Вы находитесь на странице: 1из 18

The Investment Theories of Kalecki and Keynes: An Empirical Study of Firm Data, 1970-1982 Author(s): Steven M.

Fazzari and Tracy L. Mott Reviewed work(s): Source: Journal of Post Keynesian Economics, Vol. 9, No. 2 (Winter, 1986-1987), pp. 171-187 Published by: M.E. Sharpe, Inc. Stable URL: http://www.jstor.org/stable/4538001 . Accessed: 11/12/2011 21:30
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

M.E. Sharpe, Inc. is collaborating with JSTOR to digitize, preserve and extend access to Journal of Post Keynesian Economics.

http://www.jstor.org

STEVEN M. FAZZARI and TRACY L. MOTT

The investmenttheories of Kalecki and Keynes: an empirical study of firm data, 1970-1982

I. Introduction
In the 1930s, JohnMaynardKeynes and MichalKaleckiindependently presentedtheories of firm behaviorthat emphasizedeffective demand and financial conditions as primarydeterminantsof investment. This theory has been extendedby James Duesenberry(1958), John Meyer and Edwin Kuh (1957), Hyman Minsky (1975), Josef Steindl (1976), and others. It differs from the neoclassical theory of optimal capital accumulation,in which investmentis modeled as the adjustmentof a capital aggregate to an optimal level, assuming profit-maximization, perfect competition, and well-behaved neoclassical productionfunctions. Rather,in Keynes and Kalecki, investmentin fixed capital primarilydependson a firm's demandexpectationsrelativeto its existing capacityand its abilityto generateinvestmentfundingthroughinternal cash flow and external debt financing. This paper analyzes the determinantsof investmentfrom the perspective of Keynes and Kaleckiusing a large sampleof U.S. manufacturingfirm datafrom 1970 to 1982. The historyof estimatingthis kind of investment model began with Kalecki's own work (see Kalecki, 1969), but the first extensive econometricstudyalong these lines was Decision (1957). SubsequenteconoMeyer and Kuh's TheInvestment
The authorsare Assistant Professors of Economics at WashingtonUniversity, Saint Louis, and the University of Colorado, Boulder,respectively.They would like to thank Paul Davidson, ArthurDenzau, EdwardGreenberg,Donald Harris, and Hyman Minsky for helpful comments and suggestions.
Journal of Post Keynesian Economics//Winter 1986-87, Vol. IX, No. 2 171

172

ECONOMICS JOURNALOF POST KEYNESIAN

metricworkusing whatMeyer andKuhcalled the "accelerator-residual funds hypothesis" was undertakenby Meyer and Glauber(1964), Kuh(1963), Anderson(1964), Resek (1966), andEvans(1967), among others. Coen (1971) analyzedthe effect of cash flow on the adjustment speed of actual to desired capital. The theory was among the alternatives tested in several comparative studies of empirical investment functions (see Jorgenson, 1971; Bischoff, 1971; Elliott, 1973; and Clark, 1979), but it has been overshadowedin recentyears by neoclassical models or models based on James Tobin's "q" theory of investment. The empirical results obtained with the accelerator-residual funds model have been criticized on the groundsthat the flow variablesused to representliquidity,such as profitsor cash flow, succeedonly because they are highly correlatedwith the outputor sales variablesthat appear in neoclassical investmentmodels. This studyaddressesthis criticism directly by including both sales and flow liquidity variables in the estimatedequations. This approachmay not have been satisfactoryfor earlierstudiesthatused aggregatetime series data,becausecollinearity and small sample sizes made it difficult to distinguishclearly between the independenteffects of liquidity and acceleratorvariables in time series regressions. Using a large pooled time series cross-section sample, however,we obtainedresultsthatclearlyindicatea strongindependent influence for internal liquidity on investment, even when sales variablesare includedin the equation.This supportsan importantpart of the Kaleckianand post-Keynesiantheory that has not fared well in some other studies.2 Anotherdifficulty with the earlier empirical analyses of Kaleckian and post-Keynesianinvestment models and the comparative studies cited above is that they used datageneratedprimarilyin the 1950s and 1960s. The generally tranquilexpansion that characterizedthe U.S. economy from the KoreanWar to the late 1960s resulted in data that investshowed little of the volatilitywhich is centralto post-Keynesian ment theory. In the 1970s, however, the environmentchanged. The
1'"Ouroverall conclusion is that where internalfinancevariablesappearas sigificant determinantsof desired capital, they representthe level of output" (Jorgenson, 1971, p. 1133). 2Steindl(1976, pp. 129-130) specifically recognizes this point. He writes that time series studies of investmentuse "a particularlyhazardousprocedure . .. since utilization in practice is very closely correlatedwith profits" (p. 129). He proposes a cross-sectional study of investmentsimilar to what we present in this paper.

AND KEYNES INVESTMENT THEORIES KALECKI OF

173

in of conditions amplitude fluctuation sales,prices,andfinancial greatin datathathadbeenrelavariability ly increased, causingsignificant we overtheprevious decades.Therefore, expectthe two tivelysmooth of features thepost-Keynesian to emerge more distinguishing approach clearlyin the analysisof the recentdatathanin earlierstudies. with Theresultspresented areconsistent threemajor here empirical withthepost-Keynesian utiliidentified view: (1) capacity hypotheses zation or sales variableshave a positive effect on investment; (2) variablesmeasuring internallygeneratedfinancefor firms have a on influence firminvestment strong positiveandindependent expendiof and (3) the largerthe interestcommitments firms, ceteris ture; Theseresultsholdup in the full paribus,the lowertheirinvestment. sampleand in varioussubsamples. of The remainder this paperis organizedas follows. Section 2 summarizes theoretical the framework. Section3 presents specifithe cationof themodelwe estimate discusses empirical and the techniques we use. The regression resultsappear section4. in II. The investmentmodel for to Keynes(1936, chapter11) linkedthe demand investment the the marginal efficiencyof capital,or, equivalently, demand pricefor valueof profit as price(PD)is defined thepresent capital.Thedemand flows a firmexpectsto earnfroma marginal investment expenditure afterdeducting The level of investment will settle financingcosts. wherethedemand the pricethefirmmustpayfora priceequals supply of of investment Onedeterminant PDis somemeasure (Ps). marginal In the relevant demand. the mostcommonly literature, variable output for The usedis capacity utilization.3 reason thisis straightforward. The is Firmswantto maindirecteffectof investment on output capacity. tainsufficient to demand. However, capacity meetforthcoming capaccosts. Excesscapacity relativeto demand raises ity involvescarrying be thesecosts andreducesprofits.Investment must,therefore, plans and linkedwithprojected demand relativeto current closely capacity, utilization.4 thisprojection bebasedonobserved will levelsof capacity
3Steindl(1976, pp. 107-113, 127-130) argues that utilization is the prime determinant of expected profitabilityand thus an essential determinantof investment. Similar argumentscan be found in Baranand Sweezy (1966, pp. 82-88) and Kalecki (1971, pp. 1-14). 4Using this framework,we see that the determinantsof the demandfor replacement investment are essentially similar to the determinantsof net investment demand.

174

JOURNALOF POST KEYNESIAN ECONOMICS

Keynes (1936, p. 136 andchapter16) also positeda fall in the return on investment(and thus a fall in PD) as investmentrises, because the decline in the scarcityof capitalreducesthe quasi-rent earnedby capital (see also Davisdon, 1972, pp. 72-73). Othershave linkedthe fall in PD as investmentincreasesdirectlyto firms' perceptionsof limited output markets.5Following Steindl (1976), it is assumedthatexcess capacity exists in capitalgoods industries;hence, the costs of producinginvestment goods will not be significantlyaffected by changing investment levels. This implies that dPs/dl is approximatelyzero. The demandprice for investmentalso depends on financial conditions.6 A firm's investmentprogramcan be financedwith fundsgenerated by its own cash flow, or the firm can finance its expenditures externallyby takingon debt andsetting up cash paymentcommitments. Let F, denote the amount of internal cash flow the firm can use to finance investment. It can invest up to I = F1/Ps without additional borrowing.Any investmentexceeding Imust be financedexternally,so thatmarginalexpendituresbeyondI createnew interestobligationsfor the firm.7 There are several reasons why the cost of financing investment externallywill rise as the level of investmentincreases. Kalecki(1937; 1971, pp. 105-109) arguedfor a "principleof increasingrisk," claiming thatthe marginalrisk of fixed capitalinvestmentrises with the level of investment. This implies that lenders will demand higher interest rates as the level of a firm's investmentrises (see Mott, 1986). Keynes (1936, p. 144) analyzed a similar phenomenon, which he called "lender's risk," that "may be due to moral hazard, i.e. voluntary default or other meansof escape, possibly lawful, from the fulfillment of obligation, or to the possible insufficiencyof the marginof security, i.e. involuntarydefault due to the disappointmentof expectations." These ideas have been extendedby Minsky(1975), who emphasizesthe
This is consistent with the results of Feldstein and Foot (1971) and Eisner (1978), which show that replacementand modernizationinvestmentare positively correlated with investmentfor expansion, althoughthey are less variable. 5See Duesenberry(1958, pp. 49-85), Kalecki (1971, pp. 110-123), Mott (1982, pp. 46-49), and Steindl (1976, pp. 122-124). 6See Kalecki (1937; 1971, pp. 105-123), Keynes (1936, especially chapters 11 and 12), Davidson (1972, chapter4), and Minsky (1975, chapter5). 7Wedo not consider the possibility of investmentfinance throughnew equity here since this has been shown to be insignificant. See Goldsmith(1965). Though new equity reduces the risk of illiquidity, it is more costly to the existing shareholders than borrowing. See Kalecki (1937; 1971, pp. 105-109) and Wood (1975).

OF INVESTMENT THEORIES KALECKI AND KEYNES

175

relationshipbetween internalcash flow and contractual paymentcommitments as a key determinantof the financing terms a firm can arbusiness loan is based on range. Minsky argues that a well-structured the lender's expectationthat the debt can be serviced from the future income generatedby a firm's operations.The "marginof security" for the loan is the amountby which the firm's expectedincome exceeds its total contractual paymentcommitments.Clearly,the lower the level of internal finance, the lower the margin of security for loans, and the greaterthe riskpremiumthe lenderwill require.The resultis similarto Kalecki's. As investment increases, holding the firm's financial resources fixed, lenderswill requirea higherinterestrateto compensate for the increasingrisk of default. This lowers the firm's demandprice of investmentas the level of investmentincreasesbeyond what can be financed internally. The extentto which financingcosts rise with investmentwill depend on the bankers'appraisalof a firm's abilityto carrymarginalincreases in debt. Thus, PD will fall fasterwith increasinginvestment,the more heavily leveragedthe firm is. The key measureof the firm's ability to carrymoredebt is the "marginof security,"which is linkedto the level of a firm's interestobligationsor cash commitments(CC). The higher CC, otherthingsequal, the lower the marginof securityandthe greater the increasing risk problems. Equation (1) gives our specification of the effective demandprice for investment faced by an individual firm.
(1) PD = PD(I, U, Fi,/Ps, CC)

Increases in investment(I) or cash commitments(CC) reduce the demandprice. Rising capacityutilization (u) or an increasein the investment goods that can be purchasedwith internalcash flow (FI/Ps) will increase the demand price. Equating Ps and PD gives a structural relation that determines a representativefirm's investment: (2) PD(I*, FI/Ps, CC) = Ps,

whereI* representsthe investmentlevel the firm chooses. Solving for I* gives a reduced form for investment, (3) 1* = I(u, FI/Ps, CC),

that is the basis of the empirical analysis that follows.

176

JOURNALOF POST KEYNESIAN ECONOMICS

PoDPS

PS

I
lI

I
I= F/P/

"Dl

P (I, u, FI/P, CC) I

i*

Figure 1

The model is presented graphically in Figure 1. An increase in capacity utilization shifts the PD curve outward and increases 1*. As FI rises, the PD curve also shifts out horizontally because a greater internal cash flow means that more investment can be financed without resorting to external funding. Holding FI constant, however, we see that variationsin CC affect the demandprice of investment somewhat differently. For investmentfinanced internally CC makes little difference. But higher cash commitmentsresulting from the financing of past projects increase the risk of new investment beyond the level that can be internally financed. Thus, higher CC makes the PD curve fall faster beyond 7, and I* is reduced.8
8Note that, althoughthe analysis is presentedas if the firm exhausts its entire flow of internalfinance before it borrows, such a strong assumptionis not necessary.The main point is that the cost of finance faced by a firm will rise as debt paymentcommitmentsincrease relative to internalcash flow availableto service the debt. If a firm takes on debt to finance a project that it could have financed out of internal funds, we would not expect it to face serious increasingrisk problems, so the PD curve will not fall significantlyuntil after 7 is reached. This analysis follows Davidson (1972, chapter4) by integratingfinancial conditions into the demandprice for investmentgoods. Other authorshave modeled financial constraintsthroughthe supply of funds firms face. Then, the supply price includes the capitalizedvalue of financing costs as well as the purchaseprice of investment goods. See Duesenberry(1958, pp. 49-112), Evans (1969, pp. 73-105), Minsky (1975, pp. 93-116), and Mott (1982, pp. 37-59). Both approachesresult in the same empirical investmentequation. Therefore, the distinction is not crucial for the purposes of this paper.

OF AND KEYNES INVESTMENT THEORIES KALECKI

177

III. Empiricalspecification
Three empiricalimplicationsof the model presentedin the last section are tested here: (1) the positive relationbetweeninvestmentandcapacityutilization (u), (2) the positive relation between internal finance (F1) and investment, (3) the negative relationbetween cash paymentcommitments(CC) and investment. Eachof these propositionsfollows directlyfromthe model in section 2. The basic investmentequationestimatedin the paperis a linearized version of equation (3): (4) INVST = ao + a1 L(SALES) + a2 L(IFIN) + a3 L(INTEXP) + a4 (GPLANT).

The variable definitions are: INVST: annualcapital expenditures, SALES: net annual sales, IFIN: internal finance, defined as after-taxprofits plus depreciationallowances minus common and preferreddividends, INTEXP: net annual interest expense, GPLANT: book value of gross plant. The L functions on SALES, IFIN, and INTEXP representlags, the form of which is discussed later. The remainderof this section discusses the connectionbetweenthe empiricalvariablesandthe theoretical model, the econometricestimationtechniques,andthe datasample used. The model presentedin the last section linkedthe demandfor investthis effect ment to capacityutilization. The best way of incorporating into an empirical investmentequationwould be to include a capacity utilizationvariable. However,no acceptablemeasureof capacityutilization is availablefor the individualfirm datawe use. It is not possible to determineoutputcapacityfrom financialdata. Capacityutilizationis proxiedby sales datain the empiricalspecification. Variationsin sales

178

JOURNALOF POST KEYNESIAN ECONOMICS

are highly correlatedwith variations in utilization, so this approach should yield acceptableresults. The analysisin section 2 also identifies an independent effect of the level of internalfinance(F1)on investment.The measureused for IFIN is the flow of funds availableto firms to finance capital expenditure: after-taxprofits plus depreciationallowances minus dividends. In our regressions,IFIN is lagged to accountfor the time betweeninvestment decisions and actual expenditure. Lagged interestexpense (INTEXP) is used to estimate the effect of contractualcash paymentcommitments(CC). In past studies of postKeynesianinvestmenttheory,this effect was usually modeled by variables thatrepresented stock of liquid assets or the stock of debt (see the Meyer and Kuh, 1957, and Meyer and Glauber,1964, for example). However, following Minsky (1975, chapter5) our model emphasizes the importanceof the flow of committed cash flows as the relevant constrainton investmentfinance. The results in the next section show thatthis approachresults in a more successful empiricalspecification. The GPLANT variableis included in equation(6) for econometric reasons. Because individual firm data are used, investment will be variablessimplybecause positively correlatedwith all the independent all the financialdatawill tend to be largerfor largerfirms. By includside of equation(4) ing the book value of gross planton the right-hand as a measure of the overall size of the firm, we can interpret the estimated coefficients of the other independentvariablesas the effect on investment, holding the size of the firm constant. Since GPLANT only proxies the size of the firm, it was not lagged in the estimation equations. In time series studies that use small samples of quarterlyobservations, some kind of lag distributionis often imposed to reduce the multi-collinearitycaused by including highly correlatedlags in a regression. With the data set used here, this kind of multi-collinearity will be less seriousbecausethe pooled time series cross-sectionsample of over 9,000 observations improves the ability of a regression to distinguish the effects of correlatedexplanatoryvariables. For these reasonscontemporaneous lagged valuesof the variablesare entered and as separateregressorsin our estimatedequations,unconstrained any by distributedlag assumptions. The contemporaneousvariables for IFIN and INTEXP have been omitted to avoid a possible simultaneitybias. Because all investment must be financed somehow, either internally or externally, current

INVESTMENT THEORIES OF KALECKI AND KEYNES

179

investmentis closely linked to currentfinance by definition. Omitting the contemporaneous finance variablesfrom the regression and using values alleviates this problem.9 only lagged Lags up to five years' lag were considered for each variable. The most interestingcomparisonis obtainedbetween the results with two andthreelags. The estimatedcoefficient on the thirdlag of SALESwas positive, but quite small. When a third lag was added for IFIN and INTEXP, their estimated coefficients had the opposite sign from the coefficients on the first two lags. The magnitudeof the coefficients for the first two lags also increased, so the sum of the coefficients, an estimate of the response of investment to a permanentchange in the variables,was not significantlyaffectedby includingmore independent These resultsindicatemulti-collinearityamongthe longer lags of lags. IFIN and INTEXP.Also, the estimatedregressionwith two lags had a lower root mean squarederror than the equationwith three lags. For these reasons, the results reportedin the next section includetwo lags for each variable.The conclusionsdrawnare not affected by including longer lags. Several econometric problems arise when using pooled cross-section time series data to obtain regression estimates. The time series variationfor individualfirms often generates serial correlationin the errorterms, while the cross-sectionvariationamong differentfirms is The residualsfrom an ordinaryleast likely to cause heteroscedasticity. squares(OLS) regressionusing the specificationof equation(4) show both of these characteristics.Thereforewe used an estimatedgeneralized least squares techniqueto obtain more efficient estimates. The techniqueis based on the assumptionthat the error terms for each firm follow a first-orderautoregressiveprocess of the form:
(it = QiEit-1 + Vit

where it is the error for firm i at time t and vit has a zero mean and a variance a that depends on the firm. This error term allows for both serial correlationandheteroscedasticity. equationswere estimated The using a three-stageprocedure.First we used the residuals(eit) from an OLS regression to estimate Qi as
a iite

Qi=

t=2

eit-hi/

sEi t=2

t-i

9We are indebted to Joel Prakken for making this observation.

180

JOURNAL OF POST KEYNESIAN ECONOMICS

in of whereTis thenumber observations thetimeseriesforeachfirm. Thenwe obtained residuals the OLS (vit) froma transformed regression in whichall variables werereplaced Xit - iXit-l. Using Xit by
the vit residuals, we estimate u2 by
S2 = 1 T1
T iS

t=2

2t'

Thetransformed weredividedby Si to correct data heteroscedasticity in the finalregression. moredetailed A of and analysis this approach further references theeconometric to literature be foundin Judge may et al. (1985, pp. 485-490). The data were drawnfrom the COMPUSTAT annualindustrial firms Intape. 0 Datafor U.S. manufacturing (2-digitSIC-Standard dustrial Classification-codes between20 and 39) were used for the years 1967-1982.The datafrom 1967-1969wereusedonly for conso use data structing laggedvariables, theregressions investment from 1970-1982.A firm'sobservation an individual was excluded for year fromthe sampleif one of the followingthreeconditions arose: for (1) firmdatawerenot updated the year, a or thedatareflected major for (2) merger acquisition theyear,or a necessary itemwasreported notavailable theyear. as data for (3) The autocorrelation correction that the data for procedure requires eachfirmconstitute continuous a seriesin time,andonlyfirmsthathad at least six time series observations were included.All data were for for adjusted inflation dividing theimplicit by by pricedeflator U.S. national Thetotalsample contained 9,391 observations. gross product. IV. Empiricalresults in Theexplanatory powerof theinvestment theory presented section2 anddatasample. is generally specification quitegoodforourempirical All threepropositions out at the beginning section3 are well of set in regressionsrun over the whole sample and various supported investThe between showa positiverelation subsamples. regressions mentand the variables sales and internalfinance.The representing
'Standard and Poor's COMPUSTAT service maintains extensive financial data for individual U.S. and Canadian firms.

INVESTMENT THEORIES KALECKI OF AND KEYNES

181

interestexpense variableshave the expected negative sign in our estimated equations. The estimated equation for the full sample is: (5) INVST = .0185 (SALES) + .0025 (SALES-1) + .0033 (SALES_2) (16.7) (1.6) (2.3)
+ .171 (IFIN_1) + .058 (IFIN-2) - .423 (INTEXP_1)

(16.6)

(5.4)

(012.5)

- .364 (INTEXP_2) + .144 (GPLANT) (9.4) (48.8)


.059 (D71) .033 (D72) + .003 (D74) .037 (D75)

(4.7) (3.1) (4.51)

(2.7) (1.9) (7.2)

(0.2) (0.9) (0.3)

(1.4) (2.1)

.023 (D76) - .011 (D77) - .025 (D78) + .025 (D79)

+ .055 (D80) + .091 (D81) - .004 (D82)

(t statistics in parentheses) This equation includes "fixed effects" that allow separateintercepts for each firm and each year. The annual interceptsare reported for 1971 through 1982 as the coefficients on the variables D71 through D82. These coefficients are relative to a zero value for 1970. The intercepts for each of the 835 firms in sample are not reported in equation(5). The elasticities of investmentwith respect to a 1 percent variables,evaluatedat the sampermanentincreasein the independent means, are .32 for SALES, .21 for IFIN, and -. 16 for INTEXP. ple The positive and significantcoefficient on SALES is not surprising since outputor acceleratorvariableshave been foundto have explanatory powerin investmentequationsbased on a wide varietyof theoretical foundations.The results for IFIN are striking, however, since the estimatedequationshows a strong and independenteffect for internal financeeven thoughit is includedwith sales in the regression.Withthis specification, it is unlikelythatIFIN acts only as a proxy for outputor a sales accelerator. The negative coefficients on the lags of INTEXP are also consistent

182

ECONOMICS JOURNALOF POST KEYNESIAN

with the theorypresentedearlier.Holding the other variablesconstant, we find higher lagged interestexpenses indicatethat a greaterproportion of the firm's cash flow is committedto debt service and, therefore, it is more risky to finance the firm's investment expenditures. The strengthof the effect is particularly interestingsince otherstudieshave had little success in obtainingstatistically significantresults for variables used to representsimilareffects.1 Therearetwo explanationsfor the relative strengthof our interestexpensevariable.First, other studies often focusedon stock measuresof debtburden(suchas the stock of long-term debt, for example). But as Minsky (1975, chapter 5) has pointed out, the key to "validating" the liabilities a firm incurs is for the firm to generate sufficient cash flow to meet its ongoing commitments. Thus, with other things held constant,high levels of committed cashflows, thatis, high interestexpenses, will constraininvestment,as the results indicate.12The second explanationfor the significance of the INTEXP variable is that the sample used here is drawn from a period (1970-1982) with much greater financial turbulencethan the sample periods used in earlier studies. More attentionwill be paid to the adequacyof internalcash flow to meet paymentcommitmentsin an environmentwhere the risk of default is higherand the terms available for investmentfinancing are less stable. It is also interestingto examinethe patternof the annualintercepts. They tend to move pro-cyclically, althoughthey generally lag somewhatbehindchanges in GNP. The variationexplainedby these shifting interceptsis not trivial. For example, the change from 1974 to 1975 at explainsa 4 percentdropin investment,evaluated the samplemeans. These results implythatthereis greatercyclical variationin investment than can be accounted for by changes in the independentvariables alone. This may be explainedby "animal spirits," or it could be the result of other cyclical influences on investmentnot accountedfor by our model. In any case, this issue deserves furtherinvestigation, particularly with data that could be used to analyzethis effect over short periods, such as calendarquarters. To confirm the results from the pooled time series cross-section regressions, the equationwas estimatedseparatelyfor each industryin the sample. Industrieswere divided by two-digit SIC codes. Because
"See, for example, Meyer and Kuh (1957, pp. 117-121) and Meyer and Glauber (1964, pp. 88-94). '2As in other studies, we found little significance for stock measuresof debt in our investmentequation.

THEORIES KALECKI OF AND KEYNES INVESTMENT

183

Table 1

Results of separateINVST regressions for 2-digit SIC industries


Independent variable
SALES IFIN INTEXP GPLANT

Significant* coefficients withcorrect sign


13 15 20 20

Insignificant coefficients withcorrect sign


12 1 0 0

Insignificant coefficients withincorrect sign


5 1 0 0

Significant coefficients withincorrect sign


0 3 0 0

Notes: SIC = StandardIndustrialClassification;INVST = annualcapital expenditures; SALES = net annualsales; IFIN = internal finance; INTEXP = net annual interestexpense; GPLANT = book value of gross plant. *A 5 percent significance level was used for these statistics.

the sample size for each of the 20 industriesis much smaller than the aggregatepooled sample, multi-collinearityposed more of a problem. Therefore,the results from equation(5) were used to constructdistributed lags for the independentvariables. The estimated coefficients from equation(5) were normalizedto sum to unityand thenused as lag weights. Using this distributedlag specification,we estimatedseparate equationsfor each industry.Table 1 shows that the coefficients almost always had the sign predicted by the theory and were statistically significant at the 5 percent level for the majorityof the industries. For the final empiricaltest, the samplewas divided into 13 separate cross sections for the years 1970 through 1982. The investmentequation was estimated using the lag weights discussed above. These refrom the results gressions requirea somewhatdifferent interpretation presentedso far.The pooled cross-sectiontime series with firm-specific intercepts uses data corrected for differences in the mean values between firms. Therefore, the estimates are based on variationwithin the firms. The pure cross-sectionresults, presentedin Table2, are not correctedfor differences in firm averages, so the estimates are based on variationsbetweenfirms. One would expect the estimates from the two approachesto differ. It is evident from Table 2 that sales is much more importantfor explaining changes in investment across time than for explaining changes between firms. The sales coefficients are small and have unpredictable signs in the pure cross-section regressions. Interest

184

JOURNALOF POST KEYNESIAN ECONOMICS

Table2

INVST Regressions by year


Year 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
1981

SALES -.0020t .0093 .0004t .0005t -.0031t -.0019t .0023 -.0012t -.0010t -.0045 -.0066
-.0093

Estimatedcoefficientof: IFIN INTEXP .505 .367 .455 .440 .403 .341 .333 .359 .385 .443 .481
.471

GPLANT .049 .043 .041 .094 .080 .076 .071 .077 .078 .087 .086
.044

-.01 t -.248 -.091t -.128 -.330 -.250 -.215 -.093 -.243 -.162t -.101t
-.099t

1982

-.0065

.432

-.100

.078

Note: INVST = annualcapital expenditures;SALES = net annualsales; IFIN = internal finance; INTEXP = net annualinterestexpense; GPLANT = book value of gross plant. tCoefficients not significantlydifferent from zero at the 5 percent level.

expenses also have a stronger effect in the "within" regressions althoughthe coefficients all have the expectednegative sign in the cross sections. The internal finance variable, however, has a very strong effect in the cross sections. The elasticities evaluatedat sample means for each year range from 0.32 to 0.50. This suggests that internal finance is quite importantfor explaining why different firms invest different amountsat any point in time.

V. Conclusion
The resultsof this studyare favorableto the investmenttheorypresented in section 2. All three hypotheses derived from the investment model are consistent with the empiricalanalysis. Our results reaffirm the common conclusion of other investmentstudies that sales, utilization, or acceleratorvariablesare signficiant in explaininginvestment through time. The strong, independenteffect of internal finance is more interesting, however, since it is a prediction of the investment theoriesof Kalecki, Steindl, Minsky,and othersthathas not often been

OF AND KEYNES THEORIES KALECKI INVESTMENT

185

from outputeffects in otherempiricalinvestment redistinguished search. in Thenegative is effectof interest expense alsoevident theestimated equations. to Othershave attempted isolatean effect of this kind are withlittlesuccess.Thestrong obtained duein partto theuse results of interest firms'ongoingfinancial posiexpenseas a flow measuring is variables. tionrather theuseof stockliquidity than Also, thesample a muchmore turbulence drawn froma periodin whichfinancial played covered most role thanin thetimeperiod by important in theeconomy earlierstudies. for analTheseresultshaveimportant implications macroeconomic is as in whichinvestment recognized a in theKeynesian tradition, ysis of fundamental determinant the level of aggregateactivity.Firms' conditions have a signficiant financial impacton the level of investof the here the ment.Therefore, resultspresented reinforce arguments of and that Kalecki,Keynes,Minsky, Davidson, others theavailability of financeplays a basic role in the determination macroeconomic performance. Severalextensions suggested this paperfor future are research. by or First, since this paperfocuses on the Kaleckian post-Keynesian the results shouldbe tested directly theoriesof investment alone, theoretical foundations, especially againstthosebasedon alternative the neoclassical and theoryof optimalcapitalaccumulation financial of in of theories investment the tradition Tobin's"q" theory.13 Also, the analysisof separate cross sections over time could be further withvarious refined studyhow investment to relations cyclical change in to characteristics the economy.Finally,it will be interesting intethe empiricalinvestmentspecificationdevelopedhere into grate of broader macroeconomic modelsto considerthe implications this workfor macroeconomic and stability policy design. an model Overall, this paperreestablishes empiricalinvestment of basedon the theoretical tradition KaleckiandKeynesas a serious alternative the conventional to that approaches havecome to receive in wideracceptance the recentliterature. resultsare quitestrong The andtheysuggestthe needfor a renewed effortto empirically compare the implications the Kaleckian post-Keynesian to of and alternatives otherapproaches.
'3See Fazzari and Athey (1986) for an analysis of the importanceof financing variables in neoclassical and acceleratorinvestmentspecifications.

186

JOURNALOF POST KEYNESIAN ECONOMICS

REFERENCES
Anderson, Locke. CorporateFinance and Fixed Investment.Boston: Division of Research, GraduateSchool of Business Administration,HarvardUniversity Press, 1964. Baran, Paul A., and Sweezy, Paul A. MonopolyCapital. New York: MonthlyReview Press, 1966. Bischoff, Charles. "Business Investmentin the 1970's: A Comparisonof Models." Brookings Papers, 1971, pp. 13-63. Clark, Peter. "Investmentin the 1970's: Theory, Performance,and Prediction." Brookings Papers, 1979, pp. 73-124. Coen, Robert M. "The Effect of Cash Flow on the Speed of Adjustment." In TimeIncentives and Capital Spending. Ed. by Gary Fromm. Washington,D.C.: Brookings, 1971. Davidson, Paul. Moneyand the Real World.London: Macmillan, 1972. Duesenberry,James S. Business Cycles and Economic Growth. New York: McGraw-Hill, 1958. Eisner, Robert. Factors in Business Investment.Cambridge,Mass.: National Bureauof Economic Research, 1978. Elliot, J. W. "Theories of CorporateInvestmentBehavior Revisited." AmericanEconomic Review, 1973, 63, 195-207. Evans, Michael. "A Study of IndustryInvestmentDecisions." Reviewof Economicsand Statistics, 1967, 49, 151-164. . MacroeconomicActivity.New York:Harperand Row, 1969. Fazzari, Steven, and Athey, Michael. "Investmentin ImperfectCapital Markets:Financing Constraintsand the Neoclassical Model." Workingpaper,June 1986. Feldstein, M. S., and Foot, A. K. "The OtherHalf of Gross Investment:Replacementand ModernizationExpenditures." Reviewof Economics and Statistics, 1971, 53, 49-55. Fuller, W. A., and Battese, G. E. "Estimationof Linear Models with Crossed-Error Structure."Journal of Econometrics, 1974, 2, 67-78. Goldsmith, Raymond. The Flow of Capital Funds in the Post-WarEconomy, New York: National Bureauof Economic Research, 1965. Jorgenson, Dale W. "Capital Theory and InvestmentBehavior." AmericanEconomicReview, 1963, 53, 247-259. . "The Theory of InvestmentBehavior." In Determinantsof InvestmentBehavior. Ed. by Robert Ferber.New York: National Bureauof Economic Research, 1967. . "EconometricStudies of InvestmentBehavior: A Survey." Journal of Economic Literature, 1971, 9, 1111-1147. Judge, George G.; Griffiths, W. E.; Hill, R. C.; Luktepohl,H.; and Lee, T. C. The Theory and Practice of Econometrics.2nd ed. New York:John Wiley & Sons, 1985. Kalecki, Michal. "The Principleof IncreasingRisk." Economica, 1937, 4 (new series), 440447. . The Theoryof Economic Dynamics. New York: Kelly, 1969. . Selected Essays on the Dynamics of the Capitalist Economy 1933-1970. Cambridge: CambridgeUniversity Press, 1971. Keynes, John M. The General Theoryof Employment,Interest, and Money. London: Macmillan, 1936.

AND KEYNES THEORIES KALECKI OF INVESTMENT

187

Kuh, Edwin. Capital Stock Growth:A Micro-Econometric Approach. Amsterdam:North Holland, 1963. Meyer, John, and Glauber,Robert. InvestmentDecisions, EconomicForecasting, and Public Policy. Boston: Division of Research, GraduateSchool of Business Administration,Harvard University, 1964. Meyer, John, and Kuh, Edwin. The InvestmentDecision. Cambridge,Mass.: HarvardUniversity Press, 1957. Minsky, Hyman P. John MaynardKeynes. New York: ColumbiaUniversity Press, 1975. Mott, Tracy L. Kalecki's Principleof IncreasingRisk: The Role of Finance in the PostKeynesianTheory of InvestmentFluctuations.Ph.D. dissertation, StanfordUniversity, 1982. . "Kalecki's Principleof IncreasingRisk and the Relation Among Mark-upPricing, InvestmentFluctuations,and Liquidity Preference." EconomicForum, 1986, 15, 65-76. Firms: A QuarterlyTime Series Analysis of Resek, Robert. "Investmentby Manufacturing IndustryData." Reviewof Economicsand Statistics, 1966, 48, 322-333. Steindl, Josef. Maturityand Stagnationin AmericanCapitalism. New York: MonthlyReview Press, 1976. of Witte, James G. "The Microfoundations the Social InvestmentFunction." Journal of Political Economy, 1963, 71, 441-456. Wood, Adrian. A Theoryof Profits. Cambridge:CambridgeUniversity Press, 1975.

Вам также может понравиться