ncp19 07 jb - centre for economic performance · 2011. 10. 26. · r&d centrepiece issn 1362-3761...

R&D

ISSN 1362-3761

C e n t r e Pi e c eThe Magazine of The Centre for Economic Performance Volume 16 Issue 2 Autumn 2011

Policy for innovationand growth

NEETs Process innovationTeachers’ pay Unequal BritainPostgraduates French schoolsFootball and crime Jobs in Europe

Policy-makers across Europe and theUnited States want growth and jobs,desperately seeking measures toreinvigorate their flagging recoveries.Researchers at the Centre for EconomicPerformance (CEP) have spent many yearsthinking about the fundamental questionthat underlies that search – what drivesproductivity growth?

The last issue of CentrePiece describedthe substantial body of CEP researchevidence on the important role ofcompetition in raising productivity. Oneway it does that is by promotinginnovation: in an effort to get ahead oftheir competitors, firms come up withnew ideas for products – think of SteveJobs’ Apple. Firms also find better ways ofmaking their existing products: in thisissue, Tim Leunig and Joachim Voth show the huge growth benefits of such‘process innovation’ in the cotton and car industries.

Competition policy can have asignificant impact in support of

innovation and growth, but public policycan also influence innovation moredirectly. In our cover story, CEP’s directorJohn Van Reenen outlines the Centre’swork on fiscal incentives for businessspending on research and development(R&D). The evidence indicates that R&Dtax credits can be highly effective inencouraging innovation.

The skills of the workforce areanother important contributor togrowth. Elsewhere in the magazine arearticles on the two ends of Britain’sdistribution of what economists call‘human capital’. Hilary Steedmandiscusses the NEETs – young people,often with few qualifications, who are‘not in education, employment ortraining’. And CEP’s research directorStephen Machin describes the growingtrend of students staying in universityafter graduation so as to acquire furtherqualifications.

The boom in postgraduate educationis increasing inequality among people

who have degrees as well as morebroadly in the labour market. HenryOverman and Steve Gibbons also focuson ‘unequal Britain’, examining thereality of regional disparities. Theircritique of the foundations of successivegovernments’ urban and regionalpolicies suggests that a great deal oftime and public money has been wastedon policies that do little either to reduceinequality or promote growth.

Finally, CEP’s Nobel laureate ChrisPissarides compares the records ofEurope and the United States on jobcreation, especially in the potentialgrowth sectors of healthcare, educationand business services. His findings are yetanother illustration of the powerfulimpact of public policy on employment.The key is to figure out which policieswill retard growth and which policieswill boost growth.

Romesh Vaitilingam, [email protected]

CentrePiece is the magazine of the

Centre for Economic Performance at the

London School of Economics. Articles in this

issue reflect the opinions of the authors, not

of the Centre. Requests for permission to

reproduce the articles should be sent to the

Editor at the address below.

Editorial and Subscriptions Office

Centre for Economic Performance

London School of Economics

Houghton Street

London WC2A 2AE

Annual subscriptions for one year (3 issues):

Individuals £13.00

Students £8.00

Organisations (UK and Europe) £30.00

Rest of world £39.00

Visa and Mastercard accepted

Cheques payable to London School

of Economics

CEP director, John Van Reenen

CEP research director, Stephen Machin

Editor, Romesh Vaitilingam

Design, DesignRaphael Ltd

Print, Hastings Printing Co. Ltd.

© Centre for Economic Performance 2011

Volume 16 Issue 2

(ISSN 1362-3761) All rights reserved.

Centre Piece

Editorial

mailto:[email protected]

CentrePiece Autumn 2011

1

page 6Cotton and cars:the huge gains from process innovation Tim Leunig and Joachim Voth measurethe benefits to society of two bigimprovements in the way we make things

page 14House prices and schoolquality: evidence from state and private educationin ParisGabrielle Fack and Julien Grenet findthat the presence of private schoolsmitigates the impact of state schoolperformance on Parisian house prices

in brief...

page 2Big ideas: innovation policyJohn Van Reenen describes the impact of CEP research on fiscalincentives for business spending on research and development

page 8Football matches: the effects on crimeOlivier Marie explores how the heavy police presence at footballmatches in England affects crime – in the stadium and elsewhere

page 11Young people without qualifications: how‘headline numbers’ shape policy andaspirationHilary Steedman compares England’s ‘NEET’ indicator with howFrance and Germany measure school leavers’ progression

page 16The boom in postgraduate education and itsimpact on wage inequalityJoanne Lindley and Stephen Machin find that growingnumbers of university students are staying on after their firstdegrees to get extra qualifications

page 20Teachers’ pay and pupil performancePeter Dolton and Oscar Marcenaro-Gutierrez examine theenormous variation in teachers’ pay across OECD countries and itssignificance for educational outcomes

page 23Unequal Britain: how real are regional disparities?Henry Overman and Steve Gibbons argue that regional policypays too little attention to differences in the cost of living andavailability of local amenities

page 26The future of work in EuropeNobel laureate Chris Pissarides explains why Europe remainsbehind the United States in job creation in health, education andbusiness services

page 11Young people

withoutqualifications

UNEXPECTED ITEM IN THE BAGGING AREA

page 6 Cotton and cars

Contents

page 26 The future of work in Europe


In the last CentrePiece, John Van Reenenstressed the importance of competition andlabour market flexibility for productivity growth.His latest in CEP’s ‘big ideas’ series describesthe impact of research on how policy-makerscan influence innovation more directly –through tax credits for business spending onresearch and development.

Big ideasInnovation policy


3

In the wake of the GreatRecession, the UK is hardlyalone in looking for sources ofeconomic growth. Economistsand many other commentatorsagree that technological

innovation must be at the heart of long-run growth. It is also widely understoodthat left to itself the market is unlikely toprovide enough incentives for innovation.

This ‘market failure’ is primarilybecause only a small proportion of thebenefits of invention are captured by thefirm or individual who spends money andtime on research. Most of the benefits ofinvention ‘spill over’ to other firms whocan copy the new idea without having topay the upfront research costs. Forexample, it took a lot of effort to inventthe automobile and the personalcomputer – but once they were invented,imitators crowded in.

This means that there will be too littlespent on research and development(R&D) from the point of view of societyas a whole. Intellectual property rights,such as patents and copyright, weredesigned to protect inventors andincrease their incentive to innovate. Butin most cases patents can be designedaround so they do not fully eliminate themarket failure.

So can there be a role for publicpolicy in stimulating innovation? Is itdriven by fundamental factors, such asculture and luck, which are beyond theability of governments to influenceexcept in the most minor ways?

CEP research has challenged the fatalisticattitude that innovation is not amenableto government action. One direct way toinfluence innovation is through the taxsystem, in particular by offering a taxbreak for business spending on R&D.

I started working on fiscal incentivesfor R&D in the mid-1990s after beingshocked to discover that the share of UKnational income spent on business R&Dhad declined since the late 1970s. In justabout every other developed country, ithad been rising.

The United States introduced an R&Dtax credit in 1981 under Ronald Reagan,but the UK Treasury had always resistedthe idea, arguing that firms were unlikelyto increase their R&D efforts significantlyin response. Evaluations of the US systemseemed to show, however, that after afew teething problems, American firmshad responded to these tax incentives.

Working with Bronwyn Hall of theUniversity of California at Berkeley, ourreview of all the existing evidenceshowed that when researchers used goodquality firm-level data and trackedcompanies over time, they found that taxcredits stimulated significant AmericanR&D spending (Hall and Van Reenen,2000). Were UK firms likely to be somuch more lethargic than theircounterparts across the Atlantic?

At that time, international evidenceon the effectiveness of innovation taxpolicy was almost non-existent. No onehad even collected systematic informationon the tax benefits to R&D across

countries over time – not theInternational Monetary Fund, the OECD,the World Bank or the United Nations.

Together with Rachel Griffith (nowdeputy research director of the Institutefor Fiscal Studies), we put together ateam and embarked on a major effort tomeasure the impact of the tax system onthe costs of R&D capital across all themajor economies over 20 years. Adownside of this was that we had towade through many dusty tomes ofrather tedious tax and accounting rules.

Once we had accomplished thatarduous task, we were able to show thatthere had been a major shift towardsR&D tax credits and away from directsubsidies. One of the advantages of taxcredits over the more traditional grantswas that the government could simplyset the rules and it did not have to getinvolved with ‘picking winners’.

More importantly perhaps, wecombined the tax data with informationon national R&D and showed that taxcredits had a large effect on increasingbusiness R&D. Although a 10% reductionin the tax costs only increased privatesector R&D spending by about 1% in thefirst year after an R&D tax credit wasintroduced, in the long run R&D volumesrose by a full 10% (Bloom et al, 2002).

So far, so good. But what we careabout is not R&D per se, as this is just aninput. We care about economic growth,which will increase wages andconsumption. To tackle this problem, we had to develop a new model of

s

A strong R&D base helps a country to imitate

as well as innovate

Big ideas

‘endogenous growth’ that took accountof not just the obvious effect of R&D oninnovation but also the less obvious‘second face’ of R&D, which fostersdiffusion of existing innovations.

Having more scientists helps the UKcatch up with leading-edge countriesbecause they can read and understandnew ideas, which can then be ‘absorbed’more effectively in the UK economy. For acountry like the UK, which is sadly oftenfar from the technological frontier, this isvery important. It means that just sittingback and letting other countries – theUnited States, Germany, Japan andincreasingly China – do all the innovationis unlikely to be the right strategy.

A strong R&D base helps a country to imitate as well as innovate. In a speech on the science budget last year,David Willetts, the universities andscience minister, quoted CEP’s research in this area:

‘Some 95% of scientific research isconducted outside the UK. We need to beable to apply it here – and, in advancedscientific fields, it is often necessary toconduct leading-edge research in order tounderstand, assimilate and exploit theleading-edge research of others.

‘It is this absorptive capacity which iscrucial. Indeed, Griffiths, Redding and VanReenen have shown that higher domesticbusiness R&D spend also leads to greaterproductivity being generated at homefrom foreign R&D spend as well. Andthere are powerful feedback mechanismson top of this – foreign companies cite thequality of the public research base as oneof the main reasons for locating their owninternationally mobile R&D here.’

In a series of studies with PrincetonUniversity’s Steve Redding (who wasdirector of CEP’s globalisation programmefrom 2005 to 2010), we created aneconometric model for the whole OECD,which showed how R&D stimulated


4

productivity growth through bothinnovation and imitation (Griffith et al,2004). We combined this with our R&Dtax information to simulate the effects ofintroducing an R&D tax credit in the UK.

We found that the benefits of anR&D tax credit would easily outweigh thecosts, which implied that it could be asuccessful policy (Griffith et al, 1999).But we also cautioned that, as ever, thedevil was in the detail: making the taxcomplicated could unwind its effects.

Our work appeared in academicjournals, policy pamphlets and theprinted and broadcast media. It was alsopicked up by Dan Corry (who was tobecome head of the No. 10 policy unitduring Gordon Brown’s term of office asprime minister) when he was at theInstitute for Public Policy Research. Theproposals became part of Labour’s 1997manifesto, which ushered in the firstever R&D tax credit in the UK in 2001 –initially just for small and medium-sizedenterprises, but later extended to firmsof all sizes.

In an illustration of the closeconnection between academic ideas andpolicy-making, our co-author on the taxcredit work was Nick Bloom, who wasthen my PhD student. After graduating,he was seconded to the Treasury to helpthem introduce the tax credit and itsextension to large firms (in 2003). Nick led CEP’s research programme onproductivity between 2003 and 2006and pushed forward the Centre’scontinuing investigations of the ways inwhich policy can be used to stimulateinnovation.

The R&D tax credit is under review bythe current government, but it looks likeit will remain a permanent fixture of thefiscal scene. The UK’s R&D intensitystopped declining in the mid-2000s,which coincides with the bedding downof the R&D tax credit. This could be

The UK’sR&D intensitystoppeddeclining in the mid-2000s,whichcoincideswith thebedding downof the R&Dtax credit

Big ideas


5

coincidence and more rigorousevaluations of the effects of the fiscalincentives are needed. Nevertheless, the initial findings are encouraging andour work in progress suggests that USR&D tax credits raise firms’ marketvalues, productivity and innovation(Bloom et al, 2010).

The R&D tax credit story is a usefulparable of the interaction offundamental economic research withpolicy development and implementation.It contrasts with the ‘patent box’, apoorly targeted policy that consists of taxbenefits to the royalties on patents: suchrewards create few ‘spillovers’ as theresearch is already done. The patent boxpolicy was proposed in the dying days ofthe Labour government and it will wastearound £1 billion a year at a time whenthe country can ill afford it. While thislooks like being another of the fewLabour policies that the government iscontinuing, in this case the continuity is unfortunate.

John Van Reenen is director of CEP.

Further reading

Nicholas Bloom, Rachel Griffith and

John Van Reenen (2002) ‘Do R&D Tax Credits

Work?’, Journal of Public Economics 85:

1-31 (earlier version available here:

http://cep.lse.ac.uk/textonly/people/

vanreenen/papers/gvrb15_wp.pdf).

Nicholas Bloom, Mark Schankerman and John

Van Reenen (2010) ‘Identifying Technological

Spillovers and Product Market Rivalry’, CEP

Discussion Paper No. 675

(http://cep.lse.ac.uk/pubs/download/

dp0675.pdf)

Rachel Griffith, Stephen Redding and John

Van Reenen (1999) ‘Bridging the Productivity

Gap’, CentrePiece 4(3): 14-19.

Rachel Griffith, Stephen Redding and

John Van Reenen (2004) ‘Mapping

the Two Faces of R&D: Productivity Growth in

a Panel of OECD Industries’, Review of

Economics and Statistics 86(4): 883-95

(earlier version available here:


vanreenen/papers/wp0002.pdf).

Bronwyn Hall and John Van Reenen (2000)

‘Fiscal Incentives for R&D: A New Review of

the Evidence’, Research Policy 29: 449-69

(earlier version available here:


vanreenen/papers/w7098.pdf).

The ‘patentbox’ policywill wastearound £1 billion ayear at a timewhen thecountry can ill afford it

Big ideas

http://cep.lse.ac.uk/textonly/people/http://cep.lse.ac.uk/pubs/download/dp0675.pdfhttp://cep.lse.ac.uk/pubs/download/dp0675.pdfhttp://cep.lse.ac.uk/textonly/people/http://cep.lse.ac.uk/textonly/people/

In material terms at least, life today is much better than itwas in the past for almost everyone living in a developednation. Think of all the items we have now that we didnot have then – from the internet and mobile phones tomore trivial items such as apple cinnamon cheerios, afavourite American breakfast cereal.

These ‘new goods’ enrich our lives, and economists haveworked out just how much they are worth. They havefound that the internet is worth 2-3% of GDP, the mobilephone 0.5-1% and even apple cinnamon cheerios raisedthe welfare of Americans by 0.002% of GDP. Somepeople really didn’t like their previous breakfast cereal.

There are two stages in making a new good valuable tosociety. First, it has to be invented, and we rightlycelebrate inventors throughout the ages. But inventionsalso have to be adopted, and that means that they haveto offer good value to consumers. The people who comeup with ways of producing things more cheaply are alsoimportant in making us better off.

We have investigated the scale of two such innovations:mechanical cotton spinning and the motorcar assemblyline. Both led to sensational price declines and bothtransformed what had been luxury items for upper classconsumption – Indian calicoes and motorcars – into itemsof everyday consumption for a significant part of thepopulation. Workers on Ford’s Model T assembly linecould afford the cars they made; cotton spinners couldwear cotton shirts.

Henry Ford did not invent the motorcar. Nor was hisModel T a particularly good motorcar. Ford was not even the first to use a moving assembly line. But he wasthe first to ‘mass produce’ a car, a phrase he was also thefirst to use.

The effects were impressive: the time taken to assemble aFord chassis fell from just under 12.5 hours in the springof 1913 to 93 minutes a year later. Greater efficiency ledto big falls in price: the Model T cost $950 in 1909 and$360 in 1916, a fall in real terms of more than two thirds.Ford realised his aim of building a car ‘so low in price thatno man making a good salary will be unable to own one’.

Between 1908 and 1927 Ford sold a total of 15 millionModel Ts. It was so ubiquitous that in his 1945 novelCannery Row, John Steinbeck wrote that ‘Most of thebabies of the period were conceived in Model T Fords andnot a few were born in them.’ Ford also forced other carfirms to follow suit, so that between 1908 and 1923 theaverage price of a car fell from $2,126 to $317 in 1908terms. At the same time, annual sales rose from just64,000 to 3.6 million.

One way of working out what Ford’s process innovationwas worth to the American people is to calculate howmuch extra they would have had to pay to buy the carsthey did in fact buy, at the price prevailing prior to Ford’sinnovation. This measure is known as ‘social savings’. On that basis, Ford’s value to the American people was astaggering 14.7% of GDP.

Of course, we know that not everyone who bought a carat $317 in 1923 would have been willing to pay $2,126for it. We can estimate the shape of the demand curve,and on that basis Ford’s innovation was worth around1.8% of GDP. Although much smaller than our earliernumber, it still means that the average value of a Ford carto consumers was around twice the price they had to pay.Ford made himself rich and created lots of jobs, but mostof the benefits of his innovation went to the people whobought his cars.


6

Cotton and cars: the huge gainsfrom process innovation New inventions are good for economic growth, but equally important areimprovements in the way we make things – what’s known as process innovation.Tim Leunig and Joachim Voth measure the impact of two such innovations –mechanical cotton spinning and the motorcar assembly line – on the world’smaterial wellbeing.

in brief...

Mechanising the productionprocesses of car and cotton was asvaluable as inventing the internet

The same is true of mechanising cotton spinning in the Industrial Revolution. Again, the fall in price wasspectacular. Cotton yarn that had sold for 107 pence apound in 1784 sold for just under 13 pence in 1820.

The social savings from mechanising cotton yarnproduction were of a similar order of magnitude to thoseof mechanising car production – 17.6% of British GDP. Of that, 7.5% went to British consumers, while the other10.1% went to the people around the globe who werenow able to buy cotton goods more cheaply.

The consumer surplus estimate, this time simply for thecotton used by British people, was around 2.6%, again asubstantial number. As the famous historian AJP Taylor,himself born to a cotton family, once remarked, ‘Everypiece of cotton cloth is going to make someone warmeror cleaner or more comfortable’.

These two process innovations each produced gains forconsumers that were, even when estimated conservatively,equal to the expenditure on them. Furthermore, the gainstook relatively little time to be realised.

Improving the production processes of these two existinggoods was as valuable in terms of consumer welfare asinventing the internet, and much more valuable thaninventing mobile phones. Indeed, it is hard to imagine any

product that has been invented that has had a biggereffect on welfare, as quickly, as these two improvementsin the way in which we make things.

It is sometimes said – usually wrongly – that everythingworth inventing has been invented. But even if that werethe case, economic growth could and would continue.Innovators would strive successfully, and to great effect, toproduce existing things more cheaply. This would raise ourstandard of living, allowing us to buy more of the goodsthat already exist or to spend less time at work whilehaving the same material standard of living.

Those who make existing goods cheaper should becelebrated just as much as those who invent them in the first place. Both are crucial to understanding why wecan, today, live so much better than we were able to inthe past.


7

This article summarises ‘Spinning Welfare: The

Gains from Process Innovation in Cotton and Car

Production’ by Tim Leunig and Joachim Voth, CEP

Discussion Paper No. 1050 (http://cep.lse.ac.uk/

pubs/download/dp1050.pdf).

Tim Leunig is reader in economic history at LSE

and an associate in CEP’s research programme on

globalisation. Joachim Voth is ICREA research

professor in the economics department at

Universitat Pompeu Fabra, Barcelona.

Henry Ford madehimself rich, butmost of the benefitsof his innovationwent to the peoplewho bought his cars

People who makeexisting goods cheapershould be celebratedjust as much as thosewho invent them in thefirst place

http://cep.lse.ac.uk/pubs/download/dp1050.pdfhttp://cep.lse.ac.uk/pubs/download/dp1050.pdf


The heavy police presence at football matchesin England has reduced hooliganism in thestadium – but at what cost in terms of bothpolicing budgets and under-protected placeselsewhere in the neighbourhood? Olivier Marieexamines the multiple effects of footballmatches on crime.

Football matches:the effects on crime

In the 1980s, football in Englandbecame infamous for theaggressive behaviour of some ofits fans, when hooliganismtainted the ‘beautiful game’.

To combat this affliction, security atstadiums was greatly enhanced and thepolicing of matches drastically increased.As a result, crowd violence hadsignificantly subsided by the 1990s andthe hooligan stigma has all butdisappeared from English football today.

Still, the occasional scuffle isinevitable. While the police are nowmostly able to prevent these fromdegenerating, the physical proximity ofopposing fans has the potential to resultin violent crimes during home games. This possibility – which I call the‘concentration effect’ – is the first channelthrough which football matches can havean effect on crime.

In recent years, a different publicsafety debate has emerged, which isconcerned with the very high costs ofpolicing matches. These costs are onlypartially covered by the football clubsthemselves. In 2009, during a discussionof the costs of policing football by theHouse of Commons Home AffairsCommittee, David Winnick MP notedadditional potential costs:

‘If I were involved in criminality of amore sophisticated kind… would I notwork on the assumption that the policewill be fully occupied in a particular city –it will not be difficult to find out whenthese premiership games are being played– and I could go about my unlawfulbusiness?’

This question suggests the secondchannel through which matches can affect

crime – the ‘displacement effect’. Theremay be increases in violent crime andproperty crime away from stadiumsbecause of the displacement of policepersonnel assigned to match securityduring a home game.

A third possible effect of football oncrime may stem from the ‘self-incapacitation’ of some potentialoffenders. This supposes that among thethousands of fans attending or following a game, a not insignificant number ofthem would have been criminally active if they had not been at the match. Self-incapacitation could therefore lead todecreases in violent crime and propertycrimes during both home and awaygames, especially in neighbourhoodswhere a high proportion of the populationsupports the local team.

Football matches may thus affect localcrime rates through concentration,displacement and self-incapacitation – thedirections of the three potential effects aredescribed in Figure 1. For researchers, thedifficulty is to disentangle the impact ofeach of these effects as they occursimultaneously. One solution is to considerhow they influence property crime andviolent crime separately during home andaway games.

For example, to identify the self-incapacitation effect, we simply need toconsider changes in local crime rates in afootball team’s neighbourhood duringaway games. The displacement effect canthen be measured as the difference incrime rates during home and away gamesfor property offences. The concentration

effect only affects violent crime duringhome games but will be hard to measureprecisely. We would still be able toattribute increases in violent offencesduring home games as stemming from a mix of the displacement andconcentration effects.

To obtain measures of these effects,we use the Metropolitan Crime StatisticsSystem, high frequency local area crimedata available for London for 1994-97.The system contains information on thetime, location and type of offence for allcrimes recorded by the police in the capitalduring this period. The data areaggregated at the level of the 32 boroughs in London into four six-hourwindows for property and violent crimes.

This is matched to detailed gameinformation for nine major football teamswith grounds in seven different Londonboroughs (in parenthesis): Arsenal(Islington), Charlton Athletic (Greenwich),


9

Note: Upward and downward pointing arrows represent positive and negative effects througheach of the three channels – concentration, displacement and self-incapacitation – through whichhome or away sporting events may affect local property crime and violent crime. Flat arrowssuggest no expected effect.

Figure 1:

Concentration, displacement and incapacitationeffects on property crime and violent crime duringhome and away games

Property crime Violent crime

Home Away Home Away

Concentration 3 3 1 3

Displacement 1 3 1 3

Self-incapacitation 5 5 5 5

During homematches,

opportunisticoffenders in

under-protectedlocal areas

commit moreproperty crimes


10

This article summarises ‘Police and Thieves

in the Stadium: Measuring the (Multiple)

Effects of Football Matches on Crime’

by Olivier Marie, CEP Discussion Paper

No. 1012 (http://cep.lse.ac.uk/pubs/download/

dp1012.pdf).

Olivier Marie of the University of Maastricht

is a research associate in CEP’s education and

skills programme.

Further reading

Mirko Draca, Stephen Machin and Robert

Witt (2008) ‘Panic on the Streets of London:

Police, Crime and the July 2005 Terror

Attacks’, CEP Discussion Paper No. 852

(http://cep.lse.ac.uk/pubs/download/

dp0852.pdf).

Violent crime in the localcommunity does not increase on match daysexcept during a local derby

Chelsea (Hammersmith and Fulham),Crystal Palace (Croydon), Millwall(Lewisham), QPR (Hammersmith andFulham), Tottenham Hotspur (Haringey),West Ham United (Newham) andWimbledon (Croydon).

For each match, we know the kick-offtime, the attendance, the type of game(league or cup), the result including goaldifference, the number of red and yellowcards issued and whether it is a local derby(that is, when London teams play eachother). There are 571 home and 576 awaymatches in the period for which the crimedata are available.

To identify a match effect on localcriminal activity, we exploit the variation inlocation and timing of both home andaway games. We focus on the impact oflarge variations in attendance, controllingfor weather conditions and whether thegame is played on a holiday. We also netout the possible influence of othermatches taking place at the same time,the distance of each borough to thestadium hosting a home game and thedistance of each away match.

Our results suggest that the level ofproperty crime falls by roughly 3% forevery extra 10,000 supporters attendingan away game. In accordance with ourconceptual framework (Figure 1), weinterpret the fall in property offencesduring away matches as a pure self-incapacitation effect.

But during a home game, propertycrime rises by 4% for every extra 10,000supporters. This suggests that there is animportant police displacement effect asopportunistic offenders in the under-protected areas of a borough takeadvantage of the smaller probability ofdetection to commit property crimes.

We find no measurable impact onviolent crime in the local communityexcept during a local derby. This suggestssome effect of concentration duringmatches that are reputedly the ones withthe highest levels of animosity betweenrival fans.

The overall conclusion from ourresearch is that the displacement of policeforces during football matches increasesproperty crime by almost 7% for everyextra 10,000 fans attending a game. Thisis in line with a growing body of evidenceshowing that police presence has animportant effect on reducing crime,including CEP research on the impact

on crime of extra policing following the 7/7 terror attacks in London (Draca et al, 2008).

The research also contributes to thedebate about the impact of policing ofprivate sporting events and the cost interms of the local community’s wellbeing.This is especially relevant after August’sriots and looting in London, which raisedquestions about the ability of an alreadystretched police force to maintain law andorder when the capital hosts the OlympicGames next year.

http://cep.lse.ac.uk/pubs/download/dp1012.pdfhttp://cep.lse.ac.uk/pubs/download/dp1012.pdfhttp://cep.lse.ac.uk/pubs/download/dp0852.pdfhttp://cep.lse.ac.uk/pubs/download/dp0852.pdf


11

Starting out in working lifeis getting harder for youngpeople all over Europe.Accessing the high-qualitytraining that eases thetransition and ensures

good career prospects is a particularstruggle for young people with below-average school attainments. To understandand act to improve the outcomes for theseyoung people, European governmentsneed good information on which to basepolicy decisions.

Statistics collected by nationalgovernments underpin such decisions and

these ‘headline numbers’ help to drivepolicy debate. They can also send apowerful message to young people andtheir parents about the education andtraining routes that lead to good labourmarket transitions.

Recognising the role played byheadline numbers in shaping policy andaspiration, the official French governmentadvisory body on education – the HautComité d’Education (HCE) – commissionedEric Verdier and me to coordinate acomparative study of measures ofeducation outcomes and labour markettransitions of the least well-qualified group

England’s most widely used indicator of youngpeople’s education and labour market status isthe NEET category – ‘not in education,employment or training’. Making comparisonswith how France and Germany measure school leavers’ progression and achievement,Hilary Steedman argues that NEET is nolonger good enough.

Young people without qualifications:how ‘headline numbers’ shape policy and aspiration


12

of young people. France, Germany andEngland were selected for in-depth study.

Our study reveals both markeddifferences between countries in how the‘at risk’ group of low-attaining schoolleavers is analysed statistically and thechoice of variables on which to base thestatistical description of school to worktransitions. But it also reveals someimportant convergences.

France has moved away from a strictlyeducation-based classification of school-leavers – traditionally a combination of thefinal class completed and the highestnationally recognised school diplomaobtained. The revised classification ofschool-leavers derives from greatercommitment to the importance of a post-16 diploma for all – awarded on success ina two or three year vocational or generalcourse of education and/or training.

French statistics now distinguishbetween ‘early leavers’ (sans qualification)who have no recognised certification (9%of 20-24 year olds in 2009); those (sansdiplômes) who complete a post-16 two orthree year course but do not pass theexaminations for a full certificate ordiploma (a further 9%); and those (82%) who complete and pass successfully (diplômés).

The employment rates for these three groups of young people arepublished annually. They tell a powerfulstory both to policy-makers and youngpeople about the vital importance ofgaining a full vocational certificate orgeneral diploma or, failing that, at leastcompleting the course.

The message for teachers, parents andschool students is clear: a full vocationalqualification reduces the probability of

unemployment by almost two thirdsrelative to early leavers; and achieving the Bac diploma (the French equivalent ofA-levels) reduces it by a factor of four.Even staying the full two or three year vocational course without passingthe examination is beneficial foremployment chances.

France relies, perhaps excessively, onschool-based nationally standardisedvocational and general education forlabour market preparation. This provision has inadequate capacity to reactto rapidly changing labour marketdemands. French apprenticeshipsovercome this problem but they canprovide for only a small proportion ofthose in vocational education.

Nevertheless, the emphasis onsuccessfully completing a two or threeyear course of vocational or general studyat school produces a young population inwhich two thirds hold a Bac diploma anda further 20% have two or three years ofvocational training.

The contrast with Germany could not

be greater. Well over half of all schoolleavers expect to enter apprenticeship.School marks are of importance for these16 year old German school leavers onlybecause higher marks increase thelikelihood of gaining an apprenticeshipplace.

Without the labour market recognitionthat the apprenticeship qualificationconfers, the employment outlook forGerman school leavers is grim. ‘Notqualified’ in Germany equates to ‘not trained through apprenticeship’. German labour market transition statisticsare overwhelmingly devoted to analysingand understanding access toapprenticeship places.

For almost 20 years, following theshock to the economy of the reunificationof Germany in 1990, German states andthe federal government struggled to meetthe demand for apprenticeship places fromyoung people. Many with low level schoolqualifications faced long ‘waiting periods’spent in special measures and some nevergained a place.

In France and Germany,

the headlinenumber

on schoolleavers’

transition to work

emphasisesprogression

The most high profile measure of theGerman government’s competence ineducation policy was and is the annual‘matching’ of apprenticeship applicants toplaces. Only in 2010 were apprenticeshipplaces close to matching demand fromapplicants. In 2009, 15% of 20-29 yearolds had not received apprenticeshiptraining and consequently faced poorlabour market outcomes. The remaining85% had completed either a three yearapprenticeship or three years of full-timeeducation post-16.

Education statistics published inEngland do not identify a ‘noqualifications’ category of young people,as in France, nor a group withoutrecognised occupational training, as inGermany. The most widely used indicatorof young people’s education and labourmarket status is the NEET category – ‘notin education, employment or training’.

Yet NEET young people are not a ‘noqualifications’ category. Nearly threequarters of the 15% of the age groupclassified as NEET in 2010 had some GCSE passes. Of a smaller group – thosewho have been NEET for more than 12 months – over half had some GCSEpasses (see Figure 1).

The difference from France andGermany is the emphasis in England onparticipation regardless of content or value– any job, any training, any courseremoves a young person from the ‘at risk’category. In both France and Germany, themessage from the headline number onschool leavers’ transition to work is theimportance of progression andachievement – progression to a recognisedqualification through an academic orvocational pathway in France and

progression through apprenticeship inGermany.

These measures help to set clear goalsfor 16 year old school leavers, clear goalsthat are lacking in England. In contrast toFrance and Germany where 85% reachLevel 3 equivalence, only half of all 19year olds in England achieve any Level 3qualification. Could these differences ingoal-setting account for this failure?

According to Alison Wolf’sindependent review of vocationaleducation commissioned by the UK’sDepartment for Education and publishedearlier this year, ‘at least 350,000 youngpeople in a given 16-19 cohort are poorlyserved by current arrangements. Theirprogrammes and experiences fail topromote progression into either stable

paid employment or higher level educationand training’.

As the Department for Education setsabout implementing the Wolfrecommendations for 16-19 educationand training, the negative message of the NEET category needs to be replaced bya headline number that reflects andprojects the more ambitious goals thatWolf proposes.

This article is based on Les élèves sans

qualification: La France et les pays de

l’OCDE by Ekaterina Melnik, Martine Möbus,

Noémie Olympio, Hilary Steedman, Rémi

Tréhin-Lalanne and Eric Verdier, a study

commissioned by the HCE and coordinated

by Eric Verdier and Hilary Steedman

(http://www.hce.education.fr/gallery_files/

site/21/106.pdf).

Professor Eric Verdier is at the University

of Aix-Marseille. Dr Hilary Steedman was a

senior research fellow at CEP from 1994 to

2004 and is currently an associate in the

education and skills programme.

0%

10%

20%

30%

40%

50%

>12 months

1-12 months

None reported

1-4 GCSE grades

D-G

5+ GCSE grades

D-G

1-4 GCSE grades

A*-C

5-7 GCSE grades

A*-C

8+ GCSE grades

A*-C

■ 1-12 months■ More than 12 months


13

The negative message ofNEET should be replacedby a headline numberreflecting more ambitious goals for young peoplewithout qualifications

Source: DFE Statistical Bulletin B, 1 July 2010

Figure 1:

NEET status by age 18 by highest qualification at age 16

In England, any job, anytraining, any course

removes a young personfrom the ‘at risk’ category

Per

cen

tage

s of

NE

ET

s

http://www.hce.education.fr/gallery_files/site/21/106.pdfhttp://www.hce.education.fr/gallery_files/site/21/106.pdf


14

House prices and school quality:evidence from state and private education in Paris

How much are parents willing to pay to get their childreninto what they perceive to be a better school? The answerto this question has important implications both in termsof school admission policies and urban planning.

Residence-based school admission rules often aim atensuring a degree of social diversity in schools. But ahigher willingness to pay for better schools implies thatresidential choices are more heavily influenced by theperceived quality of local schools. Contrary to the originalobjective, this phenomenon will tend to reinforce the levelof social segregation between schools and undermine theeffectiveness of school catchment areas as a policy tool toreduce educational inequalities.

If parents care about the quality of education and ifchildren are assigned to schools based on their place ofresidence, then differences in school performance shouldbe at least partly captured in house prices. The housingmarket therefore provides an indirect way of measuringthe average willingness to pay for better schools.

Following this intuition and analysing data from the UKand the United States, empirical studies (including work byCEP’s Steve Gibbons and Stephen Machin) have foundthat there is often a significant house price premiumattached to an increase in the average test scores of local schools.

A limitation of these studies is that they restrict theanalysis to state schools, ignoring the influence thatprivate schools might have on residential choices. Sinceadmission to private schools is not usually subject to aresidence requirement, private education provides anoutside option for parents who are not satisfied with thequality of the neighbouring state schools. It seems likelythat the willingness of parents to pay for better stateschools in a particular neighbourhood is smaller when thenumber of local private schools is higher.

Our study explores this issue by evaluating how house pricesreact to the quality of education offered by local schools,both state and private. We analyse comprehensive data on

middle schools (those for pupils aged 11 to 14) and housingmarket transactions in Paris over the period 1997-2004.

The organisation of secondary schooling in Paris is a usefulcontext for analysing how the interaction between stateand private schools influences residential choices becauseit combines a residence-based system of assignment tostate schools with a well-developed and almost entirelypublicly funded network of private schools.

Parents who reside in a given neighbourhood have hardlyany control over the choice of their child's state schoolbecause each catchment area contains only one schooland because very few dispensations are granted toapplicants who live outside the catchment area. In thiscontext, parents who want to avoid sending their childrento a low-performing local state school can either move toanother area or seek admission to a private school.

What makes the latter option attractive is that in additionto being entitled to select their pupils from anywhere inthe city, private schools usually charge relatively modestfees (between £1,000 and £2,000 a year). As a result,private schooling accounts for about a third of totalsecondary education enrolment in Paris.

Empirically, the issue of estimating the impact of stateschool performance on house prices is a challenge forresearchers. As in most big cities, the best state schools inParis tend to be located in the most expensive areas. Thisfeature, however, does not necessarily imply that housesare more expensive because local schools are better.

In fact, the causality might go the other way round.Wealthy neighbourhoods attract households of high socio-economic status whose children tend to perform better atschool than those in more deprived areas. This simplemechanism induces a positive correlation betweenapparent school quality and house prices, even if parentsdo not take account of the performance of local schoolswhen deciding to move into a neighbourhood.

Our research addresses the causality challenge by

It is now widely understood that the quality of state schools in aneighbourhood has an impact on local house prices. Analysing datafor Paris, Gabrielle Fack and Julien Grenet have looked deeper intothis link by exploring how the presence of private schools influencesparents’ willingness to pay to live near good state schools.

in brief...

15

comparing the value of house sales across middle schoolcatchment area boundaries – that is, sales of houseslocated close to each other but assigned to different stateschools.

The key justification for this approach is that on average,flats located in the immediate vicinity of schoolattendance boundaries have similar attributes (period ofconstruction, number of rooms, etc.) and, by definition,belong to the same neighbourhood. The only significantdifference between them is that they are assigned todifferent state schools. Hence differences in cross-boundary house prices can be interpreted as a directmeasure of the unequal valuation of the correspondingstate schools.

Based on this empirical strategy, we find that theperceived quality of state middle schools has a significantimpact on the house prices. Using various measures ofschool performance, we find that differences in schoolquality translate into differences in house prices of asimilar magnitude to those found in the UK and theUnited States. Our analysis suggests that roughly 5% oflocal variations in house prices in Paris are explained bydifferences in the perceived quality of state schools.

We also show that the impact of state middle schools onthe housing market varies with the local density of privatemiddle schools. The effect of school quality is morepronounced for residences in areas with a low density ofprivate schools and is non-existent for areas with thehighest density of private schools. Our results suggest thatin areas where private schools are scarce, the capitalisation

of state school performance in housing prices is in thesame order of magnitude as the individual cost of aprivate school four-year tuition fee in areas where they arenumerous, that is about €4,000 at the average flat price.These findings are consistent with the idea that byoffering a relatively cheap outside option to parents whoare willing to avoid the constraints of strict catchmentareas, private schools tend to mitigate the impact of stateschool performance on house prices.

In terms of policy implications, our results suggest that the co-existence of state and private schools is animportant dimension to take into account when designingschool admission policies. In particular, the effect ofalternative state school admission rules (strict residence-based assignment, relaxed school zoning, open enrolment,etc.) on residential and school segregation, pupilperformance and educational inequalities dependscrucially on how the housing market incorporates stateand private school performance.

This article summarises ‘When do Better Schools

Raise Housing Prices? Evidence from Paris

Public and Private Schools’ by Gabrielle Fack

and Julien Grenet, Centre for the Economics of

Education Discussion Paper No. 119

(http://cee.lse.ac.uk/cee%20dps/

ceedp119.pdf).

Gabrielle Fack is at Universitat Pompeu Fabra

in Barcelona. Julien Grenet of the Paris School

of Economics is a research associate in CEP’s

education and skills programme.

As elsewhere, Parisianhouse prices are

influenced by theperceived quality of

local state schools

http://cee.lse.ac.uk/cee%20dps/ceedp119.pdfhttp://cee.lse.ac.uk/cee%20dps/ceedp119.pdf


Growing numbers of university students in Britain and the United States are staying on aftertheir first degrees to invest in a postgraduatequalification. Joanne Lindley and StephenMachin document this trend and assess theimpact on wage inequality – among graduates and across the labour force as a whole.

Mass participation inhigher education hasaltered the typical pathfollowed by universitygraduates. The norm

used to be that after obtaining anundergraduate degree, people wouldfinish their studies and enter the labourmarket. These days, many more studentsstay on to invest in postgraduateeducation. Indeed, by 2009 just over 10%of the workforce in Britain and the UnitedStates – and more than a third of allgraduates – had a postgraduatequalification.

We have documented these trends inpostgraduate education and how theyrelate to rising wage inequality in the twocountries. It is now widely understood that

despite rapidly growing numbers ofuniversity-educated workers, increasedrelative demand for their skills has been akey driver of overall wage inequality. Ourresearch reveals that the changingcomposition of the graduate labour force– and widening wage differentials withinthis group – has also been a key feature ofrising inequality.

Figure 1 shows changes in theproportions of all graduates andpostgraduates in the labour force duringthe past 30 years, as well as the changingshare of postgraduates among allgraduates. In the United States, thegraduate share of employment hasincreased steadily, rising from 24% in1980 to 36% in 2009. In Britain, thegraduate share doubled between 1996

and 2009, going from 14.5% to 29%.Focusing on postgraduates, there has

been a sharp increase in both countries: inthe United States from 7.5% to 13%between 1980 and 2009; and in Britainfrom 4% to 11% between 1996 and2009. The more rapid increase in the shareof postgraduates compared with college-only workers means that the postgraduateshare among graduates has increased inboth countries: from 31% to 35% of USgraduates between 1980 and 2009; andfrom 30% to 37% of British graduatesbetween 1996 and 2009.

At the same time as postgraduatesincreased their employment share, theirrelative wages also rose. Figure 2 showshow the wage differential between thosewith a postgraduate qualification and

The boom in postgraduate educationand its impact on wage inequality


0

5%

10%

15%

20%

25%

30%

35%

40%

Postgraduate Employment Share

Graduate Employment Share

Share of Postgraduates in all Graduates

2009200019901980

Postgraduate Employment Share

Graduate Employment Share

Share of Postgraduates in all Graduates

200920001996

■ Graduate employment share■ Postgraduate employment share■ Share of postgraduates in all graduates

United States BritainSource: US Current Population Surveys and Labour Force Surveys in Britain – employmentshares are defined for people in work with 0-39 years of potential experience and aged 26-60.

Per

cen

tage

Figure 1:

Employment shares of all graduates and postgraduates

30%

25%

20%

15%

10%

5%

1980 1985 1990 1995 2000 2005 2010

■ United States■ Britain

Source: US Current Population Surveys and Labour Force Surveys in Britain –employment shares are defined for people in work with 0-39 years of potentialexperience and aged 26-60. Percentage wage differentials are calculatedcontrolling for gender, experience, region and ethnicity.

Per

cen

tage

Figure 2:

Trends in postgraduate/college-only percentagewage differentials

those with just an undergraduate degreehas increased through time.

It is evident that postgraduates havesignificantly strengthened their relativewage position in both countries. In theUnited States, the postgraduate/college-only wage differential has risen sharplyover time, more than doubling fromaround 14% in 1980 to just over 30% by 2009. In Britain, the postgraduate/college-only gap is lower but it has risenfrom 6% in 1996 to 13% by 2009.

So it seems that the relative labourmarket fortunes of postgraduate andcollege-only workers have evolveddifferently through time. The clear patternthat emerges in the two countries is of anincrease in both the employment sharesand wage differentials for postgraduatesvis-à-vis college-only workers. Risingsupply coupled with rising relative wages means that relative demand seemsto have shifted over time in favour ofpostgraduate workers compared withcollege-only workers.

Previous research has connected therelative demand shifts for differenteducation groups that have underpinnedincreased wage inequality to measures of technological change. The usualapproach is to relate the two in terms ofchanges across industries through time.This work reports that measures oftechnology – such as R&D, innovation,computer usage and investment incomputers – have been strongly correlatedwith the increased demand for moreeducated workers.

Our research also considers this,looking at shifts in labour demandseparately for postgraduate and college-only workers, and making comparisonsbetween these two groups of graduatesand with other workers without anyuniversity education.


18

It turns out that there is a strongerconnection between increases in therelative demand for postgraduates andmeasures of technological change than forcollege-only graduates. Analysis ofchanges in employment shares andchanges in computer usage in 215 USindustries and 51 British industries showsthat, for both countries, there is only apositive correlation for postgraduates.

That shifts in labour demand towardspostgraduates seem to be (at least in part)driven by technological change is alsosupported by cross-country patterns ofchanging labour demand and technology.The analysis shows that bigger shifts indemand occurred in the same industries inBritain and the United States and that thechanges in computer usage are very muchconcentrated in the same industries forthe two countries.

More evidence that employers areincreasingly demanding postgraduates canbe seen by comparing the skill sets

required by the jobs of postgraduate andcollege-only workers. Table 1 showspostgraduate/college-only differences intheir estimates of cognitive skills, problem-solving skills, people skills, firm-specificskills, the tasks they use computers forand how routine their jobs are.

It is clear that both sets of graduatesdo jobs with high skill and job taskrequirements. But in key skills areas, thelevels are significantly higher forpostgraduates. For example,postgraduates have higher numeracylevels (especially advanced numeracy),higher levels of analysing complexproblems and more specialist knowledgeor understanding.

Breakdowns of computer usage arealso striking. Postgraduate and college-only workers both report high levels ofcomputer usage, but using computers toperform complex tasks is markedly higheramong the postgraduate group.

So it seems that postgraduates possessdifferent skills and do jobs involvingdifferent (usually more complex) tasks thancollege-only workers. This is in line withthe finding that relative demand hasshifted faster in favour of thepostgraduate group, and it appears to bean important aspect of rising wageinequality among college graduates.

We have also compared theoccupations of postgraduate and college-only workers. With more recent data, wecan consider different forms ofpostgraduate degree. Table 2 shows thetop five occupations in terms of theirshare in employment for college-onlyworkers and postgraduates with master’sdegrees and doctorate degrees.

There are several notable features ofthe top five occupations of these threegroups of workers. First, the top five tendto be different occupations in the twocountries. Second, while the occupationalcategories are not quite the same acrosscountries, there are some clear similarities.

Third, the postgraduate occupationsare more segregated than the college-onlyoccupations. For postgraduates, in theUnited States the top five (out of 497occupations) account for almost half ofemployment (49%) and in Britain the topfive (out of 353 occupations) account foraround 45%. The college-only distributionis a lot more dispersed, with the top fiveaccounting for only 16% of employmentin the United States and 20% in Britain.

Table 1:

Skills and job tasks suggesting that postgraduates inBritain are more in demand than college-only graduates

Skill/job task Postgraduates College-only

Cognitive skillsLiteracy 4.067 3.763

Simple numeracy (basic arithmetic) 3.606 3.583

Advanced numeracy (maths and statistics) 3.004 2.715

Problem-solving skillsThinking of solutions to problems 4.311 4.277

Analysing complex problems 4.179 3.880

People skillsMaking speeches/presentations 3.658 3.148

Teaching people 4.023 3.843

Dealing with people 4.658 4.684

Firm-specific skillsKnowledge of products/services 3.817 3.831

Specialist knowledge or understanding 4.704 4.548

Computer usageUsing a computer or computerised equipment 4.607 4.384

Proportion that do not use a computer 1.9% 4.5%

Simple (general purpose) computer users 7.4% 10.9%

Moderate computer users 42.8% 48.6%

Complex computer users 47.9% 36.1%

Routine nature of jobPerforming short repetitive tasks 2.689 2.890

Variety in job 4.315 4.195

Sample size 257 1,095

Source: The 2006 Skills Survey. With the exception of the proportions using computers, thenumbers are based on a scale of 1-5 for questions on task performance – ‘How important is thistask in performing your current job?’ – for which the choices are 1 ‘not at all important’, 2 ‘notvery important’, 3 ‘fairly important’, 4 ‘very important’ or 5 ‘essential’.

More than a third ofgraduates inBritain and theUnited Statesnow have apostgraduatequalification


19

Overall, our findings on increasingdivergences within the group of workerswho go to university offer new evidenceon how the changing education structureof the workforce has contributed to risinginequality. Our focus is on increasingdivergences within the group of workerswho go to university.

We document that there have beensignificant increases in the number ofworkers with a postgraduate qualificationand that, at the same time as this increasein their relative supply, their relative wageshave risen strongly compared with workerswith only a college degree. Trend increasesin the relative demand for postgraduateshave acted as a key driver of increasingwithin-graduate inequality and of overallrises in inequality.

The relative demand shifts in favour ofworkers with postgraduate qualificationsare strongly correlated with technicalchange as measured by computer usageand investment. It turns out that over the

This article summarises ‘Rising Wage

Inequality and Postgraduate Education’

by Joanne Lindley and Stephen Machin,

CEP Discussion Paper No. 1075

(http://cep.lse.ac.uk/pubs/download/dp1075.pdf).

Joanne Lindley is a senior lecturer in

economics at the University of Surrey.

Stephen Machin is a professor of economics

at University College London and CEP’s

research director.

Table 2:

Top five occupations – college-only, master’s degrees and doctorates

United States, 2010, 497 detailed occupationsCOLLEGE-ONLY MASTER’S DEGREE DOCTORAL DEGREE

Top five Employment Top five Employment Top five Employment

occupations share occupations share occupations share

1 Elementary and middle 4.6% 1 Elementary and middle 11% 1 Post-secondary teachers 21%

school teachers school teachers

2 Managers, all other 3.6% 2 Secondary school teachers 4.7% 2 Physicians and surgeons 10.7%

3 Accountants and auditors 3.3% 3 Managers, all other 4.3% 3 Lawyers, judges, magistrates 10.2%

and other judicial

4 Chief executives 2.3% 4 Post-secondary teachers 3.6% 4 Psychologists 3.7%

5 First-line supervisors/managers of 2.2% 5 Education administrators 3.1% 5 Pharmacists 3.6%

retail sales workers

Britain, 2010, 353 detailed occupationsCOLLEGE-ONLY MASTER’S DEGREE DOCTORAL DEGREE

Top five Employment Top five Employment Top five Employment

occupations share occupations share occupations share

1 Primary and nursery education 5.1% 1 Secondary education 4.8% 1 Higher education 20.7%

teaching professionals teaching professionals teaching professionals

2 Marketing and sales managers 4.5% 2 Software professionals 3.6% 2 Medical practitioners 11%

3 Nurses 3.6% 3 Marketing and sales managers 3.5% 3 Bioscientists and biochemists 7.1%

4 Software professionals 3.2% 4 Management consultants, actuaries, 3.2% 4 Researchers 3.2%

economists and statisticians

5 Information and communications 3.1% 5 Information and communications 3.1% 5 Software professionals 3.1%

technology managers technology managers

years as computer use has become morewidespread in most workplaces, theprincipal beneficiaries of this revolutionhave not been all graduates, but thosewith postgraduate qualifications. As such,there has been a strong connectionbetween the increased presence ofpostgraduate workers in the labour forceand rising wage inequality over time.

The principal beneficiaries of thecomputer revolution have not been

all graduates but those withpostgraduate qualifications

http://cep.lse.ac.uk/pubs/download/dp1075.pdf


20

Why do teachers inSwitzerland earnfour times whatteachers in Israelearn? Why are

teachers in South Korea paid at the 78thpercentile of their country’s incomedistribution whereas those in United Statesare paid at only the 49th percentile? Anddo these massive variations in the waydifferent countries treat their teachersmatter for the outcomes of their pupils?

Answers to these questions are at theheart of the educational policy debate andwe can learn a lot about the relationshipbetween teacher quality and pupiloutcomes from cross-nationalcomparisons. The issue is especiallyrelevant in the context of pressures toreduce public spending. Most countriesdevote a sizeable proportion of theirbudgets to education – and around 70%of that money goes on teacher salaries.

Our research considers thedeterminants of teacher salaries acrossOECD countries and examines therelationship between the real and relativelevels of teacher remuneration and themeasured performance of secondaryschool pupils over the last 15 years.

There are two potential explanationsas to why teachers’ pay may be causallylinked to pupil outcomes. The first is thathigher pay will attract more ablegraduates into the profession. As thepotential supply of teachers rises becauseof the higher pay on offer, entry intoteaching as a profession will become morecompetitive. This in turn will mean thatthe average ability of those entering thejob will rise. Once recruited, higher relativepay and/or more performance-related paymay provide teachers with strongerincentives to improve their pupils’educational outcomes.

The second mechanism is more subtle

If you pay peanuts, do you get monkeys? Ifteachers were better paid and higher up thenational income distribution, would there be an improvement in pupil performance? Peter Dolton and Oscar Marcenaro-Gutierrezexamine the enormous variation in teachers’pay across OECD countries and its significance for educational outcomes.

Teachers’pay and pupil performance


21

– namely that improving teachers’ payimproves their standing in a country’sincome distribution and hence thenational status of teaching as a profession.As a result of this higher status, moreyoung people will want to becometeachers. This in turn makes teaching amore selective profession and hencefacilitates the recruitment of more ableindividuals.

Higher status and higher pay areinvariably linked but the two can provideseparate driving forces to engineer betterrecruits to the profession. The keyhypothesis is that better pay for teacherswill attract higher quality graduates into the profession and that this willimprove pupil performance.

The most comprehensive sources ofcomparative information about teachers indifferent countries are the OECD’s annual‘Education at a Glance’ reports. Thesepublications provide information onstarting salaries, salaries after 15 years of teaching experience and salaries at thetop of the profession.

The relative supply of teachers in acountry is measured by the number ofteachers as a fraction of the labour forceand the pupil/teacher ratio in theeducation system. An additional supplyfactor relates to the proportion of thestock of teachers who are women. Wealso control for the number of teachinghours supplied, since having a lowernumber of teachers can be compensatedfor by them working more hours.

We measure the nature of a country’sinvestment in education by the level ofeducational expenditure as a fraction ofGDP, controlling for the rate at which acountry is growing, since clearly this willconstrain its possible investments ineducation. The changing nature of the

increase in pupil performance. Likewise, a5% increase in the relative position ofteachers in the income distribution wouldincrease pupil performance by around 5-10%.

What are the policy implications ofthese findings? Most obviously, if agovernment is concerned with educationaloutcomes, then it should be aware thatthe quality of its teachers is offundamental importance. We suggest thatthe route to hiring teachers from higherup the ability distribution is to pay them ata higher point in the country’s incomedistribution.

How could this be achieved? A country with a stock of low qualityteachers cannot simply raise the pay of allteachers immediately and expect thequality of teaching to improve. Theexisting stock of teachers would clearlyhave an incentive to appropriate theseeconomic rents with no responsibility to

Better pay forteachers willattract higherquality graduatesinto theprofession andimprove pupilperformance

demand for teacher services is measuredby the demographic growth in the size ofthe population of school age.

To examine the relationship betweenteacher remuneration and educationalattainment, we use the internationallycomparable results from the OECDProgramme for International StudentAssessment (2000, 2003 and 2006) andTrends in International Mathematics andScience Study (1995, 1999 and 2003).

Figure 1 provides an insight into therelationship between teacher salaries andpupil outcomes, showing a clear statisticalassociation between higher relativeteachers’ pay and higher standardisedpupil scores across countries.

Our research with aggregate countrydata supports the hypothesis that higherpay leads to improved pupil performance.As an indication of the relative size of thiseffect, we find that a 10% increase inteachers’ pay would give rise to a 5-10%

0

20

40

60

80

100

>12 months

1-12 months

$60,000$50,000$40,000$30,000$20,000$10,000

● Turkey● Israel ● Greece

● Italy● Portugal

● Hungary● Iceland

● Czech Republic● Sweden

● Austria● UK

● Denmark

● Germany● Ireland● Belgium

● Japan● Australia

● New Zealand● Finland

● Korea● Netherlands

● Switzerland

● France● Norway

● USA● Spain

Sco

re p

erce

nti

les

Teachers’ salaries in US dollars at purchasing power parity

Figure 1:

Pupil score percentile as a function of teachersalaries after 15 years of experience

A 5% increase in the relative position of

teachers in the incomedistribution would increase pupil

performance by around 5-10%


22

become better teachers. And while thequality of new recruits to the professionwould rise as a result of this upward shiftin relative pay, it would take a long time –30 or so years – to change the quality ofthe whole stock of teachers.

The answer then must be to considerhow teacher quality can be raisedgradually. If the government were toratchet up starting pay, this would securebetter quality new teachers. But improvingthe stock of existing teachers wouldrequire continued professionaldevelopment and in-service training and/orattempting to fire the worst teachers.

Such policy measures are not withinthe scope of this study, but there is awealth of research evidence about themas possible remedies to improve theexisting stock of teachers. One solution isto provide an incentive mechanism forexisting teachers to improve quality bypaying them according to the percentileperformance (in value added terms) oftheir pupils. Another possible solution is toincrease the rate at which teachers’ payrises with their level of experience.

Another dimension of the problem is the time scale over which anyimprovement in pupil outcomes is sought. If replacing existing teachers withones of higher quality would take toolong, then a quicker fix might be toreduce the pupil/teacher ratio or increasepupil contact hours by simply employingmore teachers from the pool of inactive teachers.

Our analysis finds a clear trade-offbetween pupil/teacher ratios and teachers’pay across countries – that is, countries donot necessarily have to pay higher salariesto secure better pupil outcomes. But if acountry is not prepared to pay teachersrelatively well, then it will have to go along way down the road of reducing classsizes to compensate them – in short,governments and educationaladministrators need to know that there is‘no free lunch’ here.

The policy implications of our findingsare relevant to the recruitment of teachersand the improvement of educationalstandards. The link we find betweenteacher quality and high educationalstandards has logical implications for anygovernment’s commitment to recruit,retain and reward good teachers. In thisregard, it seems that increasing teachersalaries (and the speed at which they can

reach higher pay levels within a particularpay structure) will help schools to recruitand retain the higher ability teachers thatschools need to offer all pupils a high-quality education.

At a wider policy level, improvementsin education appear to be a commonfactor behind economic growth in recent decades in all OECD countries. The increase in human capital accountedfor more than half an extra percentagepoint of growth in the 1990s comparedwith the previous decade. One clear way to improve the stock ofhuman capital is to invest in higherquality teachers.

This article summarises ‘If You Pay Peanuts,

Do You Get Monkeys? A Cross-country

Analysis of Teacher Pay and Pupil

Performance’ by Peter Dolton and

Oscar Marcenaro-Gutierrez, Economic

Policy 26(65): 5-55, January 2011

(http://onlinelibrary.wiley.com/doi/10.1111/

j.1468-0327.2010.00257.x/full).

Peter Dolton is a professor of economics at

Royal Holloway College, University of London,

and a senior research fellow in CEP’s

education and skills programme. Oscar

Marcenaro-Gutierrez is at the University

of Malaga.

There is a cleartrade-off between

pupil/teacherratios and

teachers’ payacross countries

http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0327.2010.00257.x/fullhttp://onlinelibrary.wiley.com/doi/10.1111/j.1468-0327.2010.00257.x/full


23

Regions, cities andneighbourhoods acrossBritain seem to be veryunequal. This is true if welook at average earnings,

employment, education and almost anyother socio-economic outcome.

Take ‘gross value added’ (GVA) perperson, potentially a good indicator ofliving standards in different places. In 2005(a point in time chosen to be sometimebefore the peak of the boom), the highestranked regions in Britain were West InnerLondon and Berkshire with GVAs of£44,050 and £39,850 respectively. Thelowest ranked were Liverpool andBlackpool with GVAs half those in theSouth East: £19,800 and £21,050.

These examples are representative of abroader trend: the top ranked 10% ofregions have GVA at least 50% higher thanthe bottom ranked 10%.

Regional policy, urban policy, evenneighbourhood policy are all largely basedon concerns about these kinds ofdisparities. But these figures are simplyaggregates of the outcomes for peoplewho live and work in these places. Withoutfurther information, we do not knowwhether the outcomes for people currentlyliving in Manchester would be any differentif they lived and worked in London.

We also have no way of knowing if theproductivity of London and Manchesterwould change if these movements ofpeople actually took place. Similarly, we donot know whether replicating theeconomic, policy, institutional andenvironmental regime of London inManchester would change anythingwithout moving people. In short, it is hardto work out what these differences mean interms of the economic advantages anddisadvantages that a place offers to thepeople who live and work there.

It is also easy to assume from looking atthese aggregated figures that disparitiesbetween places are big drivers of disparitiesbetween individuals. But this clearly neednot be the case. For individuals, thedisparities within local areas could farexceed those between different areas.

Our research offers new empiricalevidence on the nature, scale and recentevolution of economic disparities in Britain(Gibbons, Overman and Pelkonen, 2010).We focus on disparities in individual wages,because wages are linked to productivityand they are an important cause ofvariation in living standards. We also havevery good individual data on wages.

Using these ‘micro’ data, we assess theextent of and persistence in wagedisparities across labour market areas in

Average earnings vary widely across theregions of Britain, a fact that has promptedmany decades of policies aimed at reducingregional disparities. But as Henry Overmanand Steve Gibbons demonstrate, suchvariation reveals little, especially if we ignoreregional differences in the cost of living andavailability of local amenities.

Unequal Britain:how real are regional disparities?

Who you are is muchmore important than

where you live indetermining earnings

Britain. We examine to what extent thesearea differences arise because of differencesin the characteristics of people who live indifferent places – ‘sorting’ – versus differentoutcomes for the same types of peopleliving in different places – ‘area effects’. Wealso consider the extent to which thesedifferences across areas contribute to overallindividual wage disparities.

Our research finds that between 1998and 2008 there were few changes in area

disparities, despite many policyinterventions. It also turns out that whoyou are is much more important thanwhere you live in determining earnings (andother outcomes). Area effects only play asmall role in the overall wage dispersion.

We can tell this by looking at wagedifferences for individuals with similar skillsliving in different areas, and comparingthem with wage differences for peoplewith different skills living in the same area.The larger area wage disparities that appearat first sight, when taking a superficial lookat the data, arise through sorting ofindividuals and the area-level aggregationof the individual skills.

So this research suggests that wagedisparities across local areas in Britain arepronounced and very persistent but thatmuch of these disparities are driven by‘people’ rather than ‘place’. Regardless,such disparities between different cities anddifferent labour markets concern policy-makers because they seem to implydifferences in standards of living andeconomic welfare. But spatial earningsdisparities are uninformative aboutdifferences in people’s overall wellbeingunless we take account of differences inthe cost of living and the availability of local amenities.

In further research, we consider theextent to which higher post-tax earnings

are offset by higher housing costs (Gibbons,Overman and Resende, 2011). Figure 1shows the results for a sub-set of 157 labour market areas. Both wages and housing costs are in £1,000s per year.The wage gaps between areas areestimated from the wage gains and lossesfor individuals who move between areas;and housing costs are measured fromhouse prices, adjusting for differences inhousing quality.

The solid upward-sloping linecorresponds to the case where housingcosts rise one-for-one with wages. Giventhat most people in Britain are free tochoose where they live, higher wagesshould translate directly into higher house

prices for places that are otherwiseidentical. That is, we should expect £1 higher wages to mean housing costsrise by £1 per person. In fact, on averagethe empirical relationship in Britain is closeto this theoretical benchmark.

But as Figure 1 shows, there is also agreat deal of variation around this generalrelationship. What drives this variation isthe fact that other things are not equal.Places differ in the local amenities that theyprovide to households. These amenitiesinclude crime, weather, pollution,entertainment and natural beauty.

For example, places with high housingcosts relative to wages must offer somekind of local amenity – better restaurantsand entertainment perhaps, or lower crimeand less pollution – which helps to offsetthe fact that real income is low in the area.Similarly, places that offer poor localamenities must ‘compensate’ people byoffering low house prices relative to wages.This suggests that we can use cost-minus-earnings differentials as a measure ofquality of life.

Rather than focusing on area rankings,it is more useful to consider what thisapproach tells us about the trade-off facedby people in Britain. The dashed line inFigure 1 does this by showing how therelationship between house prices andwages changes as we move from low to


24

Aberdeen

Barnsley

Bedford

Birmingham

Blackburn

BlackpoolBolton

Bournemouth

Brighton

Bristol

Burnley

Cambridge

Cardiff

Chelmsford

Colchester

Crawley

Doncaster

Dundee

Edinburgh

Exeter

GlasgowGloucester

Grimsby

Guildford

Hastings

Huddersfield

Hull

Ipswich

Lanark

Leeds

Leicester

London

Maidstone

Manchester

Mansfield

Milton Keynes

Newcastle

Norwich

Nottingham

Oxford

Plymouth

Poole

Reading

Sheffield

Slough

Stevenage

Swansea

Swindon

Tunbridge Wells

WakefieldWolverhampton

Worcester

Worthing

York

Ayr

Banbury

Basingstoke

Bath

Canterbury

Chichester

Crewe

Dorset

East AngliaEast Cornwall

East Highlands

East Kent

East Lincolnshire

Eastbourne

Harlow

Inverness

Kendal

Lancaster

Livingston

North Cumbria

North Devon

North Norfolk

North Scotland

North West Devon

Newbury

RugbySouth Devon

South Moray

South West Wales

South Wales

Portsmouth

Scunthorpe

Shrewsbury

Stafford

Stirling

TauntonWest CornwallWest Kent

Warwick

Bridgend

Wrexham

5

4

3

2

1

0

-1

-2

-2 -1.5 -1 -.5 0 .5 1 1.5

Figure 1:

The regional relationship between earnings and house prices in Britain

Hou

sin

g co

sts

in £

1,00

0s p

er y

ear

Net pay in £1,000s per year

Policy should be assessed by its impact on

people not places


25

high wages.Places with lower wages in Britain tend

to be rural but as wages increase, houseprices tend to fall not rise (so therelationship between house prices andwages is downward sloping). This suggeststhat places with high levels of consumeramenities tend to have few productiveadvantages for firms.

In these places, consumer demand forlocal amenities drives up land costs andhouse prices, but since these areas do notoffer productive advantages, wages mustalso be lower to induce businesses to locatethere. Households in the lowest wageplaces are willing to pay high house pricesbecause they are compensated by higherlocal amenities.

When you look at the places we aretalking about – for example, West and EastCornwall, Devon and Kendal – this clearlymakes sense. These are places in which‘underperformance’ and a lack ofdevelopment has gone hand-in-hand withpreservation of rich natural amenities thatare highly valued by consumers, eventhough wages are low. This high valuationof natural and recreational resources,reflected in housing costs, is borne out inmore detailed analysis we have done(Gibbons, Mourato and Resende, 2011).

In contrast, moving to higher wageareas – on the right-hand side of Figure 1 –we tend to see house prices increasing aswages rise (so the relationship betweenhouse prices and wages is positive). Thissuggests that local producer benefits tendto drive the relationship for higher wageareas. Firms drive up land costs in theselabour markets, and workers must be

compensated with higher wages to inducethem to live there, but with house priceshigher to offset the benefits of higherwages.

Note though, that on the far right-handside of Figure 1, in London and the SouthEast, amenities for consumers andproductive advantages for firms tend to bepositively correlated across labour markets.Housing costs rise steeply, more than one-for-one with wages, indicating thatconsumers are willing to pay over andabove the expected wage gain to live inthese areas – though here the amenities arequite different from those they expect tofind in places like the South West ofEngland and the Lake District.

So across Britain, our research showsthat increased living costs (particularly ofhousing) tend to offset completelyincreased wages for the averagehousehold. In the lowest wage areas, whichare mostly rural, differences in amenitiesdrive the trade-off between wages and thecost of living. In higher wage areas, whichare mostly urban, differences in firmproductivity drive the results.

What are the implications of thisresearch for urban and regional policy inBritain? We highlight four:

Area averages are not very usefulindicators of wellbeing: Differences inaverage incomes across neighbourhoods orregions reflect the interaction of areaeffects and the sorting of people. Highwages tend to be offset by high houseprices or low quality of life so incomedifferences on their own are not very usefulindicators of differences in wellbeing.

Policy should be assessed by itsimpact on people not places: Peopletrade off wages, cost of living andamenities, and they can move in responseto changes in any of these. As a result theimpact on observed area differences offersa very poor guide to the overall effects ofpolicy on individuals.

In practice, policy has probablybeen too heavily focused on places:Area effects mean that living in someplaces negatively affects individualoutcomes. The usual response is to try toimprove ‘bad’ areas. An alternative wouldbe to focus on improving outcomes forpeople who live in bad areas.

The focus on area differences biasespolicy towards the first response.Unfortunately, evidence suggests this hasnot been successful as area effects are very

persistent. This argues for a greater focus onimproving outcomes for individualsincluding, possibly, removing barriers thatprevent people relocating to better areas.

Policy has paid too little attention tohouse prices and amenities: Planningdecisions play a key role in generating areadisparities because people and firms sort inresponse to both wages and local costs.Similarly, disparities in amenities matterbecause high quality of life compensatespeople if wages are low relative to the costof living. Local policy-makers have policiesthat directly affect house prices andamenities and relatively few that affectwages and employment.

At a time of constrained finances andweak economic performance it is moreimportant than ever that urban and regionaleconomic policy focus on sensible objectivesusing cost effective policy levers. Reflectingthese four key insights when consideringhow policy should be targeted anddeveloped would be a useful step in theright direction.

The high valuation ofnatural amenities is

reflected in high housingcosts in relatively low

wage places like Cornwalland the Lake District

Henry Overman is director of the Spatial

Economics Research Centre (SERC),

professor of economic geography at LSE and

a research associate in CEP’s globalisation

programme. Steve Gibbons is research

director of SERC, reader in economic

geography at LSE and a research associate in

CEP’s education and skills programme.

Further reading

Steve Gibbons, Henry Overman and Panu

Pelkonen (2010) ‘Wage Disparities in Britain:

People or Place?’, SERC Discussion Paper No.

60 (http://www.spatialeconomics.ac.uk/

textonly/SERC/publications/download/

sercdp0060.pdf).

Steve Gibbons, Henry Overman and

Guilherme Resende (2011) ‘Real

Earnings Disparities in Britain’, SERC

Discussion Paper No. 65

(http://www.spatialeconomics.ac.uk/

textonly/SERC/publications/download/

sercdp0065.pdf).

Steve Gibbons, Susana Mourato and

Guilherme Resende (2011) ‘The Amenity

Value of English Nature: A Hedonic Price

Approach’, SERC Discussion Paper No. 74

(http://www.spatialeconomics.ac.uk/textonly/

SERC/publications/download/sercdp0074.pdf).

http://www.spatialeconomics.ac.uk/textonly/SERC/publications/download/se

ncp19 07 jb - centre for economic performance · 2011. 10. 26. · r&d centrepiece issn 1362-3761...

Documents