the dividends in AGIare entered in box 13c. Any single orjoint filing with dividends and otherdistributions in excess of $100 shouldcontain a completed schedule B, part 1,even if there is ultimately no dividendincome in AGI. Undoubtedly, somefilings contain a completed schedule B,part 1, even though dividends and otherdistributions were less that $100. Like-wise, some filings probably do notcontain a completed schedule B, part 1(even though required), particularly if,after the exclusion, there were no divi-dends in AGI.

Thus, the photocopied forms can beviewed as a sample of forms with com-pleted schedule B's, part 1—hence-forth referred to simply as schedule B.If schedule B's were properly completed,and only when required, the popula-tion implicit in the 1971 special samplewould include all filings with dividendsin AGI plus all filings with dividends

and distributions in excess of $100,but with dividend income below theallowable exclusion. If, as is probablythe case, some schedule B's were com-pleted even though not required andsome not completed even though re-quired, this clear interpretation becomesblurred. Although implicit in this dis-cussion, it should be pointed out ex-plicitly that the photocopied forms donot include all dividends received byindividuals; therefore, in estimating themarket value of stock held by individ-uals, a series of adjustments for theseomitted dividends were necessary.

Before describing the work done b;ythe Bureau of the Census, the extentand magnitude of any biases in thissubsample of the SOI sample will beassessed by comparing the blown-upfigures for numbers of forms in the 1971special sample and the average divi-dends reported per form with blown-upfigures from the SOI sample (see tableD). Unfortunately, figures tabulatedfrom the SOI sample are not exactlycomparable with the 1971 special sam-ple of forms with schedule B's. None-theless, there are both published andunpublished figures from the SOI sam-ple that can be used as rough checks.

Consider first the number of forms.The SOI sample for individual incometax forms in 1971 provides an estimateof the number of forms that includedthe receipt of dividends on the frontof form 1040 in box 13a. Since not all ofthese forms would have a schedule B,these numbers should be larger thanthe pupulation number of forms im-

Table D.—Comparison of Blown-Up Number of Forms and Dividends Per Form fromSOI Sample and the 1971 Special Sample by AGI

Size of AGI

Under $5,000$5,000-$9,999 .. - .$10,000-$14 999$15,000-$24,999$25,000-$49,999$50,000-$99,999. .$100,000-$199 999$200,000-8499,999$500,000 and over

Total

Number of forms

SOI sample

Withdividendsand other

dist.

2, 340, 4242, 623, 8002, 838, 5903, 118, 8561, 333, 920

334, 32766, 00314, 2722,983

12, 673, 175

Withdividendsin AGI

1, 535, 7341, 529, 9751, 428, 0731, 688, 032

967, 150290, 74462, 13913, 8582,916

7,518,621

1971specialsample

1, 595, 8451,649,4381, 629, 2542, 118, 6201,054,527

306, 18959, 76213, 2662,680

8, 429, 581

Dividends per form

SOI sample

Withdividends

in AGI

717967

1,0131,6343,4268,691

26, 87082, 143

323, 667

1971specialsample

7301,2671,0221,2922,8487,881

24, 88879, 345

362, 995

Sources: SOI, 1971: Individual Income lax Returns and the 1971 special sample.

plicit in the 1971 special sample thatwas subsequently processed by theBureau of the Census. The SOI samplealso provides population estimates offorms with dividends in AGI. Everyform in this category should have had aschedule B attached. Since some filersmay have attached unnecessarily aschedule B or were required to attachone even though no taxable dividendincome resulted, the number of formsimplicit in the 1971 special sample offorms with schedule B's would beexpected to exceed the number withdividends in AGI. Only if a substantialnumber of filers reported dividendsin AGI on the front of form 1040 andfailed to complete a schedule B wouldthis last expectation be in error.

Thus, the estimates of the number offorms with schedule B's from the 1971special sample should fall between theSOI estimates of the number of formsreporting dividends in box 13a and thenumber of forms with dividends inAGI. Table D shows that for formswith AGI of less than $100,000, theestimates of the number of forms fromthe 1971 special sample do fall betweenthe appropriate SOI estimates. Forforms with AGI in excess of $100,000or above, the estimates from the 1971special sample are marginally belowthe expected range.

Next consider dividends per form.Again tabulations based upon theSOI sample do not contain figuresexactly comparable with those from the1971 special sample with schedule B's,but perhaps conceptually the closestnumber available from the SOI sampleis dividends in AGI per form. Thisnumber differs from the correspondingnumber for the 1971 special samplein two principal respects. First, divi-dends in AGI are after deduction ofcapital gains and nontaxable distri-butions and after provision for thedividend exclusion, which could rangeup to $200 per filing. Second, the 1971special sample undoubtedly includessome forms with schedule B's, but nodividends in AGI. The first effectshould result in some tendency for thedividends per form from the 1971special sample to exceed the SOIestimate. The second effect should


November 1974


cause the reverse; but, on balance,particularly for the larger AGI classesor sampling codes, the first effect isprobably more important than thesecond.

An examination of table D disclosesthat the dividends per form as esti-mated from the 1971 special sampletend to be marginally less than thoseestimated from the SOI sample forAGI between $15,000 and $199,999.Most of the understatement in thesemiddle-income categories can be tracedto the nonbusiness forms, though thereis some evidence of a slight under-statement in the business forms. IRSpersonnel were unable to provide anyadequate explanation of these phe-nomena. For most of the analyses inthis article, the adjustments in stage 3will provide appropriate corrections.The only analysis that might be affectedis that of diversification presented inpart 5, but external figures presentedin part 5 suggest that this bias is notserious.

The second stage

Next, IRS forwarded the photo-copies to the Bureau of the Census forcoding. As pointed out above, names offilers, addresses, and social securitynumbers were deleted from these photo-copies. The Bureau prepared a file thatincluded socioeconomic and sociodemo-graphic characteristics, the names of allsources of dividends and other distri-butions listed on schedule B, and theassociated dollar amounts. From theresulting file, the Bureau prepared alist of these dividend sources and sentit to the authors. Personnel at theRodney L. White Center copied ontothis list an identification number foreach stock that was contained in theISL tapes. The ISL tapes are a standardsource of security prices and cover allNYSE and AMEX stock, roughly 400mutual funds, and more than 3,000OTC issues. In addition, a small numberof issues not listed on the ISL tapes,principally small OTC companies, wereassigned unique identification numbers.

For each of these identified issues,the Center's data files and standardfinancial publications were used todevelop stock characteristics. If thevalue of an important characteristic

Table E.—Default Values for Identified Securities by Types

Type of securities

NYSE -common

NYSE-preferred -

AMEX-common

AMEX-preferred

Mutual funds

OTC-financial-common .. .. .. ... .

OT C-financial-preferred

OTC-industrial-common. - . . ... - . -..

Size of issue 1(millions of

dollars)

500 and over100-499Under 100

100 and over15-99Under 15 .

50 and over.. . _ _10-49Under 10

100 and over.15-995-14 .Under 5

Ratio ofdivid. &dist. toprice(6/71)

3 053.143.66

4.54

3 243.393.42

6 26

3 03

2.682 803.03

2.83

2.403.043.334.25

Ratio oftotalstock

to stockwith divid.

or dist.2(6/71)

1 01721 13151 3404

1 2076

1 16931 99362 9099

1 2502

1 0049

1. 05051 21892 7138

1 0000

1. 14181 63972. 89756. 7595

Returnfrom

7/71 to 6/72(percent)

10 25 48 0

1 5

24 65 36 7

11 2

11 1

14 316 719 0

13 3

18.012 45.97.9

1. Any issue for which the size of issue was unknown was classified in the smallest category of its type.2. The ratios for banks and bank holding companies irrespective of other characteristics were 1.0025 and 1.0116, respectively.

Source: See text.

for an identified stock was missing,what is technically known as a defaultvalue was assigned. These defaultvalues, listed in table E, were usuallybased upon available data for similarkinds of assets.70

70. For any category of stock in table E, generally less than1 percent and never more than 2.5 percent of the blown-updividends used default values in ascertaining the associatedmarket values. These market values are the basic input forestimating the distribution of market value of stock heldby AGI class.

A dividend or distribution sourcewas not assigned a unique identificationnumber if the ISL tapes did not coverthe company or if the name of the sourcewas incomplete, like "First NationalBank." These sources were classified asaccurately as possible into one ofseveral generic categories by using thenames of the sources as guides. Table Flists these categories, the percentage ofsample dividends falling in each, and

Table F.—Default Values, Names, and Importance of Generic Categories

Generic category

Agency or custodial accounts .. .Agency, custodial, or trust accounts .Banks _ - ._ . - ._ ..Bank holding companies - -

Brokerage houses .....Insurance companies (stock)Investment clubs ... _ . . _ _Holding companies

Mutual funds . . . . . . _ . - - - - - - - - - - -NYSE (oil companies)NYSE (unidentified)Professional partnerships

Real estate and mortgage trustsTrusts and estates ... . .. ... _ . .Miscellaneous (preferred)Miscellaneous (unidentified) 3

Deleted items:Credit unionsInsurance companies (mutual)Other nonstock items

Percentageof sampledividendsin category

4.341.103.54.25

2.24.69.0311

.47

.08

.061.13

.0218.27

.7517.14

.0001

.43

Ratio ofdivid. &dist. to

price (6/71)

(i)0)

3.043.88

(i)2.613.102.87

3.035.003.143.00

6.76(2)

4.542.71

Ratio oftotal stock

to stockwith divid.

or dist.(6/71)

0)0)1. 00251. 0116

0)1.00001. 09161. 0558

1.00491.00001. 05201.0000

1.0000(2)1.20471. 6667

Returnfrom

7/71 to 6/72(percent)

10.410.425.318.1

10.410.410.419.2

11.110.49.0

10.4

-.19.0

40.416.4

1. The ratio of dividends and other distributions to price, and the ratio of total stock to stock with dividends ordistributions, was calculated separately for each of the AGI classes shown in table D. The first ratio was calculated as theratio of the total dividends and distributions received by filers in a given AGI class on all dividend-paying items other thanthose received through agency, custodial, and street name accounts to the market value of these items. The second ratio wascalculated as the ratio of the market value of all items other than those received through agency, custodial, and streetname accounts to the previously derived value of dividend-paying items.

2. The ratios were calculated as in the previous footnote, except that they were based only on identified NYSE issues.3. The ratios were derived from the total holdings of industrial OTC stock with no control for AGI.Source: See text.


November 1974


Table G.—Dividends, Other Distributions, and Market Value of Stockholdings of IndividualInvestors by AGI

Size of AGI

Under $5 000$5 000-S9 999$10 000-$14,999 . . .. ..

$15 000-$24,999$25,000-$49,999 - - - - ..$50 000-$99 999

$100,000-$199,999. .. - - --$200 000-$499 999$500 000 and over

Total

Divi-dendsand

otherdistri-

butions,1971

Divi-dends,

1971

Marketvalue,

1971

(millions of dollars)

1,9732,1322,068

3,1603,7182,926

1,8611,3401,143

20,322

1,8271,9321,922

2,9243,5152,812

1,8071,3031,102

19,144

65, 73164, 65670, 554

112,776143, 956126,084

85,11859, 30252, 606

780, 783

Dividends tomarket value

(ratio)

1971

.028

.030

.027

.026

.024

.022

.021

.022

.021

.025

1960

0.035.034.034

.035

.036

.033

.031

} .031

.034

Source: SOI, 1971: Individual Income Tax Returns, 1971 special sample, and Crockett and Friend, "Characteristics."

the default values of selected character-istics used in the subsequent processing.Because of the diversity of thesecategories, the miscellaneous (uniden-tified) stock are most likely to be closelyheld or small publicly traded industrialcorporations. Some items, such asinterest payments, should not havebeen reported as dividend income.These items were deleted in some of thecalculations presented in the text.

The third stage

The Census Bureau merged the stockcharacteristic file with the tax forminformation and forwarded the result-ing file to BEA for final processing. Toestimate the dividend and market valueof all stock held by individuals by sizeof AGI, the following calculations wereperformed:

1. The population estimates of thedividends and other distributions forfilers with dividends and the distribu-tions reported on schedule B's as de-rived from the 1971 special samplewere made to conform to the cor-responding SOI estimates for all filersfor each of the AGI classes given intable G. The specific adjustment wasto multiply every dividend and dis-tribution on all forms within a specificincome class by the ratio of the SOIaggregate estimate for that class 71 tothe 1971 special sample aggregate es-timate.72 This adjustment accounts forthe dividends reported on the front ofthe forms 1040 but not on schedule

71. SOI, 1971: Individual Income Tax Returns, p. 62, col. 2.

B's. It also has the desirable property ofmaking the 1971 special sample lesssensitive to any sampling bias thatmay be associated with the level ofAGI.

2. From the estimates prepared inpart 3, the dividends that should havebeen reported on schedule B's, but werenot, are estimated at roughly $336million. This sum was distributed overreported dividends and other distribu-tions in such a way that the noncom-pliance ratio for each income classwould be a multiple of that for personswith AGI of $50,000 or over. For AGIless than $10,000, the multiple was 4.0,for AGI of $10,000-$14,999, 5.5; forAGI of $15,000-$24,999, 4.5; and forAGI of $25,000-$49,999, 3.5. Theserelative ratios of noncompliance werederived from an IRS study in 1959 73

by equating the fractile ranges of AGIin 1959 with those in 1971.

3. From the estimates prepared inpart 3, it is determined that $433 millionrepresent dividends received by personsnot required to file. These dividendswere allocated to the lowest AGI class.

4. From the estimates prepared inpart 3, it is determined that $217 mil-lion of dividends were received by filerswho had dividend income less than theallowable exclusion and failed to report

72. The 1971 special sample estimate excludes items thatshould not have been reported on schedule B. A similar ad-justment, however, was not made to the SOI estimate. Thislack of adjustment will result in an approximately 0.5 per-cent overstatement of dividends and other distributions. Tooffset this overstatement, no adjustment was made for theunderreporting of capital gain and nontaxable distributions,which is roughly of the same magnitude.

73. Holland, Dividends.

them in box 13a of form 1040. Thissum was distributed according to thesame distribution by AGI as returnsthat did report dividends, but failed toexhaust the exclusion. This distributionwas taken to be proportional to thedifference in each AGI class betweenthe SOI estimates of the number claim-ing dividend exclusion 74 and the num-ber of returns with dividends in AGI.About 60 percent of such returns fall inthe AGI range $15,000-$24,999, with90 percent under $25,000.

5. To allow for dividends retained byestates and trusts for their beneficialowners, each dividend from a trust wasincreased by 57 percent. This adjust-ment moves the market value of thesekinds of assets implicit in the 1971special sample to $130 billion, which isin rough conformity with the externalestimate developed in part 3.

6. All but $7.5 million of dividendsreported as received from publiclytraded brokerage firms were reclassifiedas dividends received on stock held instreet name accounts.

With these adjustments, the 1971special sample implies that individualsreceived $20.3 billion in dividends andother distributions. Table G shows thebreakdown by AGI class. After sub-tracting the SOI estimates of capitalgain and nontaxable distributions,75 thedividends received by individuals, in-cluding retentions by estates and trusts,are estimated at $19.1 billion (seetable G). The dividends and otherdistributions, together with the stockcharacteristics and the default valuesin tables E and F, imply a market valueof individual stockholdings of $780billion.

Finally, table G gives the dividendyield rates that were used in analyzingthe change in the concentration ofholdings over time in part 4. Forcomparison, table G also presentsdividend yield rates for 1960 that werecalculated conceptually in the same wayas those for 1971.

74. This fails to allow for the probable increase, as incomerises, in the average dividend of those falling short of theexclusion. However, the distribution of dividends receivedby this group cannot be determined from available datawithout an arbitrary assumption as to the average exclusionby income class on joint returns for those with dividendsin AGI.

75. SOI, 1971: Individual Income Tax Returns, p. 62.

U.S. GOVERNMENT PRINTING OFFICE : 1974 O - 562-873


November 1974

Documents

Stockownership in the United States: …...Trends in institutional stockowner-value of stock owned bv U'S' indi- the information from the 1960 special ship 24 viduals (and by individuals