99 technology, employment, and the business cycle 1999

8/2/2019 99 Technology, Employment, And the Business Cycle 1999

1/23


2/23


3/23

251VOL. 89 NO. 1 GALI : TECHNOLOGY, EMPLOYMENT, AND THE BUSINESS CYCLE

mirrors the one obtained for the UnitedStates, with the main results holding forevery G7 country but Japan.

In Section IV, I examine the implications of

the estimated decomposition regarding the roleplayed by technology shocks as sources ofpostwar business cycles. Section V explorespossible ways of reconciling that evidencewith the RBC paradigm. Finally, Section VIsummarizes the main results of the paper andconcludes.

I. Labor-Market Dynamics in a Sticky

Price Model

In this section I develop and analyzea monetary model with monopolistic com-petition, sticky prices and variable labor ef-fort.4 I assume two exogenous drivingforces: technology and monetary shocks.The focus of the analysis is on the joint re-sponse of productivity and hours to eachof those disturbances. The model is deli-berately stylized, in order to convey the ba-sic point in the simplest possible way (inother words, it is not meant to provide acomplete account of the mechanisms under-

lying business cycles). Thus, capital accu-mulation is ignored, and nominal pricerigidities are introduced by having firmsset their prices before shocks are realized.The assumptions on functional forms andstatistical properties of the shocks make itpossible to derive an exact closed-form rep-resentation of the equilibrium processes forthe variables of interest in terms of the ex-ogenous driving forces.

4 Recent examples of dynamic general equilibriummodels of the business cycle with nominal rigidities in-clude Jean-Olivier Hairault and Franck Portier (1993) ,Jean-Pascal Benassy ( 1995) , Jang-Ok Cho and ThomasCooley (1995), Jinill Kim (1996), Robert G. King andAlexander L. Wolman (1996), and Rotemberg (1996).Examples of business-cycle models with variable effortand/or utilization can be found in Craig Burnside et al.(1993) and Mark Bils and Cho (1994). Robert J. Gordon( 1990) , Argia M. Sbordone (1995) , Susanto Basu

(1996), and Matthew D. Shapiro (1996), among others,discuss the implications of that phenomenon for the cy-clical behavior of productivity measures.

A. Households

The representative household seeks tomaximize

( 1 )

MttE b log C / l log 0 H(N, U)0 m t t Ptt0subject to the budget constraint

1

P C di / Mit it t 0

W N / V U / M / Y / Pt t t t t 01 t t

for t 0, 1, 2, ... . Ct is a composite consump-tion index defined by

1 /0 10 1/C (C ) dit it

0

where Cit is the quantity of good i [0, 1]consumed in period t, and 1 is the elas-ticity of substitution among consumptiongoods. The price of good i is given by Pit, and

1 1/ 10 10 P (P ) dit it

0

is the aggregate price index. Mdenotes ( nom-inal) money holdings. Function H measuresthe disutility from work, which depends onhours (N) and effort ( U). The following func-tional form is assumed

l ln u1/ s 1/ sn uH(N , U) N / U .t t t t 1 / s 1 / sn u

Y and P denote, respectively, monetarytransfers and profits. W and V denote the( nominal) prices of an hour of work and a unitof effort, respectively. b (0, 1) is the dis-count factor. lm , ln , lu , sn , su , are positiveconstants.

The first-order conditions associated withthe household problem are

0PitC C( 2 ) it t Pt


4/23

252 THE AMERICAN ECONOMIC REVIEW MARCH 1999

1 P 1 Pt t l / b E( 3 ) m t C M C Pt t t/ 1 t/ 1

Wt sn l C N( 4 ) n t tPt

Vt su l C U .( 5 ) u t tPt

B. Firms

There is a continuum of firms distributed

uniformly on the unit interval. Each firm isindexed by i [0, 1], and produces a differ-entiated good with a technology

aY Z L .it t it

Li may be interpreted as the quantity of ef-fective labor input used by the firm, which isa function of hours and effort:

u 10 uL N Uit it it

where u (0, 1).5 Z is an aggregate technol-ogy index, whose growth rate is assumed tofollow an independently and identically dis-tributed (i.i.d.) process {ht}, with ht N(0,

Formally,2s ).z

Z Z exp(h ) .t t0 1 t

At the end of period t 0 1 (i.e., before pe-riod ts realization of the money supply andtechnology is observed) firm i sets the price

Pit at which it will be selling good i duringperiod t, taking as given the aggregate pricelevel Pt. Once the shocks are realized, eachfirm chooses Nit and Uit optimally, given Wtand Vt. Given an output level Yit, cost mini-mization requires

U 1 0 u Wit t .( 6 ) N u Vit t

5 Notice that we can write Lit Nit(Uit/Nit)10 u , which

implies that effective labor input is proportional to hours(as in the standard model) whenever effort per hour isconstant.

Furthermore, as long as the marginal cost isbelow the ( predetermined) price Pit, each firmwill find it optimal to accommodate anychanges in the demand for its product, and will

thus choose an output level

0PitY C.( 7 ) it t

Pt

Hence, when setting the price the firm willseek to maximize

max E { ( 1 /C) ( P Y 0 W N 0 V U ) }t0 1 t it it t it t it Pit

subject to (6 ) and (7 ). The correspondingfirst-order condition is given by

E { (1 /C) ( au P Y 0 mW N ) } 0( 8 ) t0 1 t it it t it

where m / 0 1.6

C. Monetary Policy

The quantity of money Ms in the economyis assumed to evolve according to

s sM M exp(j / gh )( 9 ) t t0 1 t t

where {jt} is a white noise process orthogonalto {ht} at all leads and lags, with jt N( 0,

Notice that whenever gx 0, the monetary2s ) .mauthority is assumed to respond in a systematicfashion to technology shocks.

D. Equilibrium

In a symmetric equilibrium all firms will set

the same price Pt and choose identical output,hours, and effort levels Yt, Nt, Ut. Goods mar-ket clearing requires Ct Cit Yit Yt, forall i [0, 1], and all t. Equilibrium in themoney market implies Mt/Mt0 1 exp(jt /ght), for all t. Using both market-clearing con-ditions, one can rewrite (3) (after some alge-braic manipulation) as

6 Notice that in the absence of uncertainty (8) simpli-

fies to Pit

m(WtNit/auYit), which is just the familiar op-timal price condition for a monopolist facing an isoelasticdemand schedule.


5/23


MtC F(10) t

Pt

where F 0 b /01 1 2 2 2 7l [1 exp { / (s g s ) } ] .m 2 m zFurthermore, (4 ), (5 ), and (6 ) imply Ut

where A [ln(1 01/a( 10 u) ( 1/s ) / ( 1/s )n uA N ,t

That result allows us toa(1 0 u)/(1/s )uu) /l u] .uwrite the following reduced-form equilibrium re-lationship between output and employment:

wY AZ N(11) t t t

where w au / a( 1 0 u) ( 1 / sn ) / ( 1 / su ) .Finally, evaluating (4 ) and ( 8 ) at the sym-

metric equilibrium and combining them

with (11) and (10) one can derive a set ofexpressions for the equilibrium levelsof prices, output, employment, and pro-ductivity, in terms of the exogenous driv-ing variables. Letting lower-case letters de-note the natural logarithm of each variable,and dropping uninteresting constants, wehave:

Dp j 0 (1 0 g )h(12) t t0 1 t0 1

Dy Dj / gh / ( 1 0 g) h(13) t t t t 0 1

1 1 0 gn j 0 h(14) t t t

w w

1Dx 1 0 Dj(15) t t w

1 0 g/ /

g ht w1

/ ( 1 0 g) 1 0 ht0 1 wwhere x y 0 n is the log of (measured) laborproductivity.

7 We assume b / 1. The constant1 2 2 2exp{ / ( s g s ) }2 m z

velocity associated with (10) is a consequence of our as-sumption of i.i.d. money growth rates, which in turn im-plies a constant nominal rate.

The equilibrium responses of p , y , n , and xto each shock, represented by (12 ) ( 15) , arediscussed next. A monetary shock has a tran-sitory impact on output, employment, and pro-

ductivity, and a permanent effect on the pricelevel. More specifically, and in response to anunanticipated monetary expansion ( jt 0 ) ,output and employment go up, reverting backto their original level after one period. Thesign of the (also transitory) response of laborproductivity x depends on the size ofw, and ispositive whenever w 1. As made clear by(11), the latter condition corresponds to thenotion of short-run increasing returns to la-bor emphasized in the literature on the cy-

clical behavior of productivity (e.g., Gordon,1990). For that condition to be satisfied werequire: (a ) sufficiently productive effort(low u ), (b) a sufficiently low elasticity ofefforts marginal disutility relative to that ofemployment ( su sn ), and (c) a sufficientlyhigh elasticity of output with respect to effec-tive labor input (high a ). Finally, note that theonly variable that is permanently affected bythe exogenous increase in the money supplywill be the price level, which will adjust pro-portionally (though with a one-period lag).

A (positive) technology shock( ht 0) hasa permanent, one-for-one effect on output andproductivity, as can be seen in (13) and (15).The same shock will have a permanent nega-tive effect on the price level as long as g 1,i.e., if the degree of monetary accommodationis not too strong. Most interestingly, if thesame condition is satisfied, a positive technol-ogy shock will have a negative short-run effecton the level of employment. The intuition forthat result is straightforward. Consider, for the

sake of exposition, the g 0 case ( exogenousmoney). In that case, the combination of aconstant money supply and predeterminedprices implies that real balances ( and, thus, ag-gregate demand) remain unchanged in the pe-riod when the technology shock occurs. Eachfirm will thus meet its demand by producingan unchanged level of output. If the technol-ogy shock is positive, producing the same out-put will require less labor input, and a declinein hours will be observed. Clearly, the sign ofthat short-run response of hours to a technol-

ogy shock stands in stark contrast with the pre-dictions of the basic RBC model. Furthermore,


6/23


unchanged output and lower hours will lead toan unambiguous increase in measured laborproductivity in response to the same shock. Inthe following period, firms adjust their prices

downward (since marginal cost is lower), ag-gregate demand and output will go up, and em-ployment returns to its original level. The signof the associated change in labor productivitydepends again on whether w is greater or lessthan one (i.e., on whether the change in outputis more or less than proportional to the changein hours), which, in turn, determines whetherthe immediate response of productivity to atechnology shock overshoots or not its long-run level. By looking at ( 12) ( 15) it should

be clear that the qualitative effects of a tech-nology shock described above will remain un-changed so long as g [0, 1), a parameterrange which includes both exogenous mone-tary policy as well as a monetary rule aimedat smoothing price and employment changes.8

It is important to stress that the possibilityof a decline in hours in response to a positivetechnology shock does not hinge on the as-sumptions of predetermined prices or absenceof capital accumulation, both made here forexpository convenience. Thus, Rotemberg

(1996) obtains a similar response in a modelwith quadratic costs of price adjustment, andfor sufficiently high values of the parameterindexing the magnitude of those costs. A sim-ilar response is found in King and Wolman(1996) in a similar model with capital accu-mulation and a price-setting structure origi-nally found in Guillermo Calvo (1983).Finally, King and Watson (1996) also reporta negative contemporaneous correlation be-tween multifactor productivity and hours in

their calibrated sticky price model with capitalaccumulation.

The unconditional covariances among thegrowth rates of output, labor productivity, andemployment implied by the above model areeasily computed using ( 12) ( 15) :

8 More generally, the choice of the policy rule will onlyhave a permanent effect on prices, but it will affect thesize and/ or the dynamic pattern of the responses of output,

employment, and productivity. In particular, the monetaryauthority will face a trade-off between employment andprice volatility.

cov(Dy , Dn )(16) t t

2 22s / ( 1 0 g) ( 1 0 2g) sm z

w

cov(Dy , Dx )(17) t t

2 22 ( w 0 1 ) s / ( g / w 0 1 ) sm z

w

cov(Dn , Dx )(18) t t

22 ( w 0 1 ) s m

2w

2(1 0 g) [(2 0 w) / 2g(w 0 1)]sz0 .

2w

Whenever g [ 0, 1/2) and/ or exogenousmonetary shocks are a sufficiently important(relative to technology), the model predictsthat hours growth should be procyclicalaproperty which is a robust feature of the data.Furthermore, w 1 is a sufficient conditionfor measured labor productivity to be procy-clicalanother strong feature of the dataindependently of the relative importance ofthe two shocks.

The sign of the comovement between hoursand productivity growththe focus of our at-tentiondepends on the size of w, the policyparameter g, and the relative importance ofshocks. It is useful to look first at the sign ofthe conditional covariances. Letting cov( Dnt,Dxtz ) denote the covariance between Dnt and

Dxt conditional on technology being the onlysource of fluctuations , we have:

cov( Dn , Dx z)t t

(1 0 g )2

0 [ ( 2 0 w) / 2g(w 0 1 ) ] s .z2w

Under the assumptions g [0, 1 ) and w (1 , 2 ) i t is easy to ch eck th at co v ( Dnt ,Dxtz ) 0, i.e., technology shocks generatea negative comovement between hours andproductivity growth. On the other hand, the


7/23


analogous covariance conditional on mone-tary shocks being the only source of fluctu-ations , denoted by cov ( Dnt, Dxtm ) , isgiven by

2 (w 0 1 )2cov(Dn , Dx m ) st t m2w

w ho se s i gn d ep en ds t he s iz e o f w. Ifw 1, monetary shocks will generate apositive comovement between the samevariables.

The case of most interestand a plausibleone, in my viewcorresponds to w (1, 2),and g [0, 1), i.e., it combines some short-

run increasing returns to labor with a not-too-strong endogenous money response. Inthat case the models predictions regarding thesigns of the unconditional comovementsamong output, hours, and productivity areconsistent with the evidence, and potentiallyclose to those predicted by standard RBCmodels. Yet the two models have very differ-ent implications regarding the conditionalcomovements between hours and productivitygrowth. In particular, if technology shocks arethe only source of fluctuations, the sticky price

model predicts a negative correlation betweenhours and productivity growth, whereas thecorresponding comovement conditioned onthe nontechnology shocks is positive. Such aresult is in stark contrast with the prediction ofstandard RBC models with multiple shockswhere, for the reasons described in the intro-duction, technology shocks are a source of apositive comovement between hours and pro-ductivity, while nontechnology shocks gener-ate a negative comovement.

Next I propose a simple empirical frameworkthat allows me to estimate the conditional cor-relations in the data, and thus assess the relativemerits of the two classes of models.

II. An Empirical Model

In order to estimate their conditional co-movements, the components of hours andproductivity variations associated with tech-nology and nontechnology shocks must bedisentangled. My approach involves the use

of a structural VAR model, identified by therestriction that only technology shocks may

have a permanent effect on the level of pro-ductivity. As argued below, that restrictionis satisfied by a broad range of models, in-cluding RBC models, and models with nom-

inal rigidities. The conditional correlationsof hours and productivity variations can thenbe computed using the estimated coefficientsof the structural moving average ( MA )representation.

A. Assumptions Underlying theIdentification Strategy

Next I discuss three assumptions which are jointly sufficient to yield the identifying re-

striction used, and which implicitly determinethe range of models that the framework belowcan embrace.

ASSUMPTION 1: Output is determined ac-c ording to a homogeneous of degre eone , strictly concave , aggregate productionfunction

Y F(K , Z L )(19) t t t t

where Y is output, K and L denote the ef-fective capital and labor-input servicesemployed (thus allowing for possible unob-servable variations in the utilization rate ofboth inputs), and Z is an exogenous tech-nology parameter following a stochastic pro-cess with a unit root (i.e., some technologyshocks have permanent effects on the levelof Z) .

ASSUMPTION 2: The capital -labor ratio( measured in efficiency units ) Kt/ZtLt followsa stationary stochastic process .

The previous assumption is not hard to jus-tify. Letting rt denote the return on physicalcapital, profit maximization (combined withAssumption 1 and other standard assump-tions) implies

KtF , 1K

Z Lt tr 0 depreciation rate.(20) t

markup


8/23


Thus, under the maintained assumption ofstationarity (or constancy) of the markup andthe depreciation rate, the capital-labor ratiowill be stationary whenever the sequence of

returns { rt} is stationary.9 The latter propertyis consistent with most dynamic models of thebusiness cycle, which display fluctuationsaround a steady state ( or balanced growthpath) that corresponds to that of the Ramsey-Cass-Koopmans model or the Solow-Swanmodel.10 Most importantly, that assumptionalso appears to be consistent with the empiricalcharacterizations of the time-series propertiesof asset returns found in the literature.11

ASSUMPTION 3: Effective labor input Lt isa homogeneous of degree one function ofhours and effort:

L g (N , U)(21) t t t

and effort per hour Ut/Nt follows a stationarystochastic process .

Homogeneity is required if effective laborinput is to be proportional to hours whenevereffort per hour is constant. Stationarity of Ut/

Nt seems empirically plausible and is certainlyconsistent with existing business-cycle modelswith variable effort ( e.g., Burnside et al.[1993], or the model of Section I).

Combining Assumptions 13 one can de-rive the following expression for measuredla-bor productivity:

Y Y L K U t t t t t X Z F , 1 g 1,t t

N L N Z L N t t t t t t

9 Clearly, the same property would hold if we were toallow for adjustment costs or any other departure from( 20) , as long as the stochastic component of that deviationwas stationary.

10 Alternatively, one could have assumed stationarity ofthe capital-output ratio, as in Shapiro and Watson ( 1988) .Combined with (19), either assumption corresponds to astationary real rate.

11 The evidence on stock returns points to small depar-tures from white noise (see, e.g., Kenneth J. Singleton,1990). The evidence on real interest rates is less strong,

but generally tends to reject the presence of a unit root(see, e.g., Frederic S. Mishkin, 1992), and the evidencereported below).

or, taking logs,

x z / z(22) t t t

where zt log F( Kt/ZtLt, 1) g(1, Ut/Nt) isstationary under the above assumptions. Equa-tion (22) holds the key to the identification oftechnology shocks, for it implies that only per-manent changes in the stochastic componentof the technology parameter z can be thesource of the unit root in productivity. Put itdifferently, under the assumptions above, onlytechnology shocks can have a permanent effecton the level of labor productivity, even thoughany other shock impinging on the economy

can affect labor productivity temporarilythrough its effects on effort per hour and thecapital-labor ratio.12

The previous condition provides the keyidentifying restriction in the structural VARmodel estimated here. Notice that such a re-striction allows both types of shocks to havepermanent effects on the levels of hours andoutput, and thus does not mislabel as tech-nology any other shock that may have suchpermanent effects.13 From that viewpoint myidentifying restriction is different from the one

originally proposed by Blanchard and DannyQuah ( 1989) , and which restricted demandshocks (in their terminology) not to have per-manent effects on the level of output. Also,notice that in contrast with Shapiro andWatson ( 1988) I do not restrict technologyshocks notto have permanent effects on hours.Though with a different motivation and objec-tives, the identification strategy adopted here

12 The investment-specific form of technologicalchange assumed in Jeremy Greenwood et al. ( 1997) is notnested in the present framework, though it is easy to checkthat the identifying restriction proposed here would alsohold in a version of their model with a unit root in theinvestment-specific technology parameter (i.e., the rela-tive price of equipment), since the capital-labor ratio (andthus labor productivity) is fully pinned down by the (sta-tionary) interest rate and the value of that technologyparameter.

13 Examples of such shocks include permanent changesin government purchases (Marianne Baxter and King,1993), permanent labor-supply shocks (Shapiro and

Watson, 1988), or even monetary shocks in an insider-outsider model of the labor market (Olivier J. Blanchardand Lawrence H. Summers, 1986).


9/23


is closer to that in Edward N. Gamber andFrederick L. Joutz (1993), who restrict labor-supply shocks not to have a permanent effecton real wages, while allowing labor- demand/

technology shocks to have such a permanenteffect.

B. Specification and ConditionalCorrelations Estimators

My empirical model interprets the observedvariations in (log) productivity (xt) and (log)hours ( nt) as originating in two types of ex-ogenous disturbancestechnology and non-technology shockswhich are orthogonal to

each other, and whose impact is propagatedover time through various unspecified mech-anisms. That idea is formalized by assumingthat the vector [ Dxt, Dnt] can be expressedas a (possibly infinite) distributed lag of bothtypes of disturbances:

Dxt(23) Dnt

11 12 zC (L) C (L) t

21 22 m

C (L) C (L) t

C(L ) t

where and denote, respectively,z m{ } { }t tthe sequ en ces o f techn olog y and n on -technology shocks. The orthogonality as-sumption ( combined with a standard nor-malization) implies I. Furthermore,E t tthe identifying restriction that the unit rootin productivity originates exclusively in

technology shocks corresponds to C12

( 1 ) 0. In other words, the matrix of long-runmultipliers C (1) is assumed to be lowertriangular.

The specification in (23) is based on theassumption that both productivity and hoursare integrated of order one, so that first-differencing of both variables is necessaryto achieve stationarity. That assumption ismotivated by the outcome of standard aug-mented Dickey Fuller ( ADF ) tests which donot reject the null of a unit root in the levels

of either series, but do reject the same nullwhen applied to the first-differences (at the

5-percent significance level) .14 Notice thatwhile my identification strategy hinges crit-ically on the presence of a unit root in pro-ductivity, it can accommodate both I( 0 )

and I( 1 ) hours. Thus, and in order to checkthe robustness of the results, I also estimatean analogous model for [ Dxt, nt] , where ntdenotes deviations of ( log ) hours from a fit-ted linear time trend.

Consistent estimates of the coefficients ofC(L ) in (23) are obtained as functions of theestimated parameters of a reduced-form VARfor [ Dxt, Dnt], following a standard proce-dure.15 Given an estimate for C(L ) (whichembeds the impulse response coefficients ) , es-

timates of conditional correlations can be ob-tained using the following formula (with thepopulation coefficients are replaced by theircorresponding estimates) :

r(Dx , Dn i )(24) t t

1i 2i C Cj j

j 0___________________ var(Dx i)var( Dn i)t t

for i z , m , where var(Dxti ) j 0and var( Dnti ) are con-

1i 2 2i 2( C ) (C )j j 0 jditional variances of hours and productivitygrowth.16

III. Evidence

This section reports and discusses the evi-dence on conditional productivity-labor inputcomovements. First, I report evidence basedon a bivariate model estimated using postwar

U.S. data. Then I show how the main quali-tative results obtained in that benchmarkmodel also hold for an augmented model thatincludes a number of monetary and financial

14 Tables with a detailed description of unit root testscan be found in Gal ( 1996a) or in the Appendix availableupon request.

15 Detailed formulas for consistent estimator of C(L )in VAR models with recursive long-run restrictions canbe found in Gal (1996b) or in the Appendix available

upon request.16 Of course, in practice the sums in (24) are truncatedat a large (but finite) lag.


10/23


variables ( in addition to productivity andlabor-input measures), as well as for most ofthe remaining G7 countries.

A. Evidence from a Bivariate Model

The bivariate model ( 23 ) is estimatedusing U.S. quarterly data covering theperiod 1948:11994:4. The baseline seriesfor labor input is the log of total employee-hours in nonagricultural establishments( hours ). The baseline series for laborproductivity was constructed by subtractingthe previous variable from the log of GDP.In addition, I also report results obtained us-

ing the log of the employed civilian laborforce ( employment) as a labor-inputmeasure, with the corresponding productiv-ity measure constructed analogously. All se-ries were drawn from Citibase.

Table 1 reports estimates of both uncondi-tional and conditional correlations between thegrowth rates of each labor-input measure andthe corresponding measure of productivity.Standard errors are reported in parentheses,and significant estimates highlighted with one( 10-percent level) or two ( 5-percent level)

asterisks.17 The first panel reports results basedon an estimated VAR model for [ Dxt, Dnt],with the second panel reporting the corre-sponding results based on the [ Dxt, nt ] specification.

Estimates of the unconditional correlationof labor input and productivity are small,slightly negative, and only significant whenhours are used. As argued in Christiano andEichenbaum ( 1992) the absence of a largepositive correlation between those variables

conflicts with a key prediction of the basicRBC model driven by technology shocks, butcan in principle be reconciled with multiple-shock versions of the same model, since non-technology shocks are predicted to generate anegative correlation that may offset the posi-

17 Standard errors for conditional correlations and im-pulse responses were computed using a Monte Carlomethod to sample from the estimated asymptotic distri-

bution of the VAR coefficients and the covariance matrixof the innovations. Reported standard errors correspond tothe standard deviation of each statistic across 500 draws.

tive comovement induced by technologyshocks. Our benchmark estimates of the con-ditional correlationsreported in the secondand third columns are, however, inconsis-

tent with that explanation: in all cases, the es-timates point to a large negative correlationbetween the technology-driven components oflabor input and productivity growth, whereasthe corresponding nontechnology componentsdisplay a positive correlation. Furthermore,and with the exception of the specification us-ing detrended employment, all the estimatesare statistically significant at conventional lev-els. Figure 1 provides a graphical counterpartto the previous evidence, by displaying scat-

terplots of the original productivity and hoursseries ( in growth rates) , as well as their tech-nology and nontechnology components recov-ered from the identified VAR.18

The previous results are, however, consis-tent with the predictions of models with im-perfect competition, sticky prices, and variableeffort, as exemplified by the stylized modeldeveloped in Section I. As shown there, theshort-term rigidity in aggregate demand re-sulting from the stickiness of the price levelleads technology shocks to generate a negative

comovement between hours and productivity,while unobserved effort variations can accountfor the positive comovement induced by de-mand shocks.

In order to understand the source of the pre-vious results it is useful to look at the estimateddynamic responses of productivity, output,and hours to each type of shock. Figure 2 dis-plays the estimated impulse responses basedon the model with first-differenced hours, to-gether with their associated two-standard error

confidence bands. In response to a positivetechnology shock of size equal to one-standarddeviation, labor productivity experiences animmediate increase of about 0.6 percent, even-tually stabilizing at a level somewhat higher.Output also experiences a permanent increase,but the initial rise appears to be more gradualthan that of productivity. The gap between the

18 The dramatic contrast between those estimates and

the predictions of standard RBC model can be by com-paring Figure 1 here to Charts 2 and 4 in Hansen andWright (1992).


11/23


TABLE 1CORRELATION ESTIMATES: BIVARIATE MODEL

Unconditional Conditional

Technology Nontechnology

Panel A: First-differenced laborHours 00.26** 00.82** 0.26**

(0.08) (0.12) (0.12)Employment 00.02 00.84** 0.64**

(0.07) (0.26) (0.13)

Panel B: Detrended laborHours 00.26** 00.81** 0.35*

(0.08) (0.11) (0.20)Employment 00.02 00.35 0.38

(0.07) (0.49) (0.56)

Notes: Table 1 reports estimates of unconditional and conditional correlations between the growth rates of productivityand labor input (hours or employment) in the United States, using quarterly data for the period 1948:1 1994:4. Standarderrors are shown in parentheses. Significance is indicated by one asterisk (10-percent level) or two asterisks (5-percentlevel). Conditional correlation estimates are computed using the procedure outlined in the text, and on the basis of anestimated bivariate VAR for productivity growth and labor-input growth (Panel A) or productivity growth and detrendedlabor input (Panel B). Data sources and definitions can be found in the text.

initial increase in labor productivity and the(smaller) increase in output is reflected in ashort-lived, though persistent ( and signifi-cant), decline in hours. The fact that the bulk

of the joint variation in employment and pro-ductivity arising from a technology shocktakes place on impact, with both variablesmoving in opposite directions, is largely re-sponsible for the negative conditional corre-lation reported above.19

Figure 2 also displays the estimated dy-namic responses to a nontechnology shock, asidentified by the empirical framework above.Such a shock is shown to have a persistentpositive effect on output, hours, and produc-

19 A decline in hours (or, alternatively, an increase inunemployment) resulting from a positive technologyshock can also be detected in other structural VARs in theliterature. Since the purpose of those exercises is generallyunrelated to the issue at stake here, the presence of sucha result often appears to go unnoticed or, at most, is brieflymentioned in the text. Some of the papers where that resultcan be found are: Blanchard ( 1989 Figure 1.b ) , Blanchardand Quah (1989 Figure 6), Gamber and Joutz (1993 Fig-ure 1 ), Blanchard et al. ( 1995 Figures C and D) , Cooleyand Mark Dwyer (1995 Figure 1), and Mario Forni and

Lucrezia Reichlin (1995 Figure 3). The latter two papersprovide a longer discussion of the finding, interpreting itas being consistent with the traditional Keynesian model.

tivity. Interestingly, while the effect on pro-ductivity vanishes over time ( by assumption) ,the shock has a sizable (and statistically sig-nificant) permanent impact on both hours and

output, thus emerging as the main source ofthe unit root detected in hours. The large pos-itive comovement of productivity and hours onimpact is the main source of the positive signin the estimated correlation conditional onnontechnology shocks reported in Table 1.

Most of the qualitative patterns in the im-pulse responses just presented are preservedwhen detrended hours (i.e., deviations of loghours from a fitted linear time trend ) areused in the estimated VAR, as displayed inFigure 3. The only significant difference liesin the absence (by construction) of a perma-nent effect of the nontechnology shock on thelevel of hours (and, consequently, on output,given the identifying restriction ) . Further-more, similar results ( not reported ) obtainwhen employment is used instead of hours asa labor-input measure.20

20

Impulse responses using employment data can befound in Gal (1996a Figure 3.b) and in the Appendixavailable upon request.


12/23


FIGURE 1. PRODUCTIVITY VS . HOURS : DATA,

TECHNOLOGY COMPONENT, AND NONTECHNOLOGYCOMPONENT

B. Evidence from a Five-Variable Model

As a robustness check I estimate a higherdimensional ( five-variable ) VAR model,which allows for four orthogonal nontechnol-ogy shocksstill identified as shocks that donot have a permanent effect on the level of

labor productivity. Even though I make noattempt to identify each of those shocks sep-

arately (which would require imposing addi-tional, possibly controversial, restrictions ) ,the estimated model provides interesting in-formation regarding the effects of technology

shocks on a larger number of variables thanwas the case for the bivariate VAR.

The specification considered uses data onmoney, interest rates, and prices, in additionto the productivity and labor-input series usedin the bivariate model. My measure of thestock of money, denoted by m , is the ( log) ofM2. The price measure (p) is the (log) of theconsumer price index (CPI). The nominal in-terest rate ( r) is the three-month Treasury Billrate. Because of limited availability of M2 data

the sample period begins at a later date ( 59:194:4).In preliminary data analysis, standard ADF

tests did not reject the null of a unit root inmoney growth ( Dm) , inflation ( Dp), and thenominal rate (r) at a 5-percent significancelevel, but did reject the same hypothesis fortheir respective first-differences, as well as forD( mt 0 pt) (the growth rate of real balances),as well as rt 0 Dpt (the real interest rate).

21

That characterization suggests estimating aVAR model for [ Dxt, Dnt, Dmt 0 Dpt, rt 0

Dpt, D 2pt].22 Using the estimated VAR, to-gether with the assumption that only technol-ogy shocks have a permanent effect on x , andthe orthogonality between technology andnontechnology shocks, one can recover esti-mates of the dynamic responses to a technol-ogy shock, as well as the components of thevariation in each time series associated withthose shocks and as a residual the sum ofthe components driven by the remaining fournontechnology shocks.

21 That characterization is consistent with the findingsof many other authors (see, e.g., Shapiro and Watson[1988] and Gal [1992]). Details of the tests can be foundin Table 1 of Gal (1996a) and in the Appendix availableupon request.

22 As a robustness check to make sure that none of thequalitative results hinged on the cointegration assumptionsimplicit in the specification of the VAR, I repeated theexercise using the estimates of a VAR in first-differences (as would be appropriate in the absence ofcointegration) , i.e., a VAR for the five-variable vector

[ Dxt, Dnt, D2

mt, Drt, D2

pt]. The results obtained werevery similar to those reported in the text, and can be found

in the Appendix available upon request.


13/23


FIGURE 2. ESTIMATED IMPULSE RESPONSES FROM A BIVARIATE MODEL : U.S. DATA, FIRST-DIFFERENCED HOURS ( POINTESTIMATES AND {2 STANDARD ERROR CONFIDENCE INTERVALS )

Table 2 displays the corresponding esti-mates of the productivity-labor input correla-tions conditional on each type of shock. As

before, I report results using both Dnt and ntin the estimated VAR. The estimates largelyconfirm the results from the bivariate model:technology shocks induce a high, statisticallysignificant negative correlation between pro-ductivity and hours ( or employment) , whereasthe ( composite) nontechnology component ofthe same variables shows a positive correlation(also significant in three out of the fourspecifications).

Figure 4 displays the responses of a numberof variables to a technology shock. The pattern

of responses of productivity, output, and em-ployment is very similar to that obtained in the

bivariate model: a positive technology shockleads to an immediate increase in productivitythat is not matched by a proportional change

in output ( the latters response building upmore slowly over time) , implying a transi-torythough persistentdecline in hours.One small difference vis-a-vis Figures 2 and 3can be detected, however: the initial negativeeffect on hours is now more than fully reversedover time, leading to a positive, though quan-titatively small long-term effect.23

23

That reversal does not occur, however, when em-ployment is used as a labor-input measure (impulse re-sponses not reported here).


14/23


FIGURE 3. ESTIMATED IMPULSE RESPONSES FROM A BIVARIATE MODEL : U.S. DATA, DETRENDED HOURS ( POINTESTIMATES AND {2 STANDARD ERROR CONFIDENCE INTERVALS )

Notice that the gradual response of outputparallels the slow buildup of real balances overtime. The response of the real rate to the im-

provement in technology is positive and per-sistent, in accordance with theory, given thehigher returns to capital accumulation associ-ated with that improvement. Most interest-ingly, the estimates point to a persistentnegative impact on inflation (as opposed to aonce-and-for-all drop in the price level ) .While the direction of the price change is re-assuring (since it is consistent with the predic-tions of a broad class of models), the dynamicpattern seems consistent with the hypothesisof sluggish adjustment of prices over time,

thus strengthening the new Keynesian in-terpretation suggested above.

C. Evidence from Other IndustrializedEconomies

This section reports estimates of productivity-employment correlations for the remaining G7countries: Canada, the United Kingdom, Ger-many, France, Italy, and Japan.24 For eachcountry I estimate a bivariate VAR model forproductivity and employment. The employ-ment measure is the (log) employed civilianlabor force, drawn from the OECD Quarterly

24 Evidence for Spain using a related approach can be

found in Gal (1996b). The intriguing results obtained inthat project were the main impulse behind the presentinvestigation.


15/23


TABLE 2CONDITIONAL CORRELATION ESTIMATES:FIVE-VARIABLE MODEL

Conditional on: Technology Nontechnology

Panel A. Growth ratesHours 00.75** 0.22**

(0.04) (0.09)Employment 00.82** 0.29**

(0.08) (0.08)

Panel B. DetrendedHours 00.65** 00.02

(0.05) (0.02)Employment 00.88** 0.26**

(0.07) (0.01)

Notes: Table 2 reports estimates of conditional corre-lations between the growth rates of productivity and

labor input (hours or employment) in the United StatesStandard errors are shown in parentheses. Significanceis indicated by one asterisk (10-percent level) or twoasterisks (5-percent level). The conditional correlationestimates are based on the partially identified estimatedfive-variable VAR described in the text. The VAR isestimated using quarterly data for the period 1959:11994:4, and includes series for productivity, hours (oremployment), real balances, real interest rates, and in-flation. Panel A displays the results for the specificationthat includes labor-input growth. The results using de-trended labor input are shown in Panel B. Data sourcesand definitions can be found in the text.

Labor Force statistics. The latter was sub-tracted from (log) GDP (drawn from theOECD Quarterly National Accounts) in orderto construct the series for (log) labor produc-tivity. All data are quarterly, and seasonallyadjusted. Sample periods vary across coun-tries, depending on data availability.25

Standard ADF unit root tests were appliedto each series used.26 With one exception, the

tests did not reject at the 5-percent significancelevel a unit root in the (log) levels of employ-ment and productivity. The exception was em-ployment in France, for which the unit rootnull was rejected. That led me to estimate a

25 The sample periods are as follows: Canada (62:194:4) , the United Kingdom ( 62:1 94:3) , Germany(70:194:4), France (70:194:4), Italy (70:194:3),and Japan (62:194:4).

26

A more detailed discussion of those tests can befound in Gal (1996a) or in the Appendix available uponrequest.

VAR for [ Dxt, nt] for France, and [ Dxt, Dnt]for the remaining countries. Identification andestimation of conditional correlations pro-ceeds as in the bivariate U.S. model.

Table 3 reports, for each country, the esti-mated unconditional and conditional correla-tions of employment and productivity growth.The unconditional correlations are very smallin absolute value (and largely insignificant),with the exception of Italy ( 00.47). The av-erage correlation is 00.11. Thus, and in ac-cordance with the estimates based on U.S.data, there is no clear evidence of the largepositive correlations between productivityand employment predicted by the basic,

technology-driven RBC model.Most interestingly, the estimated condi-tional correlations for most countries displaythe same sign pattern as in the United States.Thus, and with the exception of Japan, the es-timates point to a negative correlation betweenthe technology components of employmentand productivity, with an average value of00.56 ( 00.75 if Japan is excluded). On theother hand, the nontechnology componentsshow a positive correlation (again, with ex-ception of Japan) , which is significant in most

cases, and has an average value of 0.26 (0.43when Japan is excluded).27

Figure 5 displays, for each country, the es-timated impulse responses of employment(solid line) and productivity (dashed line)to both types of shocks. With the exceptionof Japan, those responses show many of thequalitative features detected for the UnitedStates. In particular, the estimates point to apersistent decline in employment followinga positive technology shock, as well as an

increase in productivity accompanying anexpansion driven by a nontechnology shock.Nevertheless, some differences are evidentin a number of cases. Thus, technologyshocks seem to have larger and more persis-tent effects on employment in Germany, theUnited Kingdom, and Italy. By way of con-trast, in Canada the short-run negative im-pact of a positive technology shock on

27

Notice, however, that even though the pattern ofsigns of the conditional correlations is reversed for Japanthe estimates are not statistically significant.


16/23


FIGURE 4. ESTIMATED IMPULSE RESPONSES FROM A FIVE-VARIABLE MODEL : U.S. DATA, FIRST-DIFFERENCED HOURS( POINT ESTIMATES AND {2 STANDARD ERROR CONFIDENCE INTERVALS )

employment is fully reversed by the third

quarter after the shock, and ends up havinga strong positive effect asymptotically. Howshall one interpret those differences? Giventhat none of the employment responses to thetechnology shock are statistically significantin the long run, one may be tempted to down-play the differences in point estimates.28 Al-ternatively, one may want to interpret thepersistence of those responses in some of theEuropean countries as evidence of hyster-

28 A complete set of impulse responses with confidenceintervals can be found in Gal (1996a) .

esis in labor markets, along the lines sug-

gested by Blanchard and Summers (1986).In a simple version of their model, wages areset in advance by unions/ insiders so that, inexpectation, next periods employmentequals current employment. As a result, anyshock that affects current employment willchange the level of employment perma-nently. That mechanism could also underliethe permanent effects on employment re-sulting from nontechnology shocks that areobserved in most countries, though moreconventional explanations are available in

that case, since those long-run effects mayresult from permanent shifts in the labor sup-


17/23


TABLE 3CORRELATION ESTIMATES: INTERNATIONAL EVIDENCE

Unconditional Conditional

Technology Nontechnology

Canada 00.12* 00.59* 0.57*(0.08) (0.32) (0.32)

United Kingdom 00.11 00.91** 0.45**(0.13) (0.16) (0.14)

Germany 0.08 00.55** 0.23**(0.10) (0.28) (0.09)

France 0.00 00.81** 0.66**(0.11) (0.27) (0.29)

Italy 00.47** 00.93** 0.27(0.12) (0.13) (0.30)

Japan 00.07 0.41 00.60(0.08) (0.47) (0.42)

Average 00.11 00.56 0.26Average (excluding Japan) 00.12 00.75 0.43

Notes: Table 3 reports estimates of unconditional and conditional correlations between thegrowth rates of productivity and employment for Canada (62:1 94:4), the United Kingdom(62:1 94:3), Germany (70:1 94:4), France (70:1 94:4), Italy (70:1 94:3), and Japan(62:194:4). Standard errors are shown in parentheses. Significance is indicated by oneasterisk (10-percent level) or two asterisks (5-percent level). The conditional correlationestimates are computed using the procedure outlined in the text on the basis of an estimatedbivariate VAR for productivity and employment growth (detrended employment forFrance). Data sources and exact definitions can be found in the text.

ply whether exogenous ( as in Shapiro andWatson, 1988) , or induced by permanent fis-cal policy changes (as in Baxter and King,1983).

IV. Do Technology Shocks Generate

Recognizable Business Cycles?

The bulk of the evidence in the previoussection focused on the joint comovement

conditional and unconditionalof productiv-ity and labor-input measures. In this section Iturn briefly to the corresponding comovementsbetween output and labor input.

A strong positive comovement of GDPand labor input is a central feature of busi-ness cycles in industrialized economies. Anytheory of business cycles which failed tocapture that feature would be viewed as em-pirically irrelevant and would arise little at-tention from the profession, so it is thus notsurprising that a high positive correlation of

output and hours lies among the key predic-tions of the basic RBC model driven by tech-

nology shocks. Yet, whether technologyshocks in actual economies are responsiblefor the pattern of GDP and labor-input fluc-tuations associated with business cycles re-mains an open question, and one whichshould provide a critical test of the relevanceof a research program that aims to interpretthe bulk of aggregate fluctuations as result-ing from those shocks. The empirical frame-work developed above can address that

question by allowing one to decompose thehistorical time series for GDP and hours (oremployment ) into technology and nontech-nology shocks.

The outcome of that exercise is displayed inFigure 6. In order to save space, I report onlyresults for the United States, based on the bi-variate VAR for [Dxt, Dnt]. The figures dis-play the estimated components of (log) GDP(solid line) and (log) hours (dashed line), af-ter being detrended ( ex post) using a HP filter( l 1600) in order to emphasize fluctuationsat business-cycle frequencies. In addition, thefigures highlight as vertical lines the nine


18/23


FIGURE 5. ESTIMATED IMPULSE RESPONSES OF EMPLOYMENT (SOLID LINE ) AND PRODUCTIVITY ( DASHED LINE ) FOROTHER INDUSTRIALIZED ECONOMIES

NBER-dated postwar recessions. The patternsthat emerge are quite revealing in a number of

ways.29

Consider the fluctuations that the empiricalmodel identifies as having resulted from tech-nology shocks (top chart). The patterns dis-played by the two series hardly match any of

29 Results for most other specifications and countriesare qualitatively similar. In particular, an almost identicalpicture emerges when detrended hours are used in theVAR specification (see Appendix available upon re-

quest) , which implies that the results reported here do nothinge on my allowing for permanent effects of nontech-nology shocks on both output and hours.

the postwar cyclical episodes. That feature isparticularly true in one dimension: the strong

positive comovement of GDP and employ-ment that is generally viewed as central char-acteristic of business cycles is conspicuouslyabsent here; in fact, the estimated correlationbetween the two series is 00.02.

A look at the nontechnology components ofthe GDP and hours series ( bottom chart )yields a completely different picture. First,such shocks are seen to have had a dominantrole in postwar U.S. fluctuations. Second, theestimates point to an unambiguous pattern ofpositive comovements of GDP and hours as-

sociated with those nontechnology shocks,with an estimated correlation of 0.97. Third,


19/23


FIGURE 5. Continued

the resulting fluctuations account for the bulkof the decline in GDP and hours associatedwith postwar recessions.

V. Can the Evidence be Reconciled with the

RBC Paradigm?

The above results strongly suggest that U.S.business cycles have been largely driven bydisturbances that do not have permanent ef-fects on labor productivity. To the extent thatonly technology shocks can account for theunit root in the latter variable, those resultsseem to provide a picture of U.S. business cy-cles that is in stark contrast with the one as-

sociated with RBC models. That conclusionmay be strengthened by examining ( and trying

to refute) two arguments that have often beenraised in order to reconcile the previous evi-dence with the RBC paradigm.

First, one might argue that the shocks thathave been labeled all along as nontechnol-ogy shocks might also be capturing transi-tory shocks to technology, since the latterwould generally have no permanent effect onthe level of productivity. While there is noth-ing logically wrong with that interpretation, itcan hardly provide any support for RBC mod-els. For one thing, it is hard to understand howshocks to technology could be transitory, anobservation which seems to conform with thefailure to detect a significant transitory com-

ponent in measures of total factor productivity(TFP) growth, which, to a first approximation,


20/23


FIGURE 6. ESTIMATED TECHNOLOGY AND NONTECHNOLOGY COMPONENTS OF U.S. GDP AND HOURS

can be characterized as white noise.30 Most im-portantly, such an interpretation leaves unan-swered why permanent technology shockswould have the effects on the economy thatare reflected in the estimated conditional cor-

relations and impulse responses reportedabove.31

30 This characterization seems to hold even when pos-sible variations in inputs utilization rates are accounted for(see, e.g., Burnside and Eichenbaum, 1996).

31 Nonstandard RBC models characterized by slowtechnology diffusion may generate a negative response ofemployment to a positive technology shock (see, e.g.,Hairault et al., 1995 ) because of a dominant wealth effect(that makes people be willing to consume more leisure).

That mechanism is, in my view, little plausible (in addi-tion to being in conflict with the observed time-seriesproperties of multifactor productivity) .

Second, multisectoral RBC models with id-iosyncratic technology shocks and lags in thereallocation of labor across sectors are likelyto imply a short-term decline in aggregate em-ployment in the wake of a positive technology

shock in one sector ( reflected in aggregateTFP), thus inducing a negative comovementconsistent with the estimates above. In thatcontext, however, the pattern of conditionalcorrelations signs predicted by the RBC modelshould still be present in sectoral data , an im-plication that is in principle testable. Estimatesof such correlations based on two-digit U.S.manufacturing data have recently been ob-tained by Michael T. Kiley (1997) using anidentified VAR model for employment andproductivity growth based on the one pro-

posed and estimated in the present paper.Kileys estimated correlations between the


21/23


technology-driven components of those vari-ables turn up negative for the vast majority ofindustries (15 out of 17) and quite high in ab-solute value (average 00.58). The corre-

sponding estimates for the nontechnologycomponents are mostly positive (11 out of 17industries) , with an average value of 0.20.Kileys sectoral results are thus clearly notsupportive of a multisectoral RBC expla-nation for the aggregate evidence provided inthis paper.

VI. Summary and Conclusion

In recent years, many macroeconomists

have been attracted by the hypothesis that ag-gregate fluctuations can be explained, at leastto a first approximation, as the economys re-sponse to exogenous variations in technology.That view is often justified by the (largely rec-ognized) ability of RBC models to generateunconditional moments for a number of mac-roeconomic variables that display patternssimilar to their empirical counterparts.

The present paper has provided some evi-dence that questions the empirical merits ofthat class of models.32 The paper builds on the

observation of a near-zero unconditional cor-relation between productivity and employ-ment, both in the United States and in manyother industrialized economies. Proponents ofRBC models have interpreted that evidence asreflecting the coexistence of technologyshocks with other shocks. Yet, and to the ex-tent that technology shocks are a significantsource of fluctuations in those variables, wewould expect RBC models to provide at leastan accurate description of the economys re-

sponse to such shocks. For the majority of theG7 countries, however, the estimates of the ef-fects of technology shocks yield a picturewhich is hard to reconcile with the predictionsof those models: positive technology shockslead to a decline in hours, and tend to generate

32 Basu et al. (1997 ) obtain similar results using an un-related approach: they look at the response of inputs to aninnovation in a modified Solow residual series, where

the modification attempts to correct for the bias associatedwith increasing returns, imperfect competition, variableutilization, and sectoral reallocations.

a negative comovement between that variableand productivity. On the other hand, nontech-nology shocks are shown to generate a positivecomovement between hours and productivity,

in contrast with the negative comovement pre-dicted by RBC models with multiple shocks.

The results are, however, consistent with aclass of models with imperfect competition,sticky prices, and variable effort. In thosemodels a stylized version of which has beenpresented in Section 1the combination ofprice rigidities and demand constraints leadsfirms to contract employment in the face of anexogenous increase in multifactor productiv-ity, whereas the presence of variable effort

accounts for the rise in measured labor pro-ductivity in response to a demand-induced ex-pansion. Needless to say, the nature ofaggregate fluctuations and the potential rolefor policy associated with such an economyare very different from those identified withthe RBC paradigm.

REFERENCES

Basu, Susanto. Procyclical Productivity: In-

creasing Returns or Cyclical Utilization?Quarterly Journal of Economics , August1996, 111(3 ), pp. 719 51.

Basu, Susanto; Fernald, John and Kimball, Miles.

Are Technology Improvements Contrac-tionary? Mimeo, University of Michigan,1997.

Baxter, Marianne and King, Robert G. FiscalPolicy in General Equilibrium. American Economic Review, June 1993, 83(3), pp.31534.

Benas sy, Jean- Pasc al. Money and WageContracts in an Optimizing Model of theBusiness Cycle. Journal of MonetaryEconomics , April 1995, 35 (2), pp. 30316.

Bencivenga, Valerie. An Econometric Study ofHours and Output Variation with PreferenceShocks. International Economic Review ,May 1992, 33(2 ), pp. 449 71.

Bils, Mark and Cho, Jang-Ok. Cyclical FactorUtilization. Journal of Monetary Econom-ics , April 1994, 33(2 ), pp. 319 54.

Blanchard, Olivier J. A Traditional Interpre-tation of Macroeconomic Fluctuations.


22/23


American Economic Review, December1989, 79(5 ), pp. 1146 64.

Blanchard, Olivier J. and Quah, Danny. TheDynamic Effects of Aggregate Demand and

Supply Disturbances. American EconomicReview, September 1989, 79(4 ), pp. 65473.

Blanchard, Olivier J.; Solow, Robert and Wilson,

B. A. Productivity and Unemployment.Mimeo, Massachusetts Institute of Technol-ogy, 1995.

Blanchard, Olivier J. and Summers, Lawrence H.

Hysteresis and the European Unemploy-ment Problem, in Stanley Fischer, ed., NBER macroeconomics annual 1986.

Cambridge, MA: MIT Press, 1986, pp. 1578.Burnside, Craig and Eichenbaum, Martin.

Factor-Hoarding and the Propagationof Business-Cycle Shocks. American Eco-nomic Review , December 1996, 86(5 ), pp.115474.

Burnside, Craig; Eichenbaum, Martin and Re-

belo, Sergio. Labor Hoarding and the Busi-ness Cycle. Journal of Political Economy,April 1993, 101(2 ), pp. 245 73.

Calvo, Guillermo A. Staggered Prices in a

Utility-Maximizing Framework. Journalof Monetary Economics , September 1983,12(3 ), pp. 383 98.

Cho, Jang-Ok and Cooley, Thomas F. The Busi-ness Cycle with Nominal Contracts. Eco-nomic Theory , June 1995, 6(1 ), pp. 1333.

Christiano, Lawrence J. and Eichenbaum, Martin.

Current Real-Business-Cycle Theoriesand Aggregate Labor Market Fluctuations.American Economic Review , June 1992,82(3 ), pp. 430 50.

Cooley, Thomas F. and Dwyer, Mark. BusinessCycle Analysis Without Much Theory: ALook At Structural VARs. Mimeo, Uni-versity of Rochester, 1995.

Forni, Mario and Reichlin, Lucrezia. LetsGet Real: A Dynamic Factor AnalyticalApproach to Disaggregated Business Cy-cles. Centre for Economic Policy Re-s ea rc h D is cu ss io n P ap er N o. 1 24 4,September 1995.

Jordi. How Well Does the IS-LM ModelGalB,Fit Postwar U.S. Data? Quarterly Journalof Economics , May 1992, 107(2 ), pp. 70938.

. Government Size and Macroeco-nomic Stability. European Economic Re-view, January 1994, 38(1 ), pp. 117 32.

. Technology, Employment, and the

Business Cycle: Do Technology Shocks Ex-plain Aggregate Fluctuations? NationalBureau of Economic Research (Cambridge,MA ) Working Paper No. 5721, August1996a.

. Fluctuaciones y Persistencia delEmpleo en Espana, in R. Marimon, ed.,Laeconoma Espanola: Una vision diferente .Barcelona: Antoni Bosch, 1996b, pp. 13969.

Gamber, Edward N. and Joutz, Frederick L.

The Dynamic Effects of Aggregate De-mand and Supply Disturbances: Com-ment. American Economic Review ,December 1993, 83(5 ), pp. 1387 93.

Gordon, Robert J. Are Procyclical Productiv-ity Fluctuations a Figment of MeasurementError? Mimeo, Northwestern University,1990.

Greenwood, Jeremy; Hercowitz, Zvi and Krusell,

Per. Long-Run Implications of InvestmentSpecific Technological Change. American Economic Review, June 1997, 87(3), pp.

34262.Hairault, Jean-Olivier; Langot, Francois and Por-

tier, Franck. Time to Implement andAggregate Fluctuations. Mimeo, CEPRE-MAP, Paris, January 1995.

Hairault, Jean-Olivier and Portier, Franck.

Money, New-Keynesian Macroeconomicsand the Business Cycle. European Eco-nomic Review , December 1993, 37(8 ), pp.153368.

Hansen, Gary D. and Wright, Randall. The

Labor Market in Real Business CycleTheory. Federal Reserve Bank of Min-neapolis Quarterly Review , Spring 1992,16(2), pp. 212.

Kiley, Michael T. Labor Productivity in U.S.Manufacturing: Does Sectoral Comove-ment Reflect Technology Shocks. Mimeo,Federal Reserve Board, Washington, DC,1997.

Kim, Jinill. Monetary Policy in a StochasticEquilibrium Model with Real and NominalRigidities. Mimeo, Yale University, 1996.

King, Robert G. and Watson, Mark W. Money,Prices, Interest Rates and the Business Cy-


23/23


cle. Review of Economics and Statistics ,February 1996, 78(1), pp. 3553.

King, Robert G. and Wolman, Alexander L. In-flation Targeting in a St. Louis Model of the

21st Century. National Bureau of Eco-nomic Research (Cambridge, MA) Work-ing Paper No. 5507, March 1996.

Kydland, Finn E. and Prescott, Edward C. Timeto Build and Aggregate Fluctuations.Econometrica, November 1982, 50(6 ), pp.134570.

. The Computational Experiment: AnEconometric Tool. Journal of EconomicPerspectives, Winter 1996, 10(1), pp. 6985.

Mishkin, Frederic S. Is the Fisher Effect forReal? Journal of Monetary Economics ,November 1992, 30(2 ), pp. 195 215.

Rotemberg, Julio J. Prices, Output, and Hours:An Empirical Analysis Based on a StickyPrice Model. Journal of Monetary Eco-nomics , June 1996, 37(3 ), pp. 50533.

Rotemberg, Julio J. and Woodford, Michael.

Real-Business-Cycle Models and theForecastable Movements in Output, Hours,and Consumption. American EconomicReview, March 1996, 86(1), pp. 7189.

Sbordone, Argia M. Cyclical Productivity in aModel of Labor Hoarding. Mimeo, Prince-ton University, 1996.

Shapiro, Matthew D. Macroeconomic Impli-

cations of Variations in the Workweek ofCapital. Brookings Papers on EconomicActivity, 1996, (2), pp. 79119.

Shapiro, Matthew D. and Watson, Mark W.

Sources of Business Cycle Fluctuations,in Stanley Fischer, ed., NBER macroeco-nomics annual 1988. Cambridge, MA: MITPress, 1988, pp. 11148.

Sims, Christopher A. Models and TheirUses. American Journal of AgriculturalEconomics , 1989, 71(2 ), pp. 48994.

. Macroeconomics and Methodol-ogy. Journal of Economic Perspectives,Winter 1996, 10(1 ), pp. 105 20.

Singleton, Kenneth J. Specification and Esti-mation of Intertemporal Asset Pricing Mod-els, in Benjamin M. Friedman and FrankH. Hahn, eds., Handbook of monetary eco-nomics , Vol. 1. Amsterdam: North-Holland,1990, pp. 583626.

Watson, Mark W. Measures of Fit for Cali-brated Models. Journal of Political Econ-omy, December 1993, 101(6 ), pp. 1011 41.

99 technology, employment, and the business cycle 1999

Documents