2. Analysed Supply-Demand Framework for labour market
Assessed changes in relative wages for skilled Vs unskilled
What happened:
Sharp rise in relative wages of college graduates
Average wages of older workers increased relative to younger
workers (low education level)
These two points give an increase in relative wages of young male
college graduates against young males with 12 or less years of
schooling
Data from 1979 to 1987
Increase by 30%
Differences between earning of women and men decreased
Relative wages of women w.r.t. men went up by 8%
Katz & Murphy (1992)
3. Why it happened?
Technological changes (Krueger, 1991)
In 1993 Krueger showed that using computers at work increases your relative wage by 16.2%
when controlling for most variables)
‘Computer revolution’
Raised demand for more educated flexible workers
Reduced demand for physical labour
Shifts in product demand largely associated with large trade deficits in
1980s (Murphy & Welch, 1991)
Led to decline in manufacturing employment
Shift toward sectors with more educated/female intensive labour
Employment between sectors
Changes in wage-setting institutions
Decline in unionization (Thatcher/ Reagan governments)
Erosion of real minimum wages in US
Katz & Murphy (1992)
4. Key results:
Rapid increase in relative demand for skilled workers and supply of
graduates are a KEY for explanation of rising inequality and changes
in the wage structure over last 25 years
Being a college graduate rather school graduate only increase wage
differential by 3.3% a year
Changes in relative demand for skilled workers mainly account for
SBTC (between-sector employment)
No major differences in within-sector employment
Do not have major impact on changes in relative demand
Katz & Murphy (1992)
5. Found same results as Katz and Murphy
BUT:
The trend effect is a bit lower (2.7%)
Acemoglu and Autor (2012)
6. Analysed 402 British workplaces between 1984-1990
SBTC – Indirect Evidence
Used within/between decomposition
Machin (1996)
7. SBTC – Indirect Evidence
Within-establishment
component had by far
larger impact on the
changes in employment
rates for skilled workers
Machin (1996)
SBTC – Direct Evidence
16 UK manufacturing
industries
Industry R&D investment
and increase in capital stock
had a big influence on
increased employment for
non-manuals
8. Introduction of concept of ‘polarization of labour market’
Analysis based on routine jobs, non-routine jobs, and jobs substituted by
technical change.
Goos and Manning (2007)
Key Results:
Little doubt that technology has a
powerful impact on the labour market.
The hypothesis of SBTC is only a partial
truth and cannot explain all of the
important changes in the labour market
SBTC hypothesis seems best able to
explain what is happening in the top half
of the wage distribution but not its
bottom half.
TBTC shows more nuanced view about the
impact of technology proposed –
underlining the evidence that demand for
‘middling’ jobs has fallen.
Showed UK evidence of increased job
polarization that is consistent with the
TBTC hypothesis.
9. UK real wages falling
Since 2008 on average by 8%
Mainly on younger (under 25 by 15%), but still affected everyone across the
wage distribution
Why?
Unemployment exerting a larger downward pressure on wages than in
previous recessions.
The extremely poor productivity record through the recession and recovery
has not created room for wage rises
Wages of typical British workers are no longer keeping up with productivity
gains made in the economy
This stems from a growing contribution of total compensation going toward
supporting pensions and annuities (Eric French )
Highest paid (top 1%) are taking a disproportionate share of the gains from
productivity leaving room for few gains by ordinary workers (compensation
for investment bankers and senior executives)
Gregg, Machin & Fernández-Salgado
(Feb 2014)
10. Estimated the effect of police on crime rates
Faced Simultaneity Bias because:
Endogeneity of Police with respect to crime rates
Higher crime rates likely to increase the marginal productivity of police
Cities/ districts with high crime rates tend to have larger police forces
Even if police reduces crime
1994: US Congress authorizes Crime Bill
Additional federal funding on further deployment of policemen
Disproportional increase in police concentrated in election year
What Levitt does?
Builds 2SLS framework
In First-stage he uses election cycles as IV for extra police deployment
In Second-stage he recovers impact of increased policing on crime
Levitt (1997)
11. Levitt (1997)
Election cycles are
indicated to be a strong
IV for further police
deployment
When recovering
the impact of an
increase in police
on crime rates,
Levitt clears for
simultaneity bias
and outlines the
causal effect
12. 7/7 terrorist attacks
London government/ Scotland Yard deploys extra police forces in Central
London (treatment group), without any increase in police forces in Outer
London (control group)
Analysis:
Highly robust evidence on the causal effect of police on crime
Strong evidence that more police lead to reductions in susceptible crimes
Mainly property crime
Terrorist attacks can induce exogenous variations in the allocation of police resources
that can be used to estimate the causal impact of police on crime
Framework:
2 stages: IV + DiD
The scale of the police deployment is much greater than the highly localized
responses
Allows to provide new, highly robust IV-based estimates of the crime-police elasticity
Ceteris paribus dimension to the London police deployment
By temporarily extending its resources (primarily through overtime), the police service was
able to keep their force levels constant in the control group, while simultaneously increasing
the police presence in the treatment group
This provides a clean setting to test the relationship between crime and police.
Draca, Matchin and Witt (2011)
13. Results:
Clear evidence that the timing and
location of falls in susceptible crimes
closely coincide with the increase in police
deployment.
Crime rates quickly returned to pre-attack
levels after the six week “policy-on”
period
Draca, Matchin and Witt (2011)
14. Moving on to the Part II…
Here we will be looking at:
1. Explaining the much muted pace of wage inequality observed in
the 1990s in light of Machin 2002
2. And perhaps this may shed some light on how further research
ought to be undertaken
3. Lastly, we lightly touch on a couple of policy implications which
may be of interest.
Our approach in this section is by giving more emphasis on figures and
findings given that our colleague have extensively discussed the
relevant models and papers.
16. Skill-biased
technological change
(SBTC) affects the
labour market by:
Benefitting the more-skilled workers
• Increasing productivity
• Raising wages
BUT detrimental to less-skilled workers
• Lowers their wages
• Less of them are employed
OVERALL EFFECT:
Relative wages & relative employment rates
of skilled vs unskilled workers ROSE
17. The trend of wage inequality over the past three decades reflects
how great the force of ‘skill-biased technological change (SBTC)’ is.
18. Technological advances create such huge wage disparities during the 1980s.
This is mainly due to the fact that technological advances were either:
A. Complement complex, non-routine tasks
• Outward shift in supply of informational inputs (both in quantity
and quality)
• Increases marginal productivity of workers performing these
tasks
• As a result, workplace computerisation increases the demand
for problem-solving tasks.
WHILE SIMULTANEOUSLY THEY…
B. Substitutes routine tasks
• Replaces routine information processing, communications and
coordinating functions
• Examples: clerks, cashiers, telephone operators and
bookkeepers
• In economic terms: advances in IT have sharply lowered the
price of accomplishing procedural cognitive tasks (i.e. rules-
based reasoning)
19. BUT in the 1990s, rising inequality seemed to have slowed.
Is this evidence that computers have become so widespread (saturated) that a simple
headcount measure is no longer adequate?
(Machin; 2002)
• By 1997, in some industries, computer usage levels were already at high levels, near saturation and
cannot rise much on the basis of the percent using computers variable (as illustrated above using US
data).
• As such, this might pose a problem to relate skill upgrading to changes in industry computer usage.
20. RESULTS:
• 1st column showed no relation between 1990s skill upgrading and increased use of computers
in the 1990s (same finding in the US)
• This supports the notion that simple computer measure may not be a good proxy of technology
chance in the 1990s when computer use levels have reached such high levels.
• BUT once the result was broken down by importance of the computer to the job, we do see that
skill upgrading is associated with increased importance of computers
• Strongest positive (and statistically significant) association is in column (4)
• Hence, relative demand is still shifting in favour of skilled workers in industries where computers
are becoming more important, even in the 90s.
21. Policy Implications
What sort of policies could be implemented to reverse this trend in
inequality from persisting over time?
• Short –term solution is to carry out reform of labour market
institutions:
i. Minimum wage [issue: real value of minimum wage is falling
over time]
ii. Trade unions
• Meanwhile, in the long-term the best response is for
governments to build up its human capital
22. This suggests that Germany’s wages in 2008 would have been higher if union
coverage had remained the same as in 1995 level.
MAIN FOCUS: the difference is particularly large at the lower end of the wage
distribution
From: Christian Dustmann et. al; Germany’s Resurgent Economy (2014)