basu, foley - dynamics of output and employment in the us economy (2013)

7/29/2019 Basu, Foley - Dynamics of Output and Employment in the US Economy (2013)

1/30

Cambridge Journal o Economics 2013, 37, 10771106doi:10.1093/cje/bes088Advance Access publication 15 May 2013

The Author 2013. Published by Oxford University Press on behalf of the Cambridge Political Economy Society.

All rights reserved.

Dynamics of output and employment inthe US economy

Deepankar Basu and Duncan K. Foley*

This article investigates the changing relationship between employment and realoutput in the US economy rom 1948 to 2010 at the aggregate level and at somemajor industry grouping levels o disaggregation. Real output is conventionallymeasured as value added corrected or price ination, but there are some indus-tries in which no independent measure o value added is possible and existing sta-tistics depend on imputing value added to equal income. Indexes o output thatexclude these imputations are closely correlated with employment over the wholeperiod and remain more closely correlated during the current business cycle. Thisanalysis oers insights into deeper structural changes that have taken place in theUS economy over the past ew decades in a context marked by the ollowing threeactors: (i) the service (especially the fnancial) sector has grown in importance,(ii) the economy has become more globalised, and (iii) the policy orientation hasincreasingly become neo-liberal. We demonstrate an economically signifcant reduc-tion in the coefcient relating employment growth to output growth over the busi-ness cycles since 1985. Some o this change is due to sectoral shits towards services,

but an important part o it reects a reduction in the coefcient or the goods andmaterial value-adding sectors.

Key words: Okuns Law, Kaldor-Verdoorn eect, Global restructuring, measure-ment o real output

JEL classications: E12, E20

1. Introduction

One o the many remarkable challenges to received economic ideas posed by the fnan-

cial and economic crisis that hit the capitalist world in 2007 is the act that widely

accepted models o output-employment dynamics badly missed the mark in predict-ing the shape o the downturn and recovery in the United States and other advanced

capitalist economies. The two US business cycles preceding the 2007 crisis exhibited

Manuscript received 11 June 2011; fnal version received 9 April 2012.Address or correspondence: Deepankar Basu, Department o Economics, University o Massachusetts

Amherst, 1012 Thompson Hall, Amherst, MA 01003, email: [email protected]

*Department o Economics, University o Massachusetts; New School University, and Santa Fe Institute.We thank Laura Carvalho, Debarshi Das, Gilberto Lima, Thomas Michl, Lance Taylor, two anonymousreerees and participants at the Economics Department seminar at the New School or Social Research,the Analytical Political Economy Workshop at the University o Massachusetts, Amherst, the 2011 Annual

Meeting o the Eastern Economic Association in NYC and the Sixth Forum o the World Association orPolitical Economy at the University o Massachusetts, Amherst or useul comments and suggestions. Wealso acknowledge excellent research assistance provided by Michalis Nikioros.

atUNIVERSITATDEBARCELONA.BibliotecaonA

ugust30,2013

http://c

je.oxfordjournals.org/

Downloadedfrom
mailto:[email protected]://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/mailto:[email protected]


2/30

1078 D. Basu and D. K. Foley

jobless recoveries, in which employment rose much more slowly relative to measures

o output such as real gross domestic product (GDP) than models predicted. The

20078 crisis added another dimension to this anomaly, showing that GDP could be

disconnected rom unemployment in downturns as well as recoveries. In 2009, the

ofcial unemployment rate in the United States rose about twice what would havebeen predicted by conventional models o output-employment dynamics, given meas-

ured declines in output; another way o stating the same phenomenon is that in the

downturn o 2009, the all in GDP was ar lower than what would have been predicted

by conventional models, given the increase in the aggregate unemployment rate. The

pattern o jobless recoveries has also re-asserted itsel: the increase in real GDP since

the ofcial end o the Great Recession in the second quarter o 2009 has had even less

eect proportionately on the aggregate unemployment rate than in the previous two

jobless recoveries.

As we document in this article, the close relationship between output growth as

measured by real GDP and employment generation that characterised the US economy

over the two decades ater World War II has been weakening since the mid-1980s. This

has led both to jobless recoveries in which aggregate unemployment has decreased less

during the upturn phase o business cycles than what would have been predicted on

the basis o the past association between output growth and unemployment changes,

and also to severe job-loss downturns in which the aggregate unemployment rate has

increased by more during the downturn phase o the business cycle than past experi-

ence would have predicted. Thus, what seems to be at issue is a changing relationship

between aggregate demand as measured by real GDP and employment over the whole

business cycle.

The political economic implications o this change are ar-reaching in both a short- and

long-run perspective. As an immediate political issue, persistently high unemploymentrates, in the ace o modest real GDP growth and high profts,1had an enormous eect on

the 2010 mid-term elections in the United States.2

From a longer term labour perspective, the weakening o the relationship between

measured real GDP growth and employment poses serious questions about the social

and political viability o what we call neo-liberal globalising strategies or economic

development in the United States and other economies. These neo-liberal strategies

include liberalisation o international trade and capital ows, de-regulation o fnancial

markets and institutions, reliance on ination-targeting monetary policy, privatisation

o traditionally public sector economic unctions, weakening o job security and work-

ers rights to organise, reduction in job saety and health protection and unqualifedreliance on markets or the allocation o social resources. Understanding the changing

relationship between output, conventionally measured as real GDP, and job creation

(and unemployment) is a necessary frst step in ashioning policies that can simultane-

ously generate growth in employment and protect labour interests such as the right to

collective bargaining, adequate wages and benefts, acceptable working conditions and

adequate and secure pensions, within the ramework o democratic political institu-

tions as alternatives to neo-liberal economics.

1 Even as the 2010 unemployment rate remained stuck above 9%, US corporate profts increased at

record rates. See http://economix.blogs.nytimes.com/2010/11/12/a-high-water-mark-or-profts/.2 See http://thecaucus.blogs.nytimes.com/2010/07/19/mystery-or-white-house-where-did-the-jobs-go/?emc=eta1.


ugust30,2013

http://c


Downloadedfrom
http://economix.blogs.nytimes.com/2010/11/12/a-high-water-mark-for-profits/http://thecaucus.blogs.nytimes.com/2010/07/19/mystery-for-white-house-where-did-the-jobs-go/?emc=eta1http://thecaucus.blogs.nytimes.com/2010/07/19/mystery-for-white-house-where-did-the-jobs-go/?emc=eta1http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://thecaucus.blogs.nytimes.com/2010/07/19/mystery-for-white-house-where-did-the-jobs-go/?emc=eta1http://thecaucus.blogs.nytimes.com/2010/07/19/mystery-for-white-house-where-did-the-jobs-go/?emc=eta1http://economix.blogs.nytimes.com/2010/11/12/a-high-water-mark-for-profits/


3/30

Dynamics of output and employment 1079

What lies behind the changing relationship between real output and employment

in the United States and other advanced capitalist economies? This article inves-

tigates this question in several steps. First, we document the relationship between

the rapid growth o tertiary service industries and changes in employment-output

dynamics. We provide evidence that the discrepancy between predictions based onhistorical experience and actual changes has been growing, not only in the 20079

recession but also over the previous three business cycles. An important eature o

this discrepancy is the growing importance o industries, such as fnance, insur-

ance and real estate (FIRE) where, in the absence o a measurable industry out-

put, measures o output are imputed in the national income and product accounts

(NIPA) on the basis o incomes. A measurable value added (MVA) index o output

that excludes these industries is more closely related to changes in employment

over recent business cycles than is real GDP. Although MVA has a high overall

correlation with real GDP over much o the postWorld War II period, there is

evidence that this correlation is weaker at business cycle requencies and has been

alling over recent business cycles. Thus one explanation or jobless recoveries and

severe job-loss downturns is that the most widely used index o real output, real

GDP, is driting urther rom capturing changes in aggregate demand that lead to

changes in employment.

We also examine the evolution o the relation between employment and output in

industries where there are independent measures o value-added output and income,

using the broad ramework o the Kaldor-Verdoorn eect. We fnd that the elasticity

o employment with respect to output in these industries has been alling over recent

business cycles. Thus, in addition to the act that real GDP does an increasingly bad

job o measuring aggregate demand, there appears to be a weakening link between

aggregate demand and employment in the US economy.We evaluate one theory advanced in the literature to explain the phenomenon o

jobless recoveries, increasing exibilisation o labour markets and fnd it not well

supported by the evidence. Finally, we oer some alternative hypotheses regarding

the changing relationship between real output and employment in the US economy

as a way o understanding deeper processes o structural change taking place in the

US economy and its relation to the world economy under the neo-liberal regime o

fnancial globalisation.

We fnd that the service industries tend to have a lower responsiveness o employ-

ment to output than do non-service industries, in part because output is hard to

measure in some service industries and incomes in service industries such as FIREare weakly related to aggregate demand. As a result we would expect the economy-

wide responsiveness o employment to output to all as services grow as a raction

o GDP. But we also fnd strong evidence that there has been a signifcant decline in

the sectoral responsiveness o employment to output in goods and more generally

value-adding sectors o the economy, which points to a change in the structure o

US production. We associate this change with the restructuring o the US economy

due to globalisation o production.

The rest o the article is organised as ollows. Section 2 presents evidence relating to

structural change in the US economy over the post-war period; Section 3 investigates

the changing relationship between output and employment at disaggregated industry

levels, introducing our empirical model. Section 4 highlights the discrepancy in the

current and the previous two recessions between predicted and actual employment


ugust30,2013

http://c


Downloadedfrom
http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/


4/30


(or output) changes; Section 5 presents some regression-based evidence about the

changing employment-output linkage over the post-war business cycles. Section 6

looks at three dierent explanations or the weakening employment-output linkage;

the last section concludes the discussion. An appendix presents details o the con-

struction o the data used or the analysis and some results related to endogeneity.

2. Sectoral and aggregate growth

We begin our analysis by documenting two key eatures o the structural changes that

have taken place in the post-war US economy: changing distribution o real output and

employment across broad sectors and industries. To do so, we use the Annual Industry

Accounts o the US Bureau o Economic Analysis (BEA), which provides consistent

annual data or gross value added and employment at the industry level harmonised

across time according to the 2002 NAICS codes.3

2.1 Sectoral shares

Figure 1 can be used to understand the changing distribution o output and employ-

ment across the various sectors and industries o the US economy. The frst row

provides inormation about three broad divisions o the US economy: the private

goods-producing industries, the private services-producing industries and the private

industries involved in measurable value addition.

Private goods-producing industries are composed o the ollowing industries: agri-

culture, orestry, fshing and hunting; mining; construction; and manuacturing. Private

services-producing industries, on the other hand, comprise the ollowing industries:

utilities; wholesale trade; retail trade; transportation and warehousing; inormation(publishing, motion picture and sound recording, broadcasting, inormation and data

processing); fnance and insurance; real estate, rental and leasing; proessional, scientifc

and technical services; management o companies and enterprises; administrative and

waste management services; educational services; health care and social assistance; arts,

entertainment and recreation; accommodation and ood services; and other services,

except government.

The distinction between goods-producing and services-producing industries is

useul or certain purposes, but it is conceptually unsatisactory rom a Marxian or

classical political economy perspective that distinguishes productive and unproduc-

tive labour. Some o the service industries, such as wholesale and retail trade, realise

the fnal value o produced commodities, and it is more consistent to regard theirvalue added as part o commodity production. Service industries, such as utilities

and transportation and warehousing, transorm the use value o inputs and add value

like commodity-producing industries. Some other service sectors, such as inorma-

tion services, administrative and waste management services, and arts, entertainment,

accommodation and ood services produce a measurable output without imputations.

The classifcation o industries into goods-producing and services-producing sectors

does not distinguish between value-adding (or productive) sectors and value-realising

(or unproductive) sectors. Hence we have constructed a category o industries that

3 Details about this and other data sets used in the article are collected together in the appendix.


ugust30,2013

http://c


Downloadedfrom
http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/http://cje.oxfordjournals.org/


5/30


produces an independently measurable value added, which we call measurable value

added (MVA). This value-adding sector is composed o sectors in which a tangible

output (good or service) is sold in the market or a price and hence the value-added

fgure is measurable without imputations. The MVA category is composed o the

ollowing industries rom the Annual Industry Accounts: agriculture; mining; utili-ties; construction; manuacturing; wholesale trade; retail trade; transportation and

Fig. 1. Changing sectoral distribution o output and employment, 19482009


ugust30,2013

http://c


Downloadedfrom


6/30


warehousing; inormation services; administrative and waste management; art, enter-

tainment, accommodation and ood services.4

The frst row oFigure 1 shows the transormation o the US economy into a ser-

vice economy: the share o the goods-producing industries, both in terms o value

added and employment, has witnessed a secular decline.The second row o Figure 1 takes a look at the US economy at a more disaggre-

gated level and comes up with several interesting trends. First, the manuacturing sec-

tor has witnessed a spectacular decline in terms o both value added and employment;

while the manuacturing sector was the largest component o the US economy in the

early 1950s, both in terms o value added and employment, it has been overtaken by

key service-producing industries in both respects. By the mid-1980s the FIRE sector

had overtaken manuacturing as contributing the largest share o GDP; by the mid-

2000s, the proessional and business services (PBS) sector had similarly overtaken the

manuacturing sector. In terms o the share o total employment, the same process was

delayed by about two decades: only in the late 1990s did the PBS sector employ more

workers than manuacturing.

Second, the FIRE sector has increased its share o value added much more steadily than

its share o total employment. In act, the share o employment accounted or by the FIRE

sector has stagnated since the mid-1980s, but the share o output contributed by this sec-

tor has continued increasing right until the mid-2000s. A similar, though less pronounced,

trend can be observed in both the PBS and inormation services (INF) sectors too.

Third, among the goods-producing industries, only construction (CNS) industries

has managed to retain its share, both in terms o value added and employment; in act,

its share o total employment has witnessed a small increase since the early 1990s.

2.2 Real GDP and real MVA

MVA is a consistent alternative to GDP or the measurement o the value o gross output.

Under this convention, the incomes generated in the service sectors excluded rom MVA

would be treated as transers, without being added to the product side o the accounts as

imputations. The aggregate economy as measured by MVA is smaller than as measured

by GDP due to the exclusion o these imputations. In 2009 MVA was 42% o GDP.

I MVA were a constant proportion o GDP over time, it would not make much di-

erence which measure we used. But over the postWorld War II period, MVA deated

by broad indices o prices such as the GDP deator has been growing more slowly

than real GDP. How does real GDP compare with real MVA over time? To answer this

question, we have tabulated (average annual compound) growth rates or the wholeeconomy (ALL), the private sector (PVT), both using real GDP categories, and the

value-adding part o the economy (MVA) or dierent time periods inTable 1.

4 The literature on unbalanced growth stemming rom Baumol (1967) ocusses on the eect o dierentialproductivity increases on economy-wide growth, sectoral relative price movements and asymptotic relativesectoral sizes. The literature on productive and unproductive labour, o whichWol (2006) and Shaikh andTonak (1996) are leading representative examples, addresses the measurement o exploitation, labour produc-tivity and economic stagnation. Dutt (1992) surveys these and related research in depth. Our concern herewith the relationship between value-added measures o industry output as an index o eective demand andindustry employment and in changes in this relationship over time, overlaps to some degree with this work buthas a dierent ocus. In some industries, such as FIRE, the lack o an independent measure o output suggeststhat whether or not the sector is productive in the sense that classical political economists and Marx used the

term, the value added reported does not reect employment-generating aggregate demand. In other industries,such as the goods-producing industries and service industries where an independent measure o output doesexist, we are concerned with changes in the relationship between measured output and employment over time.


ugust30,2013

http://c


Downloadedfrom


7/30


Several interesting acts emerge romTable 1. First, the growth rate o real MVA was

considerably lower than the growth rate o real GDP or either the whole economy or

the private sector: between 1948 and 2008, the real GDP or the whole economy grew

at an average rate o 3.34% a year; the real MVA o the economy grew at only 2.62% a

year. The ratio o the two growth rates was about 0.78. Given the growing importance

o incomes generated in industries such as FIRE, PBS, and the education and health

sector (EHS) over this period, it is not surprising that including these incomes as

imputed output raises the measured growth rate.

Second, the growth rate dierential widened during the neo-liberal period, which we

regard as 19802008. Between 1948 and 1973, the so-called golden age o capitalism,

the growth rate o real MVA was about 82.66% o the growth rate o real GDP; between

1980 and 2008, the neo-liberal era, the growth rate o real MVA was only 72.91% o the

growth rate o real GDP. This implies that the value-transerring (or value-wasting) parto the economy has grown relative to the value-adding part during the neo-liberal era.

The picture is even more pronounced i we restrict ourselves to the private sector o the

economy. I, as we will argue, real MVA is more closely linked to both aggregate demand

and output, the growing gap between real MVA and real GDP is an important actor in the

ailure o historic patterns o employment-output dynamics to appear in recent business

cycles when output is measured by real GDP.

2.3 Real NMVA and employment

We would now like to show that real MVA is as closely correlated with employment

as real GDP over the whole postWorld War II period in the United States and moreclosely correlated over the last three business cycles.5 Unortunately, consistent esti-

mates o MVA are available only at annual requency over the whole postWorld War

II period, due to changes in the industrial classifcation used by the BEA.

Thus, we concentrate in this section on a narrow measured value added (NMVA)

aggregate, which is calculated by removing goverment (GOV), FIRE, other services

(OTH) and rest o the world (ROW) rom nominal national income (NI) and adjust-

ing or ination by dividing the resulting nominal value added by the GDP deator.6

Table 1. Average annual compound growth rates (%)

Period ALL PVT MVA

19482008 3.34 3.31 2.62

194873 3.98 3.81 3.2919802008 2.99 3.02 2.18195059 3.62 3.34 2.66196069 4.65 4.46 4.12197079 3.57 3.81 3.59198089 3.39 3.40 2.27199099 3.32 3.52 3.0620002008 2.15 2.06 1.29

5 Parts o this section closely parallel the discussion in Foley (2011).6

Thus NMVA includes agriculture, mining, manuacturing, construction, transportation and public utili-ties, wholesale trade and retail trade. The value added realised in wholesale and retail trade can be regardedas part o the value added in productive sectors.


ugust30,2013

http://c


Downloadedfrom


8/30


NI and NMVA are reported consistently with each other on a quarterly basis by the

BEA. Moreover, real NI is a very close proxy to real GDP.

Both NI and NMVA correlate closely with US nonarm employment over the post

World War II period rom 1948 to 2010, as Figure 2 shows.

Figure 2 also shows that the relationship between aggregate demand as measuredby either NI or NMVA and employment shited downwards noticeably ater 2000

(though on this scale the eect appears rather small).

A closer look at the 200110 quarterly data is provided in Figure 3. Figure 3 shows

that the historical NI-employment relationship is a much poorer guide to aggregate-

demand-employment dynamics ater 2001 than the historic NMVAemployment

relationship. An analyst using NI as a measure o aggregate demand would have seri-

ously over-estimated employment, whereas an analyst using NMVA as a measure o

aggregate demand would have estimated employment considerably more accurately.

Something like this seems to have happened in the ormulation o fscal and monetary

policy in the immediate atermath o the fnancial crash in all 2008.

Both NI and NMVA have strong historical correlations with employment. The cor-

relation o NMVA with employment continued to hold in the last two business cycles.

There is, however, a cyclical component to the deviation o NI rom NMVA. NI shows

smaller cyclical downturns than NMVA, and more rapid recoveries. This cyclical devi-

ation appears to have increased in magnitude, at least over the last two US recessions.

As a result, cyclical NI uctuations have deteriorated as a guide to employment uc-

tuations in the US economy. I the goal is to understand the severity o business cycles

as uctuations in aggregate demand and the impact o aggregate demand on employ-

ment, NMVA is a better choice than NI as an index. The superiority o NMVA as a

business cycle and employment indicator is understandable because NMVA is much

Fig. 2.NI and NMVAEMP correlations NI (right) and NMVA (let) are plotted

on the horizontal axis, with the corresponding quarters US non-arm employment onthe vertical axis. The dashed lines show the ts or data up to 2000.


ugust30,2013

http://c


Downloadedfrom


9/30


more closely related to aggregate demand than the imputed value added in service

industries like FIRE, which, as we have seen, have a growing weight in NI and GDP.

3. Employment and output at the industry level

3.1 Empirical model

We now turn to an investigation o the changing relationship between output and

employment, not only at the aggregate level but also at more disaggregated, industry lev-

els. There are two distinct strands o the literature that can be used to study the changing

relationship between output and employment, the Okuns Law (OL) literature and the

Kaldor-Verdoorn (KV) literature (Verdoorn, 1949; Kaldor, 1957,1967,1968,1975).

The OL literature studies the relationship between growth rate o aggregate output and

changes in the unemployment rate; the KV literature, on the other hand, investigates the

raltionship between the growth rate o output and growth rate o employment (or pro-

ductivity). We choose to work within the KV tradition or three reasons.First, OL as a theoretical relationship is a reduced-orm relationship; it does not

have any deeper theoretical underpinning other than the simple idea that producing

output requires the use o labour power (Okun, 1962). The KV eect, on the other

hand, can be derived rom more primitive theoretical ideas; hence, it can be plausibly

understood as a structural relationship obtaining in capitalist economies and has been

used as such within the heterodox macroeconomics tradition, or instance as a styl-

ised act in the Dixon and Thirlwall (1975) model.7

Fig. 3. NI and NMVAEMP predictions, 20012010 The 19472000 tted relation betweenNI (right) and NMVA (let) together with quarterly points rom 2001Q1 to 2010Q4.

7 The KV relationship can be derived in several ways. For instance, it can be arrived at by combining

Kaldors technical progress unction with an accelerator-type relationship. It can also be derived by ormalis-ing Allyn Youngs ideas about increasing returns to scale within a neoclassical aggregate production unctionramework (McCombie, Pugno and Soro 2003).


ugust30,2013

http://c


Downloadedfrom


10/30


Second, OL-type analyses establish a relationship between the growth o output and

changes in the aggregate unemployment rate, the relationship being mediated through

changes in the labour orce participation rate. Thus, the Okun coefcient can change i the

labour orce participation rate changes with the relationship between output growth and

employment remaining unchanged. The OL ramework is thereore not suitable i one isinterested in primarily investigating the relationship between output growth and employ-

ment changes. The KV tradition, on the other hand, by directly ocusing on the relation-

ship between output growth and employment growth oers precisely such a ramework.

Third, while the OL-type analysis is pitched at the aggregate level, the KV rame-

work naturally allows or analysis at more disaggregated levels. Since, in the absence o

a sectorally captive labour orce, unemployment rates cannot be meaningully defned

at the industry level OL-type analysis cannot be naturally extended rom the aggregate

to the industry levels.8

We use a dynamic version o the KV regression equation in Kaldor (1966) to study the

changing relationship between output and employment in the post-war US economy:

g g g g ut

E

t

Y

i

n

i t i

Y

j

m

j t j

E

t= + + + +

=

=

1 1

(1)

where gt

E stands or growth rate o employment, gt

Y stands or the growth rate o

real output and ut

is an error term.

There are two parameters o interest that emerge rom the estimation o equation

(1), one that measures the contemporaneous eect o output growth on employment

growth and the other that captures the long-run eect o output growth on employ-

ment growth. On the one hand, the crucial parameter, in equation (1), measures the

partial eect o output growth on employment output: gives the change in the growthrate o employment that will result rom a 1 percentage point change in the growth

rate o real output; we call this the short-run (or contemporaneous) KV coefcient.

On the other hand, the long-run eect o output growth on employment growth can

be measured by *, where

* =1

=1

=

1

+

i

n

i

j

m

j

(2)

* gives the change in the growth rate o employment that will result rom a 1 per-

centage point change in the growth rate o real output when we allow the eect to

completely work itsel out over time, that is, allowing lagged eects to kick in.9

3.2 Some specication issues

The model in equation (1) is similar in structure to what is reerred to in the literature

as the dynamic version o Okuns Law (Knotek, 2007; IMF, 2010); dynamics is allowed

8

An exception is Palley (1993), which presented a multi-sector approach to Okuns Law.9 In the context o OL-type analyses, a parameter like * is known as dynamic beta; or details on dynamicbetas, see International Monetary Fund (2010).


ugust30,2013

http://c


Downloadedfrom


11/30


into the model through two channels: lagged independent variable and lagged depend-

ent variable. The frst specifcation issue that we want to discuss is related to the ques-

tion o whether lagged dependent variables should be included in the empirical model.

Despite its wide use in the literature, we believe there are serious drawbacks to

including lags o the dependent variable in a model like equation (1) when the sam-ple size or estimation is not very large. In a time-series setting, inclusion o lagged

dependent variables violate key exogeneity assumptions and make the ordinary least

squares (OLS) estimates o the parameters inconsistent. Although the inclusion o

lagged dependent variables can be justifed in a large sample setting, our ocus on

business cyclelength time periods or estimation o the model recommends that we

avoid including lagged dependent variables. Oten, inclusion o lagged dependent vari-

ables is justifed as a mechanism or dealing with problems o serial correlation in

the errors; this is not necessary as heteroscedasticity and autocorrelationconsistent

(HAC) standard errors can be used to deal with problems o serial correlation o

errors without also introducing the problems o inconsistent estimation that come

with lagged dependent variables. Hence, the ocus o our analysis will be on the model

without lagged dependent variables; on the other hand, we allow or two lags o the

independent variable to capture dynamic eects. Hence, the model we estimate is

g g g g ut

E

t

Y

t

Y

t

Y

t=

1 1 2 2 + + + +

(3)

with the contemporaneous KV coefcient given by and the long-run KV coe-

fcient given by

*1 2

= + +

(4)

Even in the model without lagged dependent variables, we need to address one

important specifcation issue: possible endogeneity o the growth rate o output in

equation (3). Rowthorn (1975) and Skott (1999) have pointed to the possibility o the

endogeneity o growth rate o output in a regression like equation (3) and have asserted

that single-equation estimation methods are thereby invalid. To address this issue, we

report results rom two statistical tests o the endogeneity o the growth rate o output

in equation (3): the HAC score test (Wooldridge, 1995) and the C-statistic type test

o endogeneity (Hayashi, 2000). The idea behind both tests derives rom Hausman

(1978) and relies on comparing key statistics that would be close to each other or

cases with and without endogeneity o the relevant regressor. The null hypothesis, in

both cases, is that the relevant regressorgrowth rate o output, in our caseis exog-

enous and largep-values imply that the null cannot be rejected. Appendix Tables A1

and A2 report thep-values rom these tests or the nine post-war business cycles and

the six major sectors that are studied in this paper.10 In an overwhelming number o

cases, the highp-value suggest that the null hypothesisexogenous rate o growth o

outputcannot be rejected. This implies that estimating the parameters o equation

(1) by OLS gives reasonably accurate estimates o the true KV coefcients, in both

the short run and the long run. That the instruments used or the endogeneity tests are

indeed exogenous can be seen rom the results reported in AppendixTable A3, which

reportsp-values rom Hansens overidentifcation test; or this test, the null hypothesis

10 This analysis uses quarterly data; or details o the construction o the data set, see the Appendix.


ugust30,2013

http://c


Downloadedfrom


12/30


is that the instruments are exogenous, and thus largep-values imply that the null can-

not be rejected at standard levels o statistical signifcance.

4. The discrepancy in the current recession

We are now ready to turn to one o the motivating questions o this article: the startling

discrepancy between employment-output dynamics in the current recession and what

one would expect on the basis o past trends.

To highlight the discrepancy between actual and predicted employment changes,

we compute out-o-sample orecast or the average annual compound growth rate o

employment allowing or lagged eects o observed output growth rates and compare

it with observed growth rates o employment during the same period. To fx nota-

tions, suppose, starting in period t, a downturn lasts or n quarters. Let gt k

E

+and

gt k

E

+denote the actual and orecast quarter-over-quarter growth rate o employment

between quarter t+ k 1 and t+ k, respectively, where k = 0, 1, . . ., (n 1), and

orecasting is done using model (3) with a data set extending over the previous peak-

to-peak business cycle.

The actual average annual compound growth rate o employment in the downturn,

gE , is given by

g g gEt

E

t n

En= (1 ) (1 ) 1 4001

+ + { } +

On the other hand, the orecast average annual compound growth rate o employ-

ment in the downturn, gE , is given by

g g gEt

E

t n

En= (1 ) (1 ) 1 4001

+ + { } +

(5)

For instance, or computing the orecast average annual peak-to-trough employ-

ment growth rate or the 1991 recession, the model in equation (3) is estimated with a

data set that runs rom 1980Q1 to 1990Q3. The estimated parameters are then used to

compute orecast growth rates or each o the quarters in the recession using equation

(3). The orecast average annual growth rate o employment is then computed with the

help o equation (5). Figure 4 plots gE and gE or all the post-war downturns or the

six industry groupings under investigation.11

The comparison o the recovery period between actual and orecast employmentgrowth is carried out in a similar manner. At the time o this writing, data were avail-

able or the current recovery or only a fve-quarter period ater the trough in 2009Q2,

so we have chosen to ocus on a fve-quarter period ater each trough to compare the

orecast and actual growth rates o employment during the recovery period. For a

fve-quarter recovery ater the trough starting in period t, the actual average annual

compound growth rate o employment during the recovery, gE , is given by

g g g g g gEt

E

t

E

t

E

t

E

t

E= (1 )(1 )(1 )(1 )(1 ) 1 4001 2 3 4

5 + + + + + { } + + + +

11 In this article, we use NBER dates or business cycle turning points.


ugust30,2013

http://c


Downloadedfrom


13/30


2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

ALL SECTORS

average annual compound growth rate of employment: peaktotrough

5 4 3 2 1 0

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

CONSTRUCTION


10 5 0

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

FINANCE, INSURANCE AND REAL ESTATE


2 0 2

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

MEASUREABLE VALUE ADDED


7 6 5 4 3 2 1 0

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

PRIVATE GOODSPRODUCING INDUSTRIES


10 8 6 4 2 0

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

PRIVATE SERVICESPRODUCING INDUSTRIES


3 2 1 0 1 2

ActualPredicted

Fig. 4. Actual (grey) versus predicted (black) average annual compound growth rate (%) oemployment: peak to trough


ugust30,2013

http://c


Downloadedfrom


14/30


On the other hand, the orecast average annual compound growth rate o employ-

ment in the recovery period, gE , is given by

g g g g g gEt

E

t

E

t

E

t

E

t

E= (1 )(1 )(1 )(1 )(1 ) 11 2 3 4

5 + + + + + { }+ + + +

400

(6)

For instance, or computing the orecast trough-to-recovery (average annual)

employment growth rate or the 1991 recession, the model in equation (3) is estimated

again with a data set that runs rom 1980Q1 to 1990Q3. The estimated parameters are

then used to compute orecast growth rates or each o the fve quarters in the recovery

using equation (3). The orecast average annual growth rate o employment is then

computed with the help o equation (6). Figure 5 plots gE and gE or all the post-war

recoveries or the six sectors under investigation.

The most striking trend emerging rom Figures 4 and 5 is the aggregate evidence o a

structural break in the relationship between employment and aggregate demand some-

time during the 1980s (beginning o the neo-liberal period), especially during the recov-ery phase o business cycles. The frst chart in the frst row in Figure 5 (whole economy)

shows that or all recoveries prior to the double-dip recession in 1980, predicted and

actual employment changes in the fve post-trough quarters moved in the same direc-

tion; both were positive but dierent in magnitude. During the 1982 recovery, this trend

was reversed or the frst time: while predicted employment change was positive, actual

employment change was negative. In the next three recessions, this trend reversal has

continued unabated.

5. Evolution of industry outputemployment relationships

Our main interest is not in studying the relationship between output and employment

over the whole post-war period, but in investigating how that relationship has changed

over time. The pattern o change emerges clearly rom rolling regressions and a decom-

position analysis. In the ollowing subsections we present results rom variable-width

(business cycle length) rolling regressions and a decomposition analysis to highlight

the declining trend in the coefcient relating output growth to employment growth.

5.1 Variable-width rolling regressions

For variable-width rolling regressions, we estimate the model in equation (3) or each

post-war business cycle (peak to peak).12

In essence, thus, these are variable-width roll-ing regressions with business cycle window lengths.

These regressions give us the short-run and long-run KV coefcients, which are then

plotted across time to inspect the changing pattern o the response o employment to

output growth. Data or these rolling regression plots come rom the BEA (or aggre-

gate output) and the Bureau o Labor Statistics (BLS) (or employment). Aggregate

output is proxied by national income at the industry level; these data are available at a

quarterly requency rom NIPA Tables 6.1 B, C and D o the BEA. Nominal national

income has been deated by the GDP deator to arrive at a measure o real output.

12 The short peak-to-peak cycle between 1980Q1 and 1981Q3 is ignored; instead, the whole period rom1980Q1 and 1990Q3 is considered one peak-to-peak cycle, giving us a total o nine cycles.


ugust30,2013

http://c


Downloadedfrom


15/30


2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

ALL SECTORS

average annual compound growth rate of employment: troughtorecovery (4 qtr)

0 1 2 3 4

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

CONSTRUCTION


5 0 5

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

FINANCE, INSURANCE AND REAL ESTATE


1 0 1 2 3 4 5

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

MEASUREABLE VALUE ADDED


0 1 2 3 4 5

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

PRIVATE GOODSPRODUCING INDUSTRIES


4 2 0 2 4 6

ActualPredicted

2007

07

2001

01

1991

91

1980

80

1973

73

1969

69

1960

60

1957

57

1953

53

PRIVATE SERVICESPRODUCING INDUSTRIES


0 1 2 3 4

ActualPredicted

Fig. 5. Actual (grey) versus predicted (black) average annual compound growth rate (%) o employment:trough to recovery


ugust30,2013

http://c


Downloadedfrom


16/30


Non-arm employment data at the industry level are available at a monthly requency

rom Table B1 o the BLS; these data are converted into a quarterly requency by aver-

aging monthly data or relevant months in a quarter. Thus, the data set or the rolling

regression plots run rom 1948Q1 to 2010Q3.13

Figures 6 and 7 plot the short-run and long-run KV coefcients over each post-warbusiness cycle. The frst row oFigures 6 and 7 gives plots or the whole economy, and

the non-fnancial value-adding (NFVA) sector o the economy14; the second row cov-

ers the whole private services-producing industries and FIRE (the largest component

o the services sector); the last row displays elasticities or the whole private goods-

producing industries and construction.

Three striking trends emerge rom Figures 6 and 7. First, or the whole economy

and the NFVA sector o the economy, there has been a sharp all in both the short-run

and long-run KV coefcients; there is a discernible downwards trend in the frst rows

oFigures 6 and 7. The short-run KV coefcient has allen rom about 0.4 to around

0.1; the long-run coefcient has declined rom the region o 0.8 to around 0.3. Both

fgures display decade-long (or longer period) uctuations around the downwards

trend. The downwards trend or the long-run KV coefcient is especially pronounced

since the mid-1980s.

Second, the private services-producing industries, taken together, do not display any

declining trend or the whole post-war period. FIRE, the largest component o the pri-

vate services-producing sector, has always had a numerically small (i.e., close to zero)

short-run KV coefcient; there is no observable trend in the KV coefcient or FIRE.

The long-run KV coefcient, on the other hand, does display a signifcant downwards

trend rom the early 1980s.

Third, the private goods-producing industries behave very dierently rom the services-

producing industries. The private goods-producing industries as a whole display a signif-cant downwards trend in both the short-run and long-run KV coefcients over the whole

post-war period. Although there is an upwards trend in the early 1970s, that gets quickly

reversed and there is a pronounced decline since the mid-1970s. Manuacturing, the larg-

est component o the private goods-producing industries, displays the same trend

although we do not include the manuacturing sector in Figures 6 and 7and drives the

result or the whole goods-producing sector.

Although visual inspection o trends in Figures 4 and 5 show signifcant declines in

the values o the short- and long-run KV coefcient or the whole economy and or

most sub-sectors, Table 2 brings statistical evidence to bear on the issue o decline.

Ater all, the decline that is discerned by visual inspection might not be statisticallysignifcant; it might be driven by pure sampling error. In Table 2, we report results o

13 In deciding on the data source to use or the analysis in this article we aced a trade-o between levelo disaggregation and requency. The AIA data are available at the one- and two-digit level o NAICS but atan annual requency; the BEA national income data are available at a quarterly requency but do not reportvalues or all the NAICS codes. For the regression analysis we chose the quarterly requency data set andsacrifced some disaggregation.

14 The NFVA sector is composed o the private goods-producing industries (mining, construction andmanuacturing) and the private services-producing sector less FIRE. Note that NFVA is a broader measure

o value-added than MVA (and by implication NMVA) defned earlier. Lack o national income data atsufciently disaggregated levels prevents us rom computing MVA at a quarterly requency; hence, we useNFVA instead.


ugust30,2013

http://c


Downloadedfrom


17/30


Fig. 6. Short-run Kaldor-Verdoorn coeicient over business cycles (peak to peak)


ugust30,2013

http://c


Downloadedfrom


18/30


Fig. 7. Long-run Kaldor-Verdoorn coeicient (without lagged dependent variable) over businesscycles (peak to peak)


ugust30,2013

http://c


Downloadedfrom


19/30


testing the null hypothesis that the KV coefcientsboth short-run and long-runare

the same between the business cycles at the two ends o the period under consideration.

Suppose we wish to compare the elasticities between two business cycles, indexed

by i= 1,2. Let 1

and 2

be the OLS estimators or the true KV coefcient 1 and

2 in the two business cycles, respectively. Let s12 and s

22 be the HAC estimators o

the variance 1

2 and 22 , respectively, o the OLS estimators

1and

2. Then, the

test statistic

ts s

=( )1 2

1

2

1

2

+

(7)

is distributed asymptotically as a standard normal random variable under the null

H0 1 2

: =

The frst two columns oTable 2 report the test statistic or testing the null that the

short-run and long-run KV coefcient is same or the ollowing two peak-to-peak busi-

ness cycles: 1948Q41953Q2 and 2001Q12007Q4 (i.e., the whole post-war period);the next two columns report the test statistic or comparisons between the ollowing

two peak-to-peak business cycles: 1973Q41980Q1 and 2001Q12007Q4 (i.e., the

neo-liberal period). The alternative hypothesis is that the KV coefcient in the initial

period is higher than that in the fnal period; hence, a positive and large value o the

test statistic reported inTable 2 is strong evidence against the null o no change in the

KV coefcient over time.

Two interesting acts emerge romTable 2. First, the services-producing industries

(PSV) and its largest component, fnance, insurance and real estate (FIRE), either ail

to reject the null hypothesis or reject it in a much weaker ashion than the other sec-

tors. This seems to suggest that the growing strength o the KV eect has been much

more prominent in the non-services part o the economy. Second, the rejection o thenull hypothesis is weaker in the neo-liberal period than or the whole post-war period.

Table 2. Testing or decline in elasticities

Sector Post-war period Neo-liberal period

Short run Long run Short run Long run

ALL 4.16*** 4.20*** 2.71*** 2.12**NFVA 2.43*** 2.07** 1.86** 1.82**PGD 5.77*** 4.61*** 4.43*** 3.15***CNS 3.41*** 3.13*** 5.09*** 2.08**PSV 1.30 2.77*** 1.37 1.36FIR 1.23 0.89 0.66 2.93***

Notes: The null hypothesis is that the estimates o Kaldor-Verdoorn coefcient is the same between thebusiness cycles at the two ends o the period under consideration; the (one-sided) alternative is that thecoefcient is larger in the initial period. The entries in the table are the values o the test statistic in equation(7); a large positive value o the test statistic is evidence against the null. Signifcance levels: ***1%; **5%.


ugust30,2013

http://c


Downloadedfrom


20/30


This is because the declining trend in the KV coefcient was arrested and even sig-

nifcantly reversed or about a decade-long period beore the onset o neo-liberalism.

5.2 Decomposition o KV eect

The interpretation o the evidence presented in Figures 6 and 7 suggest that the ser-

vice sector, especially FIRE, behaves dierently rom the rest o the economy. Hence,

the ollowing question arises naturally: is the decline in the KV coefcient driven pri-

marily by the sectoral shit in the US economy? To address this question, we carry out

a simple decomposition exercise.

Suppose the economy is composed o two sectors,A and B. For instance,A could

reer to the goods-producing industries (PGD) and B could reer to the services-

producing industries (PSV); or,A could reer to the value-adding sectors (NFVA)

and B could reer to the nonvalue-adding sectors (i.e., the aggregate economy less

NFVA). Let subscripts Yand Ereer to output and employment and superscripts A

and B reer to the two sectors that make up the aggregate economy. Thus, g gY YA, andg

Y

B reer to the growth rate o output in the aggregate economy and the two sectors,

respectively; similarly, gE

reers to the growth rate o employment in the aggregate

economy, gE

A and gE

B reer to the growth rates o employment in sectors A and B,

respectively. I SE

A reers to the share o sectorA in total employment, then

g S g S gE E

A

E

A

E

A

E

B= (1 )+

Dierentiating the above expression with respect to gY

(the growth rate o aggre-

gate output), we get

=dg

dg

E

Y

= (1 ) ( )Sdg

dgS

dg

dgg g

dS

dgEA A Y

A

Y

E

A B Y

B

Y

E

A

E

B E

A

Y

+ +

+ + Sg

gS

g

gg g

S

gEA A Y

A

Y

E

A B Y

B

Y

E

A

E

B E

A

Y

(1 ) ( )

where is the marginal eect o output growth on employment growth or the whole

economy, and A and B are the marginal eects o output growth on employment

growth or sectorA and B, respectively. Note that the last step gives us an approxima-

tion because we replace dierentials (i.e., instantaneous change) with dierences (i.e.,

change over a fnite time period).

Thus, i , A and B are consistent estimators o,A andB or some time period,then we have the ollowing decomposition o in terms o the relative shares o thetwo sectors in total employment S

E

A , the relative growth rates o output in the two

sectors, g gYA

Y/ , and estimates o the marginal eects o output growth on

employment growth A and B in the two sectors:


ugust30,2013

http://c


Downloadedfrom


21/30


+ + AE

A Y

A

Y

B

E

A Y

B

Y

E

A

E

B E

A

Y

Sg

gS

g

gg g

S

g( ) (1 ) ( )

(8)

For a period when the relative employment shares in the two sectors remain

unchanged, the last term in equation (8) drops out and we have

CSE

A

E

A Y

A

Y

B

E

A Y

B

Y

Sg

gS

g

g + ( ) (1 )

(9)

where CSE

gives the value o the KV coefcient that would have approximately

arisen under a constant share o employment (hence the subscript CSE).Tables 3 and 4

give values o and CSE

, that is, both short-run and long-run KV coefcients, or

two dierent sectoral decompositions. InTable 3, the total private sector is broken up

Table 3. Decomposition o Kaldor-Verdoorn coeicients (PGD and PSV)

Cycle Contemporaneous Long run

(1) (2) (3) (4)

1948Q41953Q2 0.41 0.31 0.73 0.671953Q21957Q3 0.26 0.25 0.74 0.711957Q31960Q3 0.45 0.34 0.80 0.901960Q31969Q4 0.20 0.18 0.65 0.851969Q41973Q4 0.27 2.05 0.64 2.281973Q41980Q1 0.35 0.50 0.80 0.97

1980Q11990Q3 0.37 0.27 0.66 0.521990Q32001Q1 0.21 0.06 0.61 0.192001Q12007Q4 0.11 0.08 0.42 0.31

Notes: Columns (1) and (3) give the estimated Kaldor-Verdoorn coefcient, that is, in equation (3);columns (2) and (4) give the value o the Kaldor-Verdoorn coefcient computed under the assumption thatthe share o employment does not change over the business cycle, that is,

CSE

in equation (9).

Table 4. Decomposition o Kaldor-Verdoorn coeicients (NFVA and PVT less NFVA)

Cycle Contemporaneous Long run

(1) (2) (3) (4)

1948Q41953Q2 0.41 0.40 0.73 0.691953Q21957Q3 0.26 0.25 0.74 0.701957Q31960Q3 0.45 0.50 0.80 0.861960Q31969Q4 0.20 0.13 0.65 0.401969Q41973Q4 0.27 0.35 0.64 0.771973Q41980Q1 0.35 0.30 0.80 0.701980Q11990Q3 0.37 0.33 0.66 0.601990Q32001Q1 0.21 0.71 0.61 2.242001Q12007Q4 0.11 0.19 0.42 0.50

Notes: Columns (1) and (3) give the estimated Kaldor-Verdoorn coeicient, that is, in equation(3); columns (2) and (4) give the value o the Kaldor-Verdoorn coefcient computed under the assump-tion that the share o employment does not change over the business cycle, that is,

CSEin equation (9).


ugust30,2013

http://c


Downloadedfrom


22/30


into goods-producing and services-producing industries; inTable 4, the total private

sector is decomposed between a value-adding sector (NFVA) and a value-using sec-

tor (PVT less NFVA). In Tables 3 and 4, comparison o columns (1) and (2) or the

short-run KV coefcient, and comparison o columns (3) and (4) or the long-run KV

coefcient shows that the value o both coefcients all over the post-war period evenwhen share o employment remains unchanged between goods and services-producing

industries, and between NFVA and non-NFVA.

This decomposition, thereore, highlights the important act that the all in the KV

coefcient or broad aggregates o the US economy reects both the shit o value

added towards the service sector (which has a lower coefcient) and a all in the coe-

fcients in the subsectors themselves.15

6. Explanations

The evidence we have presented here suggests that the KV eect has become signif-

cantly stronger in the US economy in recent decades, and this is one o the actors at

the root o the changing connection between measured real output and employment.16

This raises the question o explanations or these structural changes.

Three possible explanations are: a shit in the US labour market to a greater reliance

on temporary workers (exible labour); the act that, as Kaldor noted, real output in

service-producing sectors is not independently measured rom income in those sec-

tors, so that some observed increases in real output are illusory (mis-measurement o

real output); and globalisation, which leads to restructuring o production so as to shit

low-value-added employment rom the United States to lower-wage regions o the

world (global restructuring). We discuss these possibilities in the light o the evidence

we have presented.

6.1 Flexible labour

One o the major explanations oered in the mainstream literature or the discrep-

ancy between actual and predicted employment changes that is oten construed as a

breakdown o Okuns Law, is what might be called the exibilisation o labour (or

instance, captured by the increase in the share o temporary workers in the workorce)

across industries. This argument was used by Groshen and Potter (2003) to explain

the jobless recovery ater the 2000 recession in the United States; a broader but simi-

lar argument has been made in Gordon (2010) and the 2010 World Economic Outlook

o the IMF using a cross-country ramework (IMF, 2010).This argument is straightorward. When the labour regime becomes more exible

and unionisation rates go down, frms cut down on the permanent workorce and

replace them with temporary workers. When there is an increase in demand during

business cycle upturns, frms hire temporary workers rapidly, and in a slump they shed

most o their temporary workorce. Thus, job losses are now permanent in nature as

opposed to temporary layos o permanent workers, that is, job losses lead to perma-

nent severing o employeremployee ties rather than temporary severing o these ties

15

For the 196973 business cycle, the growth rate o the goods-producing sector was large and negative;that is driving the coefcients in columns (2) and (4) inTable 3 to be so large and negative.16 An increase in the KV coefcient imples a weakening o the KV eect.


ugust30,2013

http://c


Downloadedfrom


23/30


with the implicit assumption o resuming the relationship ater the economic storm

is weathered. In the context o technological change and globalisation o production

(i.e., relocation and outsourcing), the increase in the share o the temporary workorce

increases the possibilities o structural change (i.e., permanent shit o jobs across sec-

tors, industries, regions and even countries). When demand revives during the upswingo the business cycle, the hiring process is much more subdued because some o the

jobs are simply not there anymore and creating new jobs takes more time in the ace o

increased uncertainty. Hence, recoveries are jobless.

There is certainly a ring o plausibility to this argument; the idea o exibilisation o

labour seems to ft in with the general neo-liberal turn since the late 1970s. But we dont

think this argument captures the whole story. Even though the increasing exibilisation o

the workorce

basu, foley - dynamics of output and employment in the us economy (2013)

Documents