LAWRENCE SUMMERS Harvard University CHRIS CARROLL Massachusetts Institute of Technology WhyIs U.S. National Saving So Low? LOW Americannationalsavingrate has long been a serious concern to economic policymakers.Increasingnational saving and investment was a principalobjective of the Economic Recovery Tax Act of 1981 and of the supply-sideeconomic policies that accompaniedit. Yet the nationalsaving rate, at least as measuredin the National Income and ProductAccounts, has declinedsharplyduringthe 1980s.Overthe past five years, nationalsavinghas averagedonly 2.3 percentof full-employmentGNP, comparedwith 7.4 percentduring1960-80. In 1986,American net nationalsaving was below 2 percent of GNP, less than half the ratein Britain,less than 30 percent of the rate in France and Germany, andonly 10percentof the rate in Japan. The unprecedentedU.S. governmentbudget deficits of recent years areoften singledout for blame. Governmentdissavingas a shareof fullemploymentGNP indeed increased by 2.5 percentagepoints between 1960-81and 1982-86.Still, risingdeficitscan accountfor only abouthalf of the decline in the nationalsaving rate between the two periods. The remainderis attributableto a roughlyequaldecline in the privatesaving rate. Much of that decline is in turn traceableto a fall in the personal savingrate.Low personalandprivatesavingratesareespecially striking giventhe widespreadhope thatthe tax incentives enactedin 1981would increasepersonalandprivatesaving. Theeconomiceffects of the low U. S. savingratedependon the ability of the United States to sustain large capitalinflows from abroad. Over the past several years, more than half of net domestic investment has THE Wearegratefulto DavidCutlerforresearchassistanceandto membersof the Brookings Panelformanyusefulcomments. 607 608 Brookings Papers on Economic Activity, 2:1987 been financedby internationalcapitalflows. If nationalsaving remains low, either these capital flows, with the attendantdislocations in the economy's tradedgoods sector, will continue, or investmentwill drop off precipitously.Whatfinallyhappenswill depend in largepart on the response of foreign investors and governmentsto chronic capital outflows. Neither outcomewouldimproveAmericaninternationalcompetitiveness. This report examines several issues raised by the currently low Americannationalsaving rate. Are the declines in nationalsaving real, or arethey insteadthe resultof faultymeasurementthatignoresthe huge capitalgains generatedin the stock marketfrom 1982to the fall of 1987? Does recent experience contradictthe Ricardianequivalenceidea that governmentdeficits call forth increased private saving or the supplyside idea that tax incentives can spur private saving? What forces lie behindthe apparentseculardowntrendin privatesaving?To whatextent can capitalflows fromabroadsubstitutefor domestic saving? We concludethatthe low nationalsavingrate duringthe 1980scannot be attributedto measurementproblems.It is traceableto a combination of federal deficits and a continuationof a long-termdownwardtrendin private and personal saving. Private saving would probablyhave been stilllower duringthe 1980sif the federalgovernmenthadnot encouraged savingwith new tax incentives. However, the most reliableway for the federal government to increase national saving is to reduce its own borrowing.Withoutanincreasein nationalsaving, andgiventhe increasing reluctanceof foreigninvestorsto hold Americanassets, it is unlikely that even currentlevels of investmentcan be maintained. Trends in Saving Table 1 presents saving data from the National Income and Product Accounts, measuredon both a standardand an inflation-adjustedbasis. The inflationadjustmentis necessarybecause in an inflationaryenvironment interest received (or paid) reflects in part compensationfor the erosion in the value of nominalassets and so is not properlytreatedas income.1 The table also presents measures of the ratio of private 1. For a careful discussion of inflationadjustmentand a numberof other possible adjustmentsto standardmeasuresof saving, see Derek W. Blades and Peter H. Sturm, Lawrence Summers and Chris Carroll 609 consumption to "private GNP," defined as GNP less government outlays. If depreciationis mistakenlyoverstatedin the nationalincome accounts, measurednet savingwill be understated.However, this error would not affect the ratio of consumptionto GNP, so the consumption ratioprovides a useful check on the robustnessof conclusions reached using the savingrate data. The table shows the dramaticdeclinein the nationalsavingratein the 1980s, to less than 2 percent of full-employmentGNP in 1986and less than 3 percent during 1981-86-less than half the rate during any preceding five-year interval. Abnormallylow national saving is also evidenced by the high ratio of consumptionto private GNP in recent years. The shareof total consumptionoutlays in privateGNP reacheda postwarhighin 1986afterincreasingsharplyfollowingthe 1981tax cut. As a furthercheck on the robustnessof ourconclusions, figure1plots the NIPA saving rate along with two further variants-the national savingrateinferredfromflow-of-fundsdataand the nationalsavingrate measuredinclusive of net additionsto the stock of consumerdurables.2 The flow-of-fundsdata provide an independentmeasure of saving by tallying increases in asset stocks rather than estimating a residual between income and spending. The consumer-durablesadjustment recognizes that purchases of consumer durables provide for future consumptionservices and so are a form of saving. Both adjustments confirmthat nationalsavinghas declineddramaticallyin the 1980s. The 5 percentagepoint decline in the NIPA savingratefrom 1960-81 to 1982-86may be apportionedevenly, as table 1 shows, between a 2.5 pointincreasein governmentdissavinganda 2.5 point decline in private saving. On an inflation-adjustedbasis, the results are similar,with 2.9 percentagepoints attributableto the public sector and 1.9 points, to the private sector. For 1986, a somewhat larger share of the decline in national saving may be traced to the private sector. By contrast, extendingthe intervalback to the 1950s suggests a somewhat greater "The ConceptandMeasurementof Savings:The United States andOtherIndustrialized Countries,"in FederalReserveBankof Boston, Savingand GovernmentPolicy (Boston: FRBB, 1982),pp. 1-30; or PatricH. Hendershottand Joe Peek, "PrivateSavingin the UnitedStates, 1950-85,"WorkingPaper2294(NationalBureauof EconomicResearch, June1987). 2. Fora discussionof the flow-of-fundssavingdata,see Frankde Leeuw, "Conflicting Measuresof PrivateSavings,"Surveyof CurrentBusiness, vol. 64 (November1984),pp. 17-23. 610 Brookings Papers on Economic Activity, 2:1987 Table 1. National Income and Product Accounts (NIPA) Saving and Consumption Measures, 1950-86 Percent of full-employment GNP unless otherwise indicated Feder al plus state anzdlocal Private savinzg InflatiotnReported adjusteda governmnentsaving Inflation adjusteda Reported Consumption (percent of fullemployment private GNP)b All Except goods durables Year NIPA national saving 1950 1951 1952 1953 1954 10.4 10.1 7.0 5.7 5.1 7.5 8.2 8.1 7.7 7.0 ... ... 6.5 7.2 6.8 2.9 2.0 - 1.1 -2.0 -1.9 ... ... 0.3 -1.6 -1.7 77.0 76.2 79.4 80.5 80.9 64.7 65.3 68.8 69.2 70.1 1955 1956 1957 1958 1959 8.6 9.3 8.0 4.6 7.2 7.8 8.1 7.8 7.3 7.5 6.8 6.4 6.4 6.6 6.2 0.8 1.2 0.2 -2.7 -0.3 1.7 2.8 1.5 -2.0 1.0 78.0 77.6 78.4 81.5 79.5 66.2 66.7 67.5 71.2 68.7 1960 1961 1962 1963 1964 7.2 6.3 7.3 7.8 8.5 6.6 7.1 7.9 7.6 8.9 5.7 6.5 7.0 6.8 8.3 0.6 -0.8 -0.6 0.1 -0.4 1.5 -0.2 0.2 0.9 0.1 79.7 80.2 79.3 79.0 78.7 69.2 70.3 69.0 68.3 67.8 1965 1966 1967 1968 1969 9.7 9.3 7.9 7.7 8.3 9.7 9.5 9.7 8.4 7.3 8.6 7.9 8.7 6.5 5.6 0.1 -0.2 - 1.8 -0.7 1.1 1.0 1.2 -0.9 1.0 2.6 77.8 77.8 79.1 79.5 79.0 66.6 66.6 68.0 67.8 67.6 1970 1971 1972 1973 1974 6.5 6.7 7.7 10.1 7.4 7.4 8.4 8.0 9.5 7.9 5.8 6.7 6.8 7.2 5.0 - 1.0 - 1.7 -0.3 0.6 -0.3 0.5 -0.1 0.9 2.9 2.7 80.3 79.5 78.7 76.6 78.1 69.5 68.2 67.1 65.2 67.5 role for governmentdeficits in accountingfor the low level of national saving. It seems fair to conclude that in an arithmeticsense the low nationalsavingrate in the 1980sis a reflectionof declines in both public andprivatesaving. National Saving and the Ricardian Equivalence Proposition The premisethatnationalsavingcan be analyzedfruitfullyas the sum of independentprivate and public components has been challengedby Lawrence Summers and Chris Carroll 611 Table 1. (continued) Percentof full-employmentGNP unless otherwiseindicated Private saving InflationRepor ted adjuisteda Federal plus state and local Conlsuimptioni (percenit offullemployment governmenit saving Ilifjation Repor ted adjusteda private GNP)b All Except goods durables Year NIPA national saving 1975 1976 1977 1978 1979 4.6 5.6 6.6 7.9 7.7 8.4 7.7 7.6 7.9 7.2 6.2 6.1 5.6 5.8 4.6 - 3.9 -2.1 -0.9 0.0 0.5 - 1.6 -0.5 1.1 2.2 3.1 80.2 79.2 78.4 76.9 76.8 69.4 67.9 66.9 65.7 66.1 1980 1981 1982 1983 1984 5.0 5.5 1.8 1.8 3.9 6.2 6.4 5.0 5.3 6.6 3.5 4.4 3.7 4.2 5.6 - 1.2 -0.9 -3.2 -3.5 -2.6 1.7 1.2 - 1.7 -2.2 - 1.3 78.7 77.7 81.2 81.8 80.1 68.7 68.0 71.2 71.2 69.2 1985 1986 2.3 1.7 5.5 5.1 4.4 4.6 - 3.3 - 3.2 - 1.9 -2.6 81.7 82.7 70.4 71.1 Averages 1951-55c 1956-60 1961-65 1966-70 1971-75 1976-80 1981-86 7.3 7.3 7.9 7.9 7.3 6.6 2.8 7.8 7.5 8.2 8.5 8.4 7.3 5.6 6.8 6.2 7.5 6.9 6.4 5.1 4.5 -0.4 -0.2 -0.3 -0.5 - 1.1 -0.8 -2.8 -0.3 1.0 0.4 0.9 1.0 1.5 - 1.4 79.0 79.3 79.0 79.1 78.6 78.0 81.0 67.9 68.7 68.4 67.9 67.5 67.1 70.2 1950-86 6.7 7.6 6.1 -0.8 0.4 79.1 68.2 Source: Actual saving, consumption, and price data are from the National Income and Product Accounts. Stocks of assets used to compute the adjusted series are from Board of Governors of the Federal Reserve System, "Balance Sheets for the U.S. Economy, 1946-85" (Federal Reserve Board Release C.9, 1986), and "Flow of Funds Accounts," various issues. Full-employment GNP is from Robert J. Gordon, Macroeconiomzics (Little, Brown, and Company, 1984), and calculations by the Congressional Budget Office. a. Inflation-adjusted saving is computed by subtracting from measured saving the product of the inflation rate (GNP deflator) and net interest-bearing assets held by each sector. b. Full-employment GNP less total government outlays. c. For inflation-adjusted series, 1952-55. proponentsof the increasinglypopularRicardianequivalenceproposition, which holds that the timing of governmenttax payments has no impact on an economy's level of national saving.3 If the government runsa budgetdeficit,consumerswill anticipatethe subsequentincrease 3. Theclassicmodemstatementof the Ricardianequivalencepropositionis RobertJ. Barro, "Are GovernmentBonds Net Wealth?"Journal of Political Economy, vol. 82 (November-December1974),pp. 1095-1117.A skepticalsurveyof the largeliteratureon Ricardianequivalencemay be found in B. DouglasBernheim,"RicardianEquivalence: AnEvaluationof TheoryandEvidence,"in StanleyFischer,ed., NBERMacroeconomics Brookings Papers on Economic Activity, 2:1987 612 Figure 1. National Saving as a Fraction of Full-Employment GNP, 1952-86 Percent 14~~~~~~~~~ NIPA basis plus 12 14 .-p 10 ~~~~~~~ ~~~ ~consumer 8- fb .,Purchases % 4 basis Flo w -offunds basis of pualus %~v 21955 1960 1965 1970 1975 1980 1985 Source: Full-employment GNP from Robert J. Gordon, Macroeconomics (Little, Brown, and Company, 1984), and calculations by the Congressional Budget Office. NIPA basis national saving is net national saving from National Income and Product Accounts, table 5.1. Flow-of-funds saving measures the change in the value of the total asset stock from Board of Governors of the Federal Reserve System, "Balance Sheets for the U.S. Economy, 1946-85" (Federal Reserve Board Release C.9, 1986), and "Flow of Funds Accounts," various issues. in taxes that will be necessary to repay the debt, and so will raise their saving.Privatesavingwill riseto offset the declinein governmentsaving, leavingnationalsavingunaffected. The Ricardianview may be put in a differentway: governmentbonds are not a formof wealth. Whilegovernmentbonds representan asset to those who hold them, they representan exactly offsettingliabilityto the taxpayerswho will ultimatelyredeem them. Hence deficitpolicies that increase the outstandingstock of governmentdebt do not increasetotal spending, a conclusion that contrasts sharplywith Keynesian conceptions of the effects of fiscal policies. Traditionally,Keynesians have debatedthe extent to which increasesin demandcaused by government tax cuts are crowded out. The Ricardianposition is that crowdingout takes place before the fact. There is no initialincrease in demandwhen the governmentcuts taxes. The Ricardianequivalencepropositiondependson two assumptions: Annual,1987(MITPress, 1987),pp. 263-304. Ourdiscussiondrawsheavilyon JamesM. PoterbaandLawrenceH. Summers,"FiniteLifetimesandthe Effects of BudgetDeficits on National Savings," Journal of MonetaryEconomics, vol. 20 (September1987),pp. 369-91. Lawrence Summers and Chris Carroll 613 that consumerssmooth consumptionover long periodsof time and that they foresee the long-term implications of the government's budget constraint.The validityof both assumptionsis open to question. There is substantialevidence that many families' consumption tracks their income extremely closely, and casual observation suggests that consumersrarelythinkaboutthe tax increasesthatwill be necessary to pay off a risingnationaldebt when they maketheirconsumptiondecisions.4 Ultimately,however, ajudgmentaboutRicardianequivalencemustrest on empiricalgrounds. Empirical evaluation of Ricardian equivalence has been difficult because untilrecentlythere has been relativelylittle variationin federal deficitsindependentof wars, cyclical fluctuations,andinflation,each of whichmightbe expected to have a systematicimpacton nationalsaving independentof the effect of the budget deficit. Recent federal deficits, though,are far largerthanwould be predictedon the basis of historical relationshipsbetween deficits and macroeconomicconditions, so they providea naturaltestinggroundfor Ricardianequivalence. The raw data in table 1 appear to refute decisively the Ricardian equivalence proposition. Increases in government deficits have been associated with decreases, not increases, in private saving. There is, however,alwaysthe possibilitythatthe apparentrefutationof Ricardian equivalence is spurious. Some independent development may have caused measurednationalsaving to decline in recent years, creatinga spuriouscorrelationwith the rise in budget deficits. Alternatively,the nationalincome accounts may mismeasurenationalsaving. Perhaps the most plausible argument is that NIPA saving is an inappropriatemeasure because it ignores capital gains and losses on existingassets, which representincreases or decreases in wealth and so should be treated as positive or negative saving. The increase in the stockmarketbetween 1982andthe fallof 1987,forexample,substantially increasedmeasuredhousehold wealth but was not reflected in official statistics on saving. To examine the importanceof such capital revaluationswe useddatafromthe FederalReserveBoard'sNationalBalance Sheetsto constructtimeseriesmeasuresof thenet worthof the household 4. Fora surveyof therelevantliteratureon liquidityconstraints,see R. GlennHubbard andKennethL. Judd,"LiquidityConstraints,Fiscal Policy, andConsumption,"BPEA, 1:1986,pp. 1-50, andRobertE. Hall, "RealInterestandConsumption,"WorkingPaper 1694(NationalBureauof EconomicResearch,August1985). 614 Brookings Papers on Economic Activity, 2:1987 sector. Figure2 plots the changein households'realnet worthmeasured as a fractionof potentialGNP. Because the series is extremelyvolatile, it is difficultto judge whetherthis measureof nationalsaving is abnormally low in recent years. Robert Barro and other defenders of the Ricardianequivalencepropositionhave cited this difficultyas evidence that the data do not yet permit a firmjudgment about the impact of deficitson nationalsaving.' However, simply looking at the average value of wealth saving in recent years does not provide a satisfactory test of whether wealth saving has been reducedby budgetdeficits. Logically, if the concept of saving is expanded to include the capital gains and losses associated with asset revaluations,any income measureshouldbe expandedin the same way. This adjustmenthas importanteffects. By their nature, the capitalgains and losses associated with asset revaluationsare likely to be transitory since, in well-functioningmarkets, speculation would arbitrageaway any large expected capital gains or losses. One would therefore expect that in periods when the stock marketrose sharply, makingtransitoryincome positive, wealth saving would be large. The fact that wealth saving has not been abnormallyhigh duringthe 1980s might then be evidence that budget deficits are depressing national saving. To examine this possibility and the related possibility that the low rate of national saving reported in the national income accounts was caused by macroeconomicconditions rather than budget deficits, we estimatedregressionsrelatingboththe NIPA andwealthsavingmeasures to various macroeconomic variables during 1950-81 (table 2a). The macroeconomicvariables included the GNP gap expressed as a percentage of GNP, inflationrates, and the value of real capitalgains and losses on the stock market and on the housing stock expressed as a fractionof GNP. A time trend was also included in the equations. We then used the results to forecast the nationalsavingrate during1982-86 (table2b). If the Ricardianequivalencepropositionis correctandnational saving in the 1980shas not been sharplyreduced by budget deficits, it 5. See, for example,RobertJ. Barro,"RicardianEquivalence"(HarvardUniversity, 1987).It is worth noting that the bulk of the volatilityof the wealth series comes from extremelyvariablestock marketwealth,andthatthe 1982-87increasesin such wealthare the chief explanationwhy the rate of wealth saving has not been notablylow over this period. Lawrence Sulmmers and Chris Carroll 615 Figure 2. Annual Change in Real Household Net Worth as a Fraction of Real FullEmployment GNP, 1953-86 Percenlt 30 2520 15 10 5 0 -5 - 10 - 15 j 1955 1960 1965 1970 1975 1980 1985 Source: Change in real household net worth from Federal Reserve Board, "Balance Sheets for the U.S. Economy," and "Flow of Funds Accounts." Full-employment GNP from Gordon, Macroeconomnics, and calculations by the Congressional Budget Office. shouldbe possible to findequationsthatdo not consistently overpredict saving. Results that are typical of the many equationsthat we estimatedare reportedintable2a. All of the equationssignificantlyoverpredictnational saving in the 1980s. The predictionerrorsare substantivelysignificant in mostcases andarefrequentlystatisticallysignificantwhen the official savingmeasureis examined. The errorsare frequentlyon the order of severalpercentagepoints of GNP. In the case of the equationsusing the wealth saving concept, the errors are often much larger,reflectingthe fact that previous stock market rallies, unlike the 1982-87 one, have been associatedwith highrates of nationalsaving.6 The table also shows that, as theory would predict, saving responds differentlyto capitalgains and losses than to other forms of income. In 6. An alternativeway of demonstratingthat deficits have reduced national saving wouldhavebeento show thatdeficitmeasuresreceive negativecoefficientswhen entered as independentvariablesin the equationsshown in table 2a. While deficitvariablesare consistentlystatisticallysignificantwhenenteredintoequationsexplainingnationalsaving, the size of theircoefficientis quitesensitiveto detailsof specification. to = = 00 c aZr o * b.t o Z ?5X, 2:,<~~~~~~~4. c f~~~~~~~n Z0 = ? U (:Q 00t _ < 00 00 X >X ^~~~~~~~~~~~~~~~~ ^ r_4-EE > 4? ^ 0 Z *_~~~~~~~~~~~~~~~~' . -ae .Z i~~~~~~~~~~~~~~~~~~~ O U ^ N ~00 t- 00 C) cq z~~~~CN C) O N O\ >~~~~~~~I \_ \ 00 O0 YQn Q c= _s _s _s _s ,O 00 r- C) _s _s _s Z ; 08b0 r@GC S~~~~~~~~~~~~Z 00 ',.C C) ~~~~~c ~ X t- . 40 ) N O Q E Q N O N 0 0 ^ ~~~~~~~~~~~~~ r-@ .0~~ e~~~~~~ ~~~~~) ( O 00 00 X~ ~4o ~~~~~C = a o Z Eu O t) O 00 t- tva 00 O- m O) "C - " C - - C) C) C) ceS( ce ~~Z N : r) ,Z t Z < u? C) bu aN G O ug = r 0 o o N0 C) u 'A = ;O= ' o ~~~~~= 0 - Lawrence Sulmmers and Chris Carroll 617 Table 2b. Prediction Errors Generated by Equations Explaining Private and Personal Saving Measures, 1982-86 Percent Predictioni errora Equation 1. National saving 2. National saving 3. National saving 4. National saving 5. Change in wealth 6. Change in wealth 7. Change in wealth 1982 1983 - 0.70 (-0.921) - 0.05 (-0.061) - 0.45 (-0.435) -1.15 (-1.364) -31.15 (-2.562) -17.09 (- 1.468) -17.53 (- 1.172) - 1.02 (-1.164) -0.16 (-0.159) -0.55 (-0.469) -1.36 (-1.456) -24.98 (-1.768) -13.47 (- 1.046) -13.99 (-0.831) 1984 - 1.66 (-1.999) -0.56 (-0.525) - 0.95 (-0.735) -2.00 (-2.273) -18.22 (-1.356) -8.55 (-0.706) -9.22 (-0.501) 1985 1986 - 2.92 (-3.231) -1.57 (- 1.285) -2.17 (- 1.428) -3.42 (-3.338) -26.40 (-1.791) -9.27 (-0.656) -10.06 (-0.464) - 2.22 (-2.255) -0.62 (-0.446) -1.09 (-0.682) -2.43 (-2.344) -34.82 (-2.156) -26.32 (- 1.844) -27.18 (-1.197) Source: Equations in table 2a (estimated over the period 1950-81) used to forecast saving, 1982-86. a. Errors are expressed as realized minus predicted saving rates. Numbers in parentheses are t-statistics computed using the standard error of the corresponding equation in table 2a. the equations that treat capital gains as a component of saving and income, the marginalpropensity to save out of stock market gains is close to unity. This correlationsuggests that there is little to be gained from includingcapital gains and losses in measuringsaving. Finally, even before 1981, there is evidence of a long-termdownwardtrend in nationalsaving. Whateverthetheoreticalmeritsof theRicardianequivalencedoctrine, these results refute it as an empiricalproposition about U.S. budget deficits. It is therefore legitimate to ascribe a substantialpart of the decline in nationalsaving duringthe 1980s to budget deficits. But the increasein governmentdeficitsfrom0.8 percentof GNP during1976-80 to 2.8 percentof GNP during1981-86cannot possibly explain all of the simultaneous3.8 percentdropin the nationalsavingrate. Private Saving in the 1980s Because private saving does not appear to be tied to changes in governmentborrowingpatterns,it is probablybest measuredrelativeto privateincome, which we define as the sum of disposable income and 618 Brookings Papers on Economic Activity, 2:1987 Table 3. Composition of Private Saving, 1950-86 Percent of disposable private incomea Corporate Personal Inflationand Totalprivate saving Year Inflationand NIPA Inflation- NIPA Pension- pension- NIPA Pension- pensionreported adjusted" reported adjustedc adjusted reported adjustedc adjusted ... 1950 1951 1952 1953 1954 9.7 10.7 10.8 10.3 9.7 ... ... 8.6 9.7 9.4 5.8 7.0 7.0 7.0 6.1 ... ... ... 6.2 5.3 ... ... ... 5.6 5.0 3.8 3.7 3.8 3.3 3.7 ... ... 4.1 4.4 ... 4.1 4.4 1955 1956 1957 1958 1959 10.5 11.0 10.7 10.2 10.4 9.1 8.7 8.8 9.3 8.6 5.4 6.9 7.0 7.3 6.1 4.6 6.0 6.0 6.3 5.1 3.2 3.4 3.9 5.3 3.0 5.0 4.1 3.8 3.0 4.4 5.8 5.0 4.7 3.9 5.3 6.0 5.2 4.9 4.0 5.6 1960 1961 1962 1963 1964 9.3 10.1 11.0 10.6 11.9 8.0 9.2 9.8 9.5 11.2 5.6 6.4 6.2 5.6 6.6 4.6 5.4 5.2 4.6 5.6 3.1 4.4 3.8 3.2 4.7 3.7 3.6 4.8 5.0 5.3 4.7 4.6 5.8 6.0 6.4 4.9 4.7 6.0 6.2 6.5 1965 1966 1967 1968 1969 12.7 12.4 12.9 11.2 9.9 11.3 10.4 11.5 8.7 7.6 6.6 6.4 7.6 6.7 6.2 5.5 5.3 6.5 5.5 5.0 3.8 2.7 4.6 1.9 1.7 6.0 6.0 5.3 4.6 3.7 7.1 7.1 6.4 5.8 4.9 7.5 7.7 6.9 6.8 5.9 1970 1971 1972 1973 1974 10.3 11.6 10.9 12.8 11.0 8.1 9.3 9.4 9.6 7.0 7.9 8.3 7.0 9.0 9.1 6.7 7.0 5.7 7.7 7.6 3.3 3.5 3.2 2.1 0.0 2.5 3.3 3.9 3.7 1.9 3.6 4.6 5.2 5.1 3.4 4.8 5.8 6.2 7.5 7.0 ... corporateretainedearnings.Table 3 presents estimates of the private saving rate thus measured,along with its personal and corporatecomponents. In additionto the standardmeasuresof personalandcorporate saving, the dataare reportedwith two adjustments.The rationalefor the inflationadjustment,alreadynoted,is thatneutralchangesin the inflation ratethatdo not affect realinterestrateswould otherwisehave an impact on measuredsavingrates. The pension adjustmentis necessary because the national income accounts treat all contributionsto pension plans and income earnedby pension plans as personal income. Benefits paid out by pension plans are not treated as a component of income, since doing so would be double countingin the same way as it would be double countingto treat Lawrence Summers and Chris Carroll 619 Table 3. (continued) Percent of disposable private incomea Corporate Personal Year InflationInflationand and Totalprivate saving NIPA Inflation- NIPA Pension- pension- NIPA Pension- pensionreported adjusted" reported adjustedc adjusted reported adjustedc adjusted 1975 1976 1977 1978 1979 12.0 10.9 10.6 11.1 10.0 8.9 8.7 7.9 8.1 6.5 8.9 7.4 6.3 6.8 6.6 7.2 5.6 4.5 4.9 4.5 1.9 1.9 -0.3 -0.4 - 2.1 3.2 3.6 4.3 4.3 3.5 4.8 5.4 6.1 6.2 5.5 6.9 6.8 8.2 8.6 8.6 1980 1981 1982 1983 1984 8.9 9.3 7.6 7.8 9.3 5.0 6.4 5.6 6.2 7.9 7.0 7.3 6.7 5.2 5.9 4.6 5.2 4.7 3.3 4.2 -2.6 -0.4 0.9 0.3 1.3 1.9 2.0 0.9 2.6 3.4 4.3 4.2 2.9 4.6 5.2 7.7 6.8 4.7 5.9 6.6 1985 1986 7.7 7.2 6.3 6.5 4.3 4.2 2.8 3.0 0.0 1.7 3.4 3.0 4.9 4.2 6.3 4.8 1956-60 1961-65 1966-70 1971-75 1976-80 1981-86 10.4 10.3 11.3 11.4 11.7 10.3 8.2 9.2 8.7 10.2 9.3 8.8 7.2 6.5 6.5 6.5 6.3 6.9 8.5 6.8 5.6 5.4 5.6 5.3 5.8 7.0 4.8 3.9 4.6 3.7 4.0 2.8 2.1 -0.7 0.6 3.9 3.8 5.0 4.4 3.2 3.5 2.5 4.8 4.7 6.0 5.6 4.6 5.5 4.3 4.8 4.9 6.2 6.4 6.7 8.0 5.9 1950-86 10.4 8.5 6.7 5.3 2.3 3.7 5.1 6.2 Averages 1951_55d Sources: Actual saving, income, consumption, and price data from NIPA. Adjusted personal and private series computed by the author; adjusted corporate series based on calculations by James Poterba, "Tax Policy and Corporate Saving," BPEA, 2:1987. Financial asset stock data used in computing the adjustments are from Federal Reserve Board, "Balance Sheets for the U.S. Economy," and "Flow of Funds Accounts." a. Personal disposable income plus retained earnings. b. Calculated as described in table 1. c. Adjustment for defined-benefit pension plans. The method of adjustment is defined in detail in the text. d. Averages begin with earliest printed number in the column and go through 1955. withdrawalsfrom bankaccounts or proceeds from stock sales as a form of personalincome. Such treatmentis naturalfor defined-contribution plans,in whicha workerdirectlyowns a pensionaccountthatis invested at his discretionso that his pension contributionsor reinvestmentsof pensionincomearejust anotherformof saving. Most private pension plans, however, are of the defined-benefit, variety, in which employerscommitto ratherthandefined-contribution, provideworkerswith a pensionbasedprimarilyon finalsalaryandyears of service. The employersthen fund the impliedcontractualliabilityas they see fit, and retiredworkersreceive a streamof income that bears 620 Brookings Papers on Economic Activity, 2:1987 no necessary relationship to the past saving that the employer has undertakento fund that liability.It thereforeseems most appropriateto treatbenefitpaymentsfrom defined-benefitplans as disposableincome andto regardpensioncontributionsandinvestmentincomeas the saving of employers rather than of pension beneficiaries.7This treatment precisely parallelsthe NIPA treatmentof state and local pensions and social security.8 The available data do not permit a precise adjustmentfor pension saving. Since 72 percent of pensions are of the defined-benefittype, we added 72 percent of pension benefits paid by private pensions to householdsavingand subtractedthe same figurefromcorporatesaving, and also switched 72 percent of contributionsto pension funds and imputed interest earned on pension assets from personal saving to corporatesaving.9 No matterhow the measurementissues are resolved, private saving has trendeddownwardover the past fifteen or twenty years after rising during the 1950s and early 1960s, although the downward trend is considerablymore pronouncedin the inflation-adjustedseries than in the unadjustedseries. On an inflation-adjustedbasis, the privatesaving rate has fallen by more than one-thirdfrom its high in the early 1960s. When adjustmentsare made for inflation and pensions, the average personal saving rate over the past decade has actually been negative. Further,it appearsthat most of the decline in inflation-adjustedprivate savingcan be tracedto decliningpersonalsaving. Table 4a presents regression equations directed at the question of whetherthe recent course of privatesavingis aberrantor insteadsimply reflects the continuation of secular trends and the effects of recent macroeconomicconditions. We relate both inflation-adjustedand un7. B. DouglasBernheimandJohnB. Shoven, "PensionFundingand Saving,"in Zvi Bodie,JohnB. Shoven,andDavidA. Wise,eds., Pensions in the U. S. Economy (University of ChicagoPress, forthcoming),emphasizethe importanceof pensionissues in evaluating movementsin personalsaving.Forfurtherdiscussionof the needto adjustfor pensionsin assessingpersonalandcorporatesaving,see JamesM. Poterba,"TaxPolicyandCorporate Saving,"BPEA, 2:1987. 8. This treatmenthas the well-knowndefect that the officialmeasureof government savingdoes not reflectpensionliabilitiesthatthe governmentincurs.In the sameway, the treatmentof defined-benefitplanscontemplatedhere does not treatthe pensionliabilities incurredby corporationsas an offset to theirsaving. 9. This figurecomes fromBernheimandShoven, "PensionFunding,"p. 6, table 1. Lawrence Summers and Chris Carroll 621 adjusted measures of private and personal saving to trend variables, disposableincome, inflation,and measuresof capitalgains on the stock marketand owner-occupiedhousing. The equations are estimated for 1954-81andare used to predictprivatesavingrates over the succeeding five years (table4b). Ourfindingis thatwhen historicaltrendsandcurrentmacroeconomic conditionsare takeninto account, privateand personalsaving have not been unusuallylow, and may even have been abnormallyhigh, over the past five years. The coefficients in the equationsgenerallyconfirmthe standardpresumptionthat temporaryincreases in disposable income increase the private saving rate, but neither revaluationsof corporate equity norrevaluationsof the housingstock appearto have a significant impact on the private or personal saving rate, though usually the coefficientshave the expected negativesign. The weakness of the effect of the stock marketon private saving makes it all the more strikingthat forecasts of saving based on the equationsin table 4a underpredictthe observed saving rate. 10 Onepossible factorworkingto reduceprivatesavingduringthe 1980s has been the cash payouts to shareownersassociated with corporate restructurings.1IIn 1985, the last year for which data are available, corporatesharerepurchasestotaled$27.3 billion, and cash paymentsto shareholdersin companies that were taken over totaled $94.8 billion. Sharerepurchasesand takeoversresultedin a flow of income equalto 4 percentof disposableincomefromthe corporateto the householdsector, comparedwith only 0.1 percent of disposable income in 1975and 1.3 percentin 1980. JohnShoven has demonstratedthat these paymentsfrom the corporatesectorto the householdsector have not supplantedbut insteadhave supplementeddividend payments."2What households have done with 10. Theequationsin table4a includequadratictrendtermsto capturewhatappearsto be a hump shape to the raw time series data on saving. While the extrapolationof a quadratictrendis somewhatperilous,similarresultsare obtainedwhen equationswith a lineartrendareestimatedstartingin 1965. 11. See John B. Shoven, "The Tax Consequencesof ShareRepurchasesand Other Non-DividendCash Paymentsto Equity Owners," in LawrenceH. Summers,ed., Tax PolicyandtheEconomy(MITPress, 1987),pp.29-54, fora discussionof sharerepurchases atndtakeoversas devices for passingcash in a tax-advantagedway fromthe householdto the corporatesector. The estimates cited below come from Poterba, "Tax Policy and CorporateSaving." 12. Shoven, "TheTax Consequencesof ShareRepurchases." -t: 00w. X Cb o F .t *t o oto N o X ^ o s ooF o.~~~7 .Qk ~~~~~~~00 rq W) *C a 00 .ss w 00 N1 sC<>\)_.t r .t O A W-@ O, O O O, ^ O t *Ct-= ~~~~~~~03 U S X 00Cr ~~~~~~~~~~~~~~~~0 o Q 00 oo X >t oo ~"t > tt < 4 N t- C o 'oI_ 0 = P,3to U) CoN o ?OC)- .4 ;,~~~~~~~~c N to I X m v c to=m e0 to l I N mm Q~~~~~~~~~~~~~~~~~~~~~~~~ e > 0 l <^~~~~~~~~~~~~~o Y ct = C1 X .c. rlU X .' ? o <4u,S m _ 0> i oJ a S *~~~~~ U b X~~~~~~~~~~~~~~~~~~~r- ? t1 X N o o o^ Q;*e SC O- "C O.',.t > e o < ~~~~~~ < eY ~ O o ~~~ Dto E o~~~~~~~~~~C,3C, o m S r~~~~~c, Y > Q 0 Lawrence Suimmers and Chris Carroll 623 Table 4b. Prediction Errors Generated by Equations Explaining Private and Personal Saving Measures, 1982-86 Percent Prediction esrr or a Equiationi 1. Privatesaving 2. Inflation-adjusted privatesaving 3. Personalsaving 4. Inflation-adjusted personalsaving 1982 1983 0.58 (0.435) -0.15 (-0.096) 1.11 (0.748) 0.89 (0.556) 0.75 (0.511) -0.16 (-0.095) 0.85 (0.518) 0.68 (0.389) 1984 1985 1.57 (0.954) 0.68 (0.362) 2.01 (1.104) 1.98 (1.011) 1.60 (0.828) 0.40 (0.182) 1.12 (0.521) 1.41 (0.612) 1986 2.39 (1.189) -1.02 (-0.446) 2.26 (1.014) 3.10 (1.293) Source: Equations in table 4a (estimated over the period 1950-81) used to forecast saving, 1982-86. a. Errors are expressed as realized minus predicted saving rates as a fraction of real private disposable income. Numbers in parentheses are t-statistics computed using the standard error of the corresponding equation in table 4a. this extra cash remains an open question. If they have reinvested it, repurchasesandtakeovershave not affectedthe overallsavingrate. But to the extentthathouseholdshave consumedit, the personalandprivate savingrate has been reduced. We suspect that consumptionout of cash payoutsfromcorporaterestructuringmay have been of substantialand growingimportancerecently. If households consumed 50 percent of those payouts in recent years, the personal and private saving rates wouldhave fallenbetween 1 and 2 percentagepoints. The Efficacy of Saving Incentives Some analysts have concluded from the low recent private and personalsavingrates that the IndividualRetirementAccount tax incentives for privatesaving enacted in 1981have been ineffective. Such an inferenceis premature.IRA contributionsin 1984,for example, representedless than2 percentof disposableincome.13It shouldbe clearfrom the size of the predictionerrors in table 4b that the fraction of these contributionsthatrepresentednew incrementalsavingcannotbe reliably inferredfromaggregatedata. If anything,the tendencyfor the equations to underpredictprivate and personal saving suggests the efficacy of IRAs. 13. ChrisCarrolland LawrenceH. Summers,"Why Have PrivateSavingsRates in the United States and CanadaDiverged?" Journal of Monetary Economics, vol. 20 (September1987),pp. 249-79. See table3. 624 Brookings Papers on Economic Activity, 2:1987 Severalpieces of microeconomicevidence suggestthata sizable part of IRA contributionsdoes representincrementalsaving.First, most IRA contributorshave relativelylittle wealth or capitalincome. The Federal Reserve Board's Survey of ConsumerFinances revealed that in 1983, the median-income IRA contributorhad less than $10,000 in liquid assets. A two-earnerfamily makingthe maximumcontributionwould have exhaustedthis sumin fewer thanthreeyears. Second, for the more than60percentof IRAcontributorswhocontributeless thanthe statutory maximumamount,IRAsclearlyprovideanincrementalsavingincentive. Third, a sizable fraction of IRA contributorsmake their contribution near the last possible moment, suggestingthat they are respondingto the advertisingblitz mountedby financialinstitutionseach April. IRS statistics indicate that almost half of contributionsfor 1984were made in 1985.14 Whilethe macroeconomicdatado not permitanyjudgmentaboutthe efficacyof targetedsavingincentives, they do runcounterto theoretical andempiricalargumentssuggestingthatprivatesavingrespondsstrongly to ratesof return.15 If savingwere highlyinterest-elastic,one wouldhave expected the unprecedentedlyhighrealinterestratesof the 1980sto lead to largepositive residualsin savingequations.One explanationfor why they did not is that high real interest rates were caused in part by the strengthof consumptiondemand.If so, saving rates and interest rates would move in opposite directions,even if savingwere interest-elastic. Another way to reconcile the observed data with theoretical arguments suggestingthat saving should be responsive to rates of returnis to note that measured real interest rates probably do not accurately reflect the expected returnson most of the assets in consumers'portfo14. See Steven F. Venti and David A. Wise, "Have IRAs IncreasedU.S. Saving?: EvidencefromConsumerExpenditureSurveys," WorkingPaper2217(NationalBureau of Economic Research, April 1987),and the papers cited there for a discussion of the microeconomicevidence on IRAs. See also commentsby HarveyGalper,CharlesByce, and LawrenceH. Summersin TaxNotes, vol. 31 (June2, 1986),pp. 917-21, and vol. 31 (June9, 1986),pp. 1014-16,for alternativereadingsof the microeconomicevidence. 15. Such argumentsare presentedin MichaelJ. Boskin, "Taxation,Saving, and the Rate of Interest,"Journalof Political Economy, vol. 86 (April 1978),pp. S3-27; and in Lawrence H. Summers, "CapitalTaxationand Accumulationin a Life Cycle Growth Model,"AmericanEconomicReview,vol.71 (September1981),pp.533-44;andSummers, "The After-TaxRate of ReturnAffects PrivateSavings," AmericanEconomicReview, vol. 74 (May 1984, Papers and Proceedings, 1983), pp. 249-53. Lawrence Summers and Chris Carroll 625 lios. Dramaticincreasesin price-earningandprice-dividendratiosduring the 1980smightwell have been taken as indicativeof reducedexpected returns on corporate equities.16 While these arguments have some appeal, the experience of the 1980s certainly creates doubt about the ability of economic policy to raise private saving by increasing the returnsavailableto savers.17 The Downward Trend in Private Saving The equationsin table4a generallyfinda substantialdownwardtrend in recent years in the saving rate, with the estimates suggestingthat, other things being equal, the private saving rate is currentlytrending downwardat a rateas highas 0.4 percentper year. It is the strongtrends that enable the equations in table 4a to predict reasonably accurately recent savingbehavior. Judgmentsaboutthe likely futurecourse of privatesavingdependon one's beliefs about why it has trended downward. In this part of the paper,we brieflyexaminea numberof possibleexplanationsfordeclining saving. Since our interest is in secular ratherthan cyclical movements in saving, we do not try to fit econometric equations describingconsumptionor saving. Instead, our approachis informal. We begin by focusing on the primarymotivationsfor saving: provision for old age, the possibilityof "rainydays," the desire to purchasebig-ticketitems, andthe desireto leave bequests. Thenwe examinepossible connections betweendemographicchanges and trendsin the savingrate. The most commonly adduced explanationfor saving is the need to providefor old age. The celebratedlife-cycle saving hypothesis holds thataggregatesavingsarise because the dissavingof the retiredpopulation is exceeded by the saving of the more numerousand prosperous 16. The available empiricalevidence supports this possibility. See, for example, RobertJ. Shiller, "Stock Prices and Social Dynamics," BPEA, 2:1984, pp. 457-98. A numberof studieshave also foundthatincreasesin real interestrates portendlower, not higher,stockreturns. 17. See CarrollandSummers,"WhyHave PrivateSavingsRatesin the UnitedStates andCanadaDiverged?"for a discussionas to why a comparisonof the UnitedStatesand Canadais more encouragingabout the efficacy of saving incentives and high rates of return.These factorsappearto explainwhy Canada'ssaving rate has trendedupwards throughtimewhilethe Americansavingratehas declined. 626 Brookings Papers on Economic Activity, 2:1987 young.18Thelevel of savingwilldependon the extentto whichconsumers expect that their income will fall late in life, which in turn will depend both on retirementbehaviorand on the income supportavailableto the retiredpopulation. Table 5 presents dataon changes since 1950in the relativeeconomic well-being of elderly Americans. Despite dramaticreductions in the labor-force-participation rate of marriedmen over age sixty-five, from 37.1 percentin 1960to 17.3percentin 1986,and despite the agingof the elderlypopulation,the income of the elderlyhas increasedsubstantially relative to that of the rest of the population.'9The income of the aged can increase even as they retireearlierbecause labor income accounts for about 15 percent of their income.20Primarilyresponsible for the improvementin the well-being of the elderly has been the dramatic increase in social securitybenefits. The ratio of those benefitsper aged adultto per capitadisposableincome has grownnearly50 percentin the pasttwentyyears. Social securityhas been especiallyeffective in putting a floorunderthe income of the aged. As a consequence, the shareof the elderlypoor is now lower thanthe correspondingshareof the remainder of the population. It seems reasonableto expect that the currentrelativeincome of the elderlyinfluencesthe perceptionof youngerAmericansabouthow much they need to save for retirement. The observed change of about 10 percentagepoints in the ratioof the medianincome of the elderlyto the medianincome of the rest of the populationcould easily account for a significantpart of the decline in private saving. It is noteworthy that rising private saving rates in the 1950s coincided with declines in the relative economic position of the elderly, while the turnaroundin the relative income of the elderly preceded the downwardtrend in private savingrates thatbeganin the mid-1960s.The currentimportanceof this 18. For a summaryof the life-cycle hypothesisand supportingevidence, see Franco Modigliani, "Life Cycle, IndividualThrift, and the Wealth of Nations," American EconomicReview,vol. 76 (June1986),pp. 297-313. 19. The statisticsin the tableprobablyunderstateboththe absoluteeconomicposition of the elderlyandthe improvementin theirrelativepositionthroughtimebecausethey do not take accountof taxes or the value of the medicalservices providedundermedicare. Nor do they take accountof the fact thatthe elderlytypicallylive in smallerfamiliesthan the nonelderly. 20. Economic Report of the President, February 1985, p. 170. Lawrence Summers and Chris Carroll 627 Table 5. Relative Income of the Elderly and Nonelderly, Selected Years, 1950-85 Ratio of median incomeSa Ratio of poverty ratesb Social security payment-income ratioc Year Men Women 1950 1955 0.35 0.34 0.49 0.46 n.a. 1960 1965 0.34 0.33 0.44 1.7d 2.le 0.26 0.44 1970 1975 1980 1985 0.35 0.41 0.42 n.a. 0.43 0.55 0.67 n.a. 2.2 1.3 1.2 0.9 0.30 0.37 0.40 0.40 n.a. 0.02 0.05 0.27 Sources: Medianincomesare fromSusanGrad,"Incomesof the Aged and Nonaged, 1950-82,"Social Security Bulletin, vol. 47 (June 1984),p. 9, table 6. Povertyrates for the elderlyand nonelderlyare from U.S. Department of Commerce,Bureauof the Census,Current Population Reports, seriesP-60,no. 154,"MoneyIncomeandPoverty Statusof Familiesand Personsin the UnitedStates: 1985"(GPO, 1986),p. 22, table 16. Social securitypayments to the elderlyare imputedfrom total OASDIpaymentsby weightingtotal paymentsfor each spendingcategoryby the ratioof beneficiariesover age sixty-five.WeightsandpaymentsarefromSocial Security Bulletin, variousissues, tablesM-9, M-10,and M-13. n.a. Not available. a. Medianincomeof the elderlydividedby medianincomeof the nonelderly.Data for odd years are averagesof precedingand followingeven-yeardata. b. Povertyrate of the elderlydividedby povertyrateof the nonelderly. c. Ratioof averagesocial securitypaymentper elderlypersonto per capitadisposableincome. d. Dataare for 1959. e. Data are for 1966. effect depends, of course, upon the question of what level of social securitybenefitsthose who are savingtoday expect to receive.21 Providingfor emergencies is a second motivation for saving.22As table 6 demonstrates, the extent to which the population is insured againstthe need for large medical expenditureshas increaseddramatically since 1950.Wheredirectpatientpaymentscovered 29.9 percentof hospitalizationoutlays in 1950, they covered only 9.3 percent of these outlays in 1985.There have been even largerreductionsin the fraction of physician and nursinghome care that is not covered by insurance. 21. A largeandinconclusiveliteraturedatingfromMartinFeldstein,"SocialSecurity, InducedRetirementandAggregateCapitalAccumulation," Journal of Political Economy, vol. 82 (September-October1974),pp. 905-26, has examinedthe role of social security variablesin aggregateconsumptionfunctions.Giventhatsocial security'seffect on saving depends on perceptionsabout future benefits that are likely to respond sluggishlyto legislativechanges, it is not surprisingthat the studies of year-to-yearmovements in consumptionhave not shed muchlight. 22. Theargumentconsideredinthisparagraph is discussedinthecontextof a simulation model in Laurence J. Kotlikoff, "Health Expendituresand PrecautionarySaving," WorkingPaper2008(NationalBureauof EconomicResearch,August 1986). 628 Brookings Papers oni Economic Activity, 2:1987 Table 6. Direct Medical Expenses by Patients, Selected Years, 1950-85 Totalhealth care expenses Year Nursinighome costs Percenit Percent Percent Percent Percent Percenit of total of disof total of disof total of dismnedical posable medical posable mnedical posable payment inicomne payment iniconme payment income Hospital costs Othlercostsa Percent of total medical paymenit Percent of disposable inicome 1950 1955 1960 1965 65.5 58.1 54.9 51.6 3.4 3.3 3.6 3.8 29.9 22.3 19.8 16.8 0.6 0.5 0.5 0.5 n.a. n.a. n.a. 64.5 n.a. n.a. n.a. 0.3 n.a. n.a. n.a. 74.7 n.a. n.a. n.a. 3.1 1970 1975 1980 1985 40.5 32.5 28.7 28.4 3.7 3.3 3.3 3.7 11.4 7.9 7.8 9.3 0.4 0.4 0.4 0.6 50.3 42.7 43.6 51.4 0.3 0.4 0.5 0.6 63.9 54.2 47.3 42.5 2.9 2.5 2.4 2.6 Sources: Healthcareexpendituresare fromDanielR. Waldo,KatharineR. Levit, and HelenLazenby,"National Health Expenditures,1985,"Health Care Finanicing Review, vol. 8 (Fall 1986),pp. 16-18, tables 4-8; and from RobertM. Gibson,DanielR. Waldo,and KatharineR. Levit, "NationalHealthExpenditures,1982,"Healthl Care FinancingReviewv, vol. 5 (Fall 1983),pp. 8-11, tables 4-7. Disposableincome is from EconomicReport of the President, February 1987, p. 274, table B-25. n.a. Not available. a. Includespaymentsto physiciansand otherhealthcare costs. However, because the cost of health care has risen far faster than disposable income, the share of income that consumers devote to uninsuredhealthcare has not declined. As a consequence, it is unlikely that a reduction in the need to save for possible health outlays has contributedmuchto declines in the savingrate. It may be, however, that improved disability and life insurance coverage has reduced the extent of precautionarysaving. Since 1950, there has been a modest improvementin life insuranceprotection, as shown below.23 Ratio of life insurance perfamily to disposable income 1975 1980 1985 1950 1955 1960 1965 1970 1.12 1.33 1.65 1.85 1.95 1.87 1.89 2.14 While we have not located satisfactorydata, we thinkit likely that the combined value of private and public disability insurance has also increased. Yet anothermotivationfor saving, the purchaseof big-ticketgoods, has probably grown less importantbecause it has become easier to borrowto financehousingand durablegoods. The firsttwo columns of 23. American Council on Life Insurance, Life Insurance FactBook, 1986 (Washington, D.C.: ACLI, 1986), p. 22. Lawrence Summers and Chris Carroll 629 Table 7. Required Down Payment on House Purchases, 1976-85 Down payment of first-time buyers Down payment of repeat buyers Year Percent of sales price Percent of median income Percent of sales price Percent of median income 1976 1977 1978 1979 18.0 19.2 12.4 17.6 42.9 48.7 31.1 44.1 30.8 48.3 27.6 29.0 97.6 121.3 87.8 92.3 1980 1981 1982 1983 1984 20.5 19.4 15.1 15.7 13.2 59.3 54.8 38.1 46.3 40.7 32.7 27.1 27.3 27.8 25.6 116.6 99.5 115.0 114.2 97.1 1985 11.4 30.9 32.7 125.2 Sources: Down payment to sales price ratio is from U.S. Bureau of the Census, Statistical Abstract of the United States, 1987 (GPO, 1987), p. 716, table 1293. Median income is from Economic Report of the President, February 1987, p. 278, table B-29. Median income is reported in 1985 dollars. The personal consumption deflator (from Economic Report, p. 248, table B-3) was used to form current dollar median income. table 7 present data since 1976on the average down payment by firsttimehomeowners,expressed as a fractionboth of medianfamilyincome and of the value of the purchasedhome. The decreasingneed for large downpaymentsalmost certainlyreduces some consumers'felt need for savings. More generally, as table 8 shows, consumers have been taking on increasingamounts of debt. While some of this increase is probably matchedby increased holding of assets, some probablyhas increased consumptionand reduced saving. Installmentcredit, which rose from 12.6 percent of disposable income in 1960 to 19.4 percent in 1985, is particularlylikely to representan alternativeto saving, since consumers are unlikely to take on substantialinstallment debt, which typically carries a high interest rate, while holding liquid assets. The ratio of mortgage debt to disposable income has increased as well. Recent increases in this ratio reflect the nearly $200 billion that has been borrowedon second mortgagessince 1981. A finalmotivationfor saving is provisionfor one's children.It is not clearhow the incentivefor this formof savinghas changed.Reductions in birthrateshavedramaticallyreducedthe numberof childrenforwhom parentsmust save. It may also be that the great increase in the number andqualityof publicinstitutionsof highereducationhas reducedsaving. Brookings Papers on Economic Activity, 2:1987 630 Table 8. Types of Debt Relative to Disposable Income, Selected Periods, 1956-85 Percent of disposable income Mortgagedebt Time period Average ratio Average change in real debta 1956-60 1961-65 1966-70 1971-75 1976-80 1981-85 36.9 45.1 45.0 43.8 48.3 50.3 2.7 3.1 0.8 1.1 2.9 1.6 Installmentcredit Othercredit Average ratio Average change in real debta Average ratio Average change in real debta 11.6 13.8 15.0 15.6 16.3 16.4 0.6 1.1 0.4 0.4 0.6 1.1 5.5 5.9 5.6 4.9 4.1 3.9 0.2 0.3 0.0 0.0 0.0 0.2 Sources: FederalReserveBoard, "BalanceSheets for the U.S. Economy,"and NIPA. a. Changein realdebtdividedby disposableincome.Realdebtis computedusingthe chanigein the GNPdeflator. In addition, the widespreaduse of financialaid formulasthat penalize accumulatedsaving may also have discouragedsavingfor children.We doubt, though, that these considerations have reduced saving rates much, particularlygiven thatdatafromthe 1972ConsumerExpenditure Survey reveal that marriedcouples with childrensaved 20.5 percent of their income, comparedwith 25.3 percent for marriedcouples without children. An alternativeexplanationfor the decline in saving rates is that the changingage compositionof the populationmightinfluencethe aggregate saving rate. To examine this possibility table 9 reports adjustmentsto the saving rate, constructedby combininginformationon age-specific saving rates with informationon the share of income going to different age groups.As table 10indicates,thereis considerableuncertaintyabout the age-specific pattern of saving rates. The uncertaintyreflects large recall errorsin the availablemicroeconomicdata, as well some conceptual differences between ConsumerExpenditureSurvey estimates of saving rates, which use a residual method, and Survey of Consumer Finances saving rates, which estimate saving from increases in asset stocks. Regardlessof which saving data are used, demographicchanges do not appearto accountfor largevariationsin the savingrate, in largepart because changesin the shareof income received by differentage groups are relativelymodest. From 1968to 1984,the largestchangewas the 5.7 percentdropin the shareof incomegoingto those agedforty-fiveto fiftyfour. Moretypically, changes were on the orderof 2-3 percent. 631 Lawrence Summers and Chris Carroll Table 9. Adjustments to Personal Saving Rate for Changing Demographic Composition of Income, Selected Years, 1970-84 Percent Year Consumer Expenditure Survey Survey of Consumer Finances 1970 1972 1974 1976 -0.I -0.2 - 0.4 - 0.3 0.0 0.0 0.0 -0.1 1978 1980 1982 1984 -0.3 -0.4 -0.3 0.0 -0.1 -0.2 -0.4 - 0.4 Sources: Authors' calculations. Personal saving is a weighted sum of saving rates across demographic groups, calculated as the product of the saving rate for each age group (ages 18-24, 25-34, 35-44, 45-54, 55-64, 65+) times the share of total income accruing to that age group in that year. The numbers in the table are the rate for 1968 less the rate for each year. Consumer Expenditure Survey data are reported in Bureau of the Census, Staitistical Abstract of the Uniited States 1987. Survey of Consumer Finances data are from the Board of Governors of the Federal Reserve System. Income shares are from Michael J. Boskin, Laurence J. Kotlikoff, and Michael Knetter, "Changes in the Age Distribution of Income in the United States: 1968-1984," Working Paper 1766 (National Bureau of Economic Research, October 1985). A differentdemographicexplanationfor the decliningsaving rate is thatthe rise of two-earnerfamilieshas reducedsavingrates by reducing the variabilityof family incomes. The fractionof marriedwomen in the laborforce has risen sharply,from 30.5 percent in 1960to 54.6 percent in 1986.24It is difficultto test whether that change has led to reduced saving. One negative piece of evidence is that the 1972 Consumer ExpenditureSurvey indicates that marriedcouples with both spouses workingfull time had a saving rate of 22.1 percent, comparedwith 18.6 percentfor marriedcouples with only one spouse working. Quantifyingthe separate contributionsof all these factors to the seculardownwardtrend in private saving is impossible. Ourjudgment is thatthe improvingrelativeeconomic fortunesof the elderlyprobably is the singlemost importantcause of reduced saving. Improvementsin insurancecoverageand households' increasedabilityto take on debt to purchasedurablegoods have also been at work. We doubt that these trendsare likely to be reversedin the nearfuture, thoughultimatelythe generosity of social security may have to decline. This suggests that 24. U.S. Bureauof the Census, StatisticalAbstractof the UnitedStates, 1987(GPO, 1987),table654, col. 1, p. 383. Brookings Papers on Economic Activity, 2:1987 632 Table 10. Personal Saving Rates by Age, 1963 and 1984 Percent Age group Survey of Consutmer Finances (1963) Consumer Expenditure Survey (1984) 18-24 25-34 35-44 45-54 55-64 65 + 5 13 9 14 8 -1 - 17 9 12 9 13 3 Sources: Consumer Expenditure Survey saving rates are computed from consumption and income data published in the Bureau of the Census, Statistical Abstract of the Uniited States, 1987, table 718, p. 428. Saving is defined as total expenditures minus retirement, pension, and social security expenditures divided by income before taxes minus personal taxes. Survey of Consumer Finances saving rates are from Dorothy S. Projector, Slurvey of Chaniges in Fam7?ily Finanices (Washington, D.C.: Federal Reserve System, 1968), table 7, p. 14. even if governmentdeficitsreturnto historicallynormallevels, the U.S. nationalsavingrate will remainbetween 4 percent and 6 percent. International Capital Flows and the Low National Saving Rate We next consider the implicationsof maintaininga level of national saving that is low by both historical and internationalstandards.The economic effects of a low saving rate depend criticallyon the international response to it. In textbook models of small open economies, the level of nationalsavinghas no impacton the level of nationalinvestment. Instead,decreasedsavingis translateddollarfor dollarinto international borrowing.In a closed economy, by contrast,reducedsavingis directly translatedinto lower investment. The huge U.S. currentaccount deficit of recent years and the vast internationalcapitalmarketmakeit temptingto concludethat the openeconomy model is more appropriatefor thinkingabout the effects of changes in U.S. nationalsaving. However, considerationof the recent American experience, and of the internationalhistorical experience moregenerally,raises doubtsthatinternationalcapitalflows can substitute for domestic savingon a long-termbasis. Comparedwith the vast internationaldifferences in national saving rates, the recent U.S. capitalinflows of about 3 percent of GNP do not Lawrence Slummers and C/iris Carroll 633 appearlarge. Nor do they appearlarge comparedwith the movements in capital that would be necessary to equalize internationalrates of return. With a standardCobb-Douglasproductionfunction, having a capital share of 0.25 and a capital output ratio of 2, an increase in the capitalstock equalto morethan20 percentof GNP is necessary to drive down the rate of returnby I percentagepoint. And yet observed capital flows in the United States and the associated movements in the trade deficit have been associated with huge economic dislocations as the traded goods sector of the economy has lost competitiveness. It is doubtfulthat a trade deficit of the current size would be sustainable politically,even if it were sustainableeconomically. Thejudgmentthat large-scalecapital importis not viable as a longrunstrategyis confirmed-byinternationalexperience.Figurc, illulstrates the point, firstmadeby MartinFeldsteinandCharlesHorioka,thatthere is a near-perfectassociation between national saving and investment rates.25The consistent tendency for high-savingnations to have high investmentrates and vice versa suggeststhatit would be difficualt for the United States to maintaina high investmentrate in tandemwith a low saving rate. The reasons for the close association between saving and investment are unclear, but perhaps the answer suggested by recent U.S. experience is most plausible. Capital flows cannot take place withoutlarge changes in patternsof domestic production.The consequenteconomic dislocationscreate substantialpressureto bringsaving andinvestmentinto balance,and sooneror laterthe governmentadjusts its policies accordingly.Evidence in favor of this view comes from the consistent tendency of countries with high private saving rates to run chronicbudgetdeficitsandthe tendencyfor countrieswhereinvestment exceeds privatesavingto runbudgetsurpluses. Conclusion Even if the fiscal aberrationof the Reagan years is corrected, the United States will continue to have a saving problem. Because of a seculardownwardtrendin private saving rates, nationalsaving will be 25. MartinFeldsteinandCharles Horioka,"DomesticSavingandInternationalCapital Flows," Economic Journal, vol. 90 (June1980),pp. 314-29. For a discussionof a variety of possibleexplanationsof the close associationbetweensavingandinvestmentrates, see Brookings Papers on Economic Activity, 2:1987 634 Figure 3. National Saving and Investment Rates, 1960-83 Investment (percent of GNP) 20 Japan a Greece 18 -* Austria * Switzerland 0 * Australia 16 14 - *Norway Belgium. Finland Frances *Germany Canada. *I taly 12- Sweden - 10 _ 8- * United Kingdom *United States 6 420 , , 2 4 I 6 12 10 Saving (percent of GNP) 8 14 16 . 18 20 Source: LawrenceH. Summers,"Tax Policy and InternationalCompetitiveness,"in Jacob A. Frenkel, ed., International Aspects of Fiscal Policies (Universityof ChicagoPress, forthcoming). inadequate to finance even the levels of investment that have been observed historically. The forces causing private saving to declineimprovementsin the economic well-beingof the elderly, improvements in public and private insurancethat have reduced the need to save for rainy days," and increases in the ease with which consumers can borrow-are all basicallybenign.They are not likely to be reversedover the next few years. Publicpolicy will not and probablyshould not seek LawrenceH. Summers,"Tax Policy and InternationalCompetitiveness,"in Jacob A. Frenkel, ed., InternationalAspects of Fiscal Policies (University of Chicago Press, forthcoming). Lawrence Summers and Chris Carroll 635 to reduceeconomic securityor makeit more difficultfor young families to purchasehouses. This leaves a pressingproblemfor publicpolicy. As long as the U.S. national saving rate lags far behind that of major U.S. competitors, restoringAmericancompetitivenesswill be difficult.The experience of the 1980ssuggeststhatsavingincentivescan spurprivatesavingto some extent. But it remainsthe case that changes in the government'sfiscal postureare the most potent andreliableway to increasenationalsaving. Unless new ways of encouragingprivatesaving can be found, it may be necessaryfor the federalgovernmentto runchronicbudgetsurplusesin comingyears. Comments and Discussion Alan S. Blinder: Lawrence Summersand Chris Carroll'snicely constructedand well-illustratedMichelinguide to low saving rates concentrates on the downwardsecular trend; it asks why saving rates in the 1980swere lower than in previous decades. Like the Michelinguides, thereis a lot of valuableinsightandinformationpackedinto a few pages. While I have no serious quarrelwith their answers, I want to suggest that a more pointed question mightbe: why have nationalsaving rates been so very low very recently? For something like half the problem, neither the question nor the answer is in dispute. One seemingly obvious cause of a lower national saving rate is the large dissaving of the federal government,beginning around 1983. It is tempting to take this part as "explained" by an exogenousfiscalaberrationandconcentrateon what'sleft-a temptation to which I, like Summersand Carroll,will shortly succumb. But this is a bit facile for two reasons. First, there is the possibility that debt and taxes are equivalent.The authorsconsider this hypothesis and reject it for good reasons. While I agreewith theirconclusion,I thinkthey could have offereda moredirect test by, for example, puttingthe governmentdeficit into their table 2a regressions. A working paper by Summers several years ago did this using data through1982and estimatedthat each dollarof federaldeficit raisedprivate savingby about 20-40 cents.1 I suspect that data through 1986would reduce this estimate sharply.Why not tell us? Second, both privateand governmentsaving are highly cyclical and the cyclical conditionsof 1981-86were atypical.We reallywantto know 1. LawrenceH. Summers,"Issues in NationalSavingsPolicy," WorkingPaper1710 (NationalBureauof EconomicResearch,September1985),table4. 636 Lawrence Summers and Chris Carroll 637 whether the cyclically adjustedgovernmentdeficit called forth higher cyclically adjustedprivate saving. In 1985, Angus Deaton and I asked this questionusingolderdataanda periodendingin 1984.We foundthat there was no decline at all in cyclically adjusted private saving rates through1984.2I would like to know if this still holds in the revised data and whether the years 1985-86 look differentfrom 1982-84; I suspect both are true. Again, Summersand Carrollcould easily use their table 2a regressionsto tell us. Summers and Carroll claim that the recent observations are not aberrant,and do so in an apparentlyreasonableway. They run regressions over 1950-81(see table4a), extrapolatethemto 1982-86(table4b), and ask whetherthey overpredictprivate saving rates. Not all do. So Summers and Carroll conclude that recent behavior is in line with historicalexperience;savingmay even have been surprisinglyhigh. I read the same evidence differently because I am inclined to be skeptical about extrapolatingquadratictime trends. If regression 1 is taken as representative,the effect of raising t from 1981to 1986 is to reduce the predicted saving rate by 2.47 percentage points-which exceeds the observeddrop.That,as SummersandCarrollacknowledge, is how their equations are able to track recent experience. Without a quadratictrend, the equationswould greatlyoverpredictrecent saving rates.3 If the last two years really were unusual,what mightthe reason be? I haveone explanationthatdoes not even appearin SummersandCarroll's guidebook:the fallingprice of energy. Deaton and I found in our paper thatthe relativepriceof nondurablegoods-which is, in turn,dominated by the relativepriceof energy-has a strongnegativeeffect on spending. I have since learned that energy prices have an astoundingly strong contemporaneouseffect on real spendingon energy products. Furthermore,muchof the changein real spendingon energy seems to come out of saving,ratherthanout of spendingon othergoods.4So fallingenergy 2. Alan S. Blinder and Angus Deaton, "The Time Series ConsumptionFunction Revisited,"BPEA,2:1985,p. 470, table 1. 3. It also seemsoddto omittheinterestratefroma regressionthatincludesthe inflation rate. If the interestratehadbeen entered,and earneda positive coefficient,the overpredictionswouldhavebeen greater. 4. JasonBenderlycalled this to my attention.See his "Consumptionand Housing," in The ConferenceBoard, U.S. EconomicOutlook1987-88, ResearchBulletin211 (May 1987). 638 Brookings Papers on Economic Activity, 2:1987 prices are a possible explanationof falling saving rates in 1985-86. I'd give it at least one Michelin star. In 1974-75, it should be noted, when energyprices were risingrapidly,savingrates were surprisinglyhigh. Let me now turn to Summers and Carroll'smajorfocus: the longterm decline in saving rates-if, indeed, there is one. Summers and Carrollguideus intelligentlythrougha list of possible causes of declining savingby goingback to basics. They awardthe coveted starto very few. Stunningly,theirlist of basics never even mentionsintertemporalchoice and rates of return, both of which had featuredprominentlyin earlier Summersguides. I say this not in criticism, by the way. Flexibility of mindis admirablewhen it comes from samplingthe data. In fact, I findlittle to quarrelwith in the moderate1987Summersview of this issue. He is right,for example, that the fact that the medianIRA contributorin 1983 had less than $10,000 in liquid assets raises the possibility that IRAs mighthave become a marginalincentivefor many people after a few more years. But I don't thinkthe fact that most IRA contributionsare made at the last minute is germane. In fact, it may arguethe other way: it paints a pictureof lots of people with loose cash that they can toss into tax-shelteredaccounts at the last minute. But thereis a farmorebasic point. If IRAs serve as a savingincentive, they must do so by raisingthe after-taxrate of return.Yet, as Summers and Carrollcorrectlynote, titanicincreases in rates of returnduringthe 1980sfailed to raise private saving. This suggests that the response of saving to the rate of returnmay not even be positive, much less large. And if that is the case, providinga marginalincentivewill do little good. Summersand Carroll'slist of causes of the decliningsavingrategives two starsto the increasedrelativeaffluenceof the elderly,whichsuggests a reduced need to save for old age. The argumenthere is eminently reasonable. But does it explain the facts? That depends on what facts we want to explain. It certainlycannot explain an abruptdropin saving duringthe last year or two. It is a more promisingexplanationfor any secular decline in saving that may exist. However, their table 5 shows that the biggest jump in social security benefits relative to disposable income came in the late 1950s,just before saving rates soared in the 1960s. Of course, there are lags. So I am not suggestingthat we reject the explanation-only that we downgradeit to one star. LaawrenceSummers and Chris Carroll 639 GeneralDiscussion Thomas Juster agreed with Alan Blinder about the importance of distinguishingbetween hypotheses aboutthe seculartrendof the saving rate and those about the trend in recent years. He noted that previous revisions of the NIPA raised estimates of saving; the apparentlylow savingrate duringthe past few years may be higheronce the NIPA are revised. Evidence of a seculardecline would be a more serious matter. However, Justerquestionedthe appropriatenessof Summersand Carroll's inflationadjustmentto the personal saving rate, which appeared to be a majorcontributorto the seculardecline. The nonadjustedseries shows a declineinjust the pastfew years. Whileagreeingthatan inflation adjustmentmay be appropriatefor the corporatesector, Justerargued thatit is by no means clearthat householdsbase decisions on real rather thannominalinterestrates. A numberof participantsdiscussed the logic of the authors'pension adjustment.Justernoted that theirapproach,regardingcontributionsto defined-benefitpension plans as corporateratherthanpersonal saving, assumes that households do not take into account their claim to future pension benefits in making their saving decisions. The size of this adjustmenthas grown over recent decades, thereby contributingto the apparentdecline in household saving. While agreeingthat households do not know the precise value of theiraccumulatedpension benefits, he arguedthat they do know whether the benefits are vested and for how manyyears they have credit, and they make pretty good guesses about thefractionof theirincometo whichtheywillbe entitleduponretirement. Hence he believed it plausiblethat accumulatedpension benefits affect householdbehavior. WilliamBrainardagreed about the desirabilityof an adjustmentto recognize the futureobligationsincurredby corporations and the balancing claims of households on future benefits. He arguedthatsuch claimsare in effect annuitiesowned by householdsand noted that takinginto account pension obligationsincurredin a given yearby corporationssubstantiallychanges the pictureof corporateand householdsaving. James Poterba, while agreeing that the calculations made by the authors(similarto ones made in his own paper)leave out the accrualof 640 Brookings Paper-s on Economic Activity, 2:1987 pension benefits by households, noted that constructingan aggregate series reflectingthe accrual would requirea large numberof arbitrary assumptions. Summersstressed that this issue does not affect conclusions about total private saving, but only the division of private saving between corporateand personal saving. Summersalso arguedthat the logic of Brainardand others would seem to imply the accrualof social security benefits in income and saving, a procedurehe believed most would agree is inappropriate. Olivier Blanchardwas not convinced by the authors' dismissal of RobertBarro'sview that the low savingrate may be due to the increase in the marketvalue of assets. He arguedthat Barro'sview is consistent with standardconsumptionfunctions,such as the one in the MPSmodel, according to which consumptiondepends on both labor income and wealth. An increase in expected dividendsis capturedin an increase in wealth, which should thereforebe associated with an increase in consumptionand a decrease in measuredsaving. Blanchardsuggestedthat it would be informativeto know how well such consumptionfunctions had performed in recent years and whether the income and wealth coefficientsappearedto have changed. RobertHall disagreedwith Blanchard's, andthe "standard," view of the effects of stock marketappreciationon consumption.Consumption functionslike thatin the MPSmodeltypicallyassumethatthe coefficient on wealth is constant. However, some changes in wealth, for example those caused by a change in the discount rate, would not result in an increase in the consumption of a long-lived household. If the stock market appreciates, but the real return is proportionatelylower, the coefficient on wealth times the value of wealth remainsunchanged.In other words, Hall argued,the expected flow of dividendsneed not have changed even though the value of stocks has risen. With a long time horizon, consumptionand dividendswill be approximatelyequal. A numberof participantsquestioned various of the authors' explanations for the decline in the personal saving rate. James Duesenberry was skepticalthatthe increasein the affluenceof the elderlywouldresult in a lower savingrate. He noted thatthe increasein life expectancy may offset the higherincomesof the elderly.Althoughsocial securitybenefits are like an annuity,Duesenberryreasoned, to the extent thatthe elderly rely on other income sources, the need to save for a longerlifetimemay outweighthe increasedsocial securitypayments. Lawrence Summers and Chris Carroll 641 Poterbanoted that changes in social security benefits are likely to have differenteffects upon the saving of various age groups. He questionedthe authors'conclusionthatthe currentrelativewell-beingof the elderlyis likely to reduce saving on the partof youngerindividuals.He believed that recent changes in social security benefits may encourage individualswho are nearretirementto reducetheirsaving. In supportof this view Henry Aaron cited public opinionpolls that show that young people anticipatelittle or no returnfromthe social security system. He agreedthat lower saving rates for youngerhouseholds are not likely to be the resultof recentimprovementsin social securitybenefits.Summers discounted the validity of survey responses that suggest a lack of confidencein the social security system. In his view, individualswho observe that their parents enjoy a good retirementincome from social securitywill be less inclinedto save for theirown retirement. Juster,while suggestingthat the evidence on saving rates across age groupsis not conclusive, notedthatthe datasuggestthatthe elderlymay save more than any other age group. He believes that more careful analysisof the characteristicsof savingat the microeconomiclevel, both across age groupsand across time, will be necessary for understanding the reasonsfor changesin the overall savingrate. Aaronarguedthat the relevantcomparisonfor explainingthe saving of the elderly is that between their currentand prior economic status ratherthanthat between the economic status of the aged and nonaged. If the economic status of the elderly has improved, Aaron reasoned, theirconsumptionwould be high. He cited a study by John Shoven and MichaelHurdthatindicatesthatthe aged, on average,areableto sustain their preretirementstandardof living. Aaron wondered whether this findingrepresentsa changefromearlierperiods.1 WilliamNordhaussuggestedthatthe focus uponthe life-cycle model, with the emphasisupon the representativeindividualor family, has led to the neglect of the role of distributionin the explanationof aggregate saving.Whileagreeingthat the life-cycle paradigmwas useful, with its emphasisupon individualmaximizingbehavior, Nordhaus noted that workby Kotlikoff,Spivak, and Summersappearedto demonstratethat 1. MichaelHurdand John B. Shoven, "Real Income and Wealthof the Elderly," American Economic Review, vol. 72 (May 1982, Papers and Pr-oceedings, 1981), pp. 314-18. 642 Brookings Papers on Economic Activity, 2:1987 the life-cycle model could explain only a small fraction of aggregate saving.2He remindedthe panel of Keynes's view, as expressed in The General Theory, that the distribution of income and wealth is an importantdeterminantof aggregatesaving and suggestedthat it would be useful to examine the effect of recent changes in the distributionof income upon saving. Nordhaus observed, however, that accordingto earlier theories, the increased dispersion in incomes should have increased saving. GlennHubbardaddedthat several other changes in the environment facing households have helped reduce saving. He cited the evidence providedby the authorsthatdown paymentshave declinedover time as one exampleof a changeincapitalmarketsthatmayhavebeen important. A second change is the role of governmentprogramsthat contain an insurance component and thereby reduce the need for precautionary saving. 2. LaurenceJ. Kotlikoff,Avia Spivak,and LawrenceH. Summers,"The Adequacy of Savings," American Economic Review, vol. 72 (December 1982), pp. 1056-69.
© Copyright 2024 Paperzz