Title: Trade with Unemployment Part 2: New Insights from Old Trade Theory
1Trade with UnemploymentPart 2 New Insights
from Old Trade Theory
- Carl Davidson
- Michigan State University
2Old Trade Theory
- For simplicity, I will focus on the 2 sector HOS
and Ricardian GE models - Here the industry is the appropriate unit of
measure (firms within an industry are all the
same) - All markets are perfectly competitive, there is
full employment
3Old Trade Theory
- Focus on the two biggest questions
- What causes trade?
- Who gains and who loses from trade? That is, what
is the link between trade and factor rewards? - How does adding unemployment change the answers?
4What Causes Trade (Ricardo)?
- Interaction of cross-sector differences and
cross-country differences - In Ricardian model, labor productivity varies
across countries and, for at least one country,
across sectors -
5What Causes Trade (HOS)?
- Sectors differ in their use of factors (factor
intensities, measured by aij) - Countries differ in endowments of resources
- Differences in factor use across sectors coupled
with differences in endowments across countries
leads to trade between countries (HO Theorem)
6HO Theorem
- If a country has a relatively large endowment of
labor, labor will be relatively cheap in that
country - In that country, it will therefore be relatively
cheap to produce good that are labor intensive in
production (i.e., the autarkic price of that good
will be low relative to ROW)
7HO Theorem
- Countries will have a comparative advantage in
the good that makes relatively intensive use of
its relatively abundant factor
8Adding Unemployment
- Basic point The structure of the labor market
can influence trade patterns - In our search framework, workers cycle between
periods of employment and unemployment the
employment process is risky - Workers take this risk into account when choosing
an occupation
9Adding Unemployment
- Suppose that two countries are identical in all
aspects except for the structure of their labor
markets - In the labor markets, matching technologies and
job security may differ for a variety of reasons
(info flows, policies related to hiring and
firing,) - What are the implications?
10Adding Unemployment
- Suppose that in one country the matching
technology is relatively less efficient in a
sector than it is in the same sector in other
countries - Then, workers face more risk looking for a job in
that sector than they do in other countries
they must be paid a compensating differential to
seek such a job - The compensating differential pushes up autarkic
prices and makes it less likely the country can
export that good
11Adding Unemployment
- Suppose that in one country jobs are less secure
in a sector than they are in the same sector in
other countries - Then, workers face more risk looking for a job in
that sector than they do in other countries
they must be paid a compensating differential to
seek such a job - The compensating differential pushes up autarkic
prices and makes it less likely the country can
export that good
12Bottom Line
- Employment risk affects wages
- Wages affect autarkic prices and trade patterns
- A country is more likely to have a comparative
advantage in a good that is produced in a sector
with a relatively low duration of unemployment
and a relatively high job duration
13Empirical Evidence?
- Substantial evidence that job creation/job
destruction rates vary across sectors and
countries (DHS 1996 and the work that followed) - Evidence that employment risk affects
compensating differentials (Abowd and Ashenfelter
1981)
14Empirical Evidence?
- Using two sources of data on turnover (DHS for
job turnover, BLS for worker turnover) we found
strong evidence that industry trade position is
strongly and negatively tied to job destruction
rate that is, the higher the job destruction
rate, the more you import (Davidson Matusz RIE
2004)
15Empirical Evidence?
- Alternative story a surge of imports destroys
jobs, increases job destruction rate (workers in
more open sectors face less job security) - What is the direction of causality?
16Causality?
- One story (ours) is about LR behavior look at
relationship between trade position and average
turnover rate - Second is a SR story look at trade position and
deviation of turnover rate from the mean - In all regressions, relationship much stronger
for mean turnover rate
17Additional Evidence
- Janiak (Working Paper 2006) Looks at how changes
in trade patterns (caused by liberalization)
affects turnover rates findings similar to
ours, but story of causation reversed - Cunat and Melitz (Working Paper 2006) Argue that
labor market flexibility shapes comparative
advantage find that countries with flexible
labor markets specialize in high volatility
industries
18Trade and Factor Rewards (R)
- In Ricardian model, some factors are sector
specific, some are mobile - Free trade benefits factors specific to the
export sector and harms those specific to the
import sector - Impact on mobile factors is ambiguous
- For specific factors, industry affiliation is all
that matters
19Trade and Wages (HOS)
- Many results in HOS model based on manipulation
of market clearing conditions - L0 LX LY aLXX aLYY
- K0 KX KY aKXX aKYY
- PX aLXwx aKXrx
- PY aLYwy aKYry
20Unit input requirements
Kj
akj
X(L,K)1
Lj
aLj
21Trade and Wages
- Totally diff. one product market clearing
expression to get
But, since unit input requirements are optimal,
the last two terms sum to zero
22Trade and Wages
- Subtract the two expressions yields
where ? measures the relative factor
intensities of the two sectors (in value terms)
? gt 0 means that sector X is relatively more
labor intensive than sector Y
23Trade and Wages
- Suppose that the world price of a labor-intensive
good rises (PX increases) - Prod. of X rises. Factors released from sector Y
are less labor intensive than those absorbed in
sector X. - Demand for L rises, demand for K falls.
- w rises (L gains), r falls (K loses), regardless
of where it is employed
24Trade and Wages
- Is there another effect? After all, all firms
become more K intensive (the aij terms change) - But, changing the aij terms has no impact on
prices since the aij terms were optimal (Envelope
theorem)! - This will be useful shortly
25Trade and Wages
- Note also The price increase is a convex comb.
of factor changes
Thus, labor gains in real terms
26Stolper-Samuelson
- Trade benefits a countrys abundant factor and
harms its scarce factor (industry affiliation
does not matter) - Protection of an industry benefits the factor
used relatively intensively in that sector
27More Details on HOS
- Totally diff. the factor market clearing
expressions yields
where ? measures the relative factor
intensities of the two sectors (in physical
terms) ? gt 0 means that sector X is relatively
more labor intensive than sector Y (with perfect
competition, ?? gt 0).
28 Px/Py
RS
PM clearing
RD
X/Y
w/r
FM clearing
45o line
w/r
29Adding Unemployment
- Yesterday I noted that with unemployment the
product market clearing conditions can be written
as - Px aLxwXu aKxrXv
- Py aLywYu aKyrYv
- Differentiate to get
30Adding Unemployment
- Key question Are the unit input requirements
optimal? If so, last two terms sum to zero
again. - If so we get an extension of Stolper-Samuelson
31Adding Unemployment
- It is now the unemployed factors that move to
clear factor markets - An increase in a price attracts more unemployed
factors to that sector - This changes the mix of unemp. factors
- Usual SS result, but the implication is now for
the return to unemployed factors
32Extended Stolper-Samuelson
- If a country protects a good produced in a labor
intensive sector, all unemployed labor benefits
and all idle capital is harmed
33An Important Exception
- But, this version only holds if unit input
requirements (which include unemployed factors)
are optimal - Why?
- When you differentiate factor market clearing
conditions the envelope theorem no longer applies
can get feedback effects on prices this is
due to externalities in the search process
34Implications?
- The free trade equilibrium may not be optimal
equilibrium unemployment may be too high or too
low - The relative supply curve need not be upward
sloping it can bend back, leading to multiple
free trade equilibria - Expectations about economic activity become quite
important - See DMM (1987, 1988, 1991)
-
35Why not emphasize these results?
- Based on controversial, hard to measure search
externalities it is unclear if such
externalities are important, hard to predict
which way they go more research needed - Other interesting insights are present even when
unit input requirements are optimal
36Back to Stolper-Samuelson
- Hard to test extended SS cannot observe return
to unemployed or idle factors - What about the return to employed factors?
- We get a convex combination of SS and Specific
Factors (SF, or Ricardian) effects
37Trade and Wages with Unemployment
- When factors are employed, frictions create an
attachment to that sector this makes them
quasi-fixed factors this generates SF effects - All employed workers realize that they will spend
a fraction of life unemployed this generates SS
effects
38Predictions
- Trade and factor rewards will be linked by a
convex comb. of SF and SS effects - SS effects will be strong in high turnover
industries - SF effects will be strong in low turnover sectors
39Empirical Evidence
- How to test this?
- Some have tried to test SS by looking at
political contributions - SS predicts that factor abundance determines
views on trade policy - SF predicts that industry affiliation determines
view on trade policy
40Empirical Evidence
- In the past it has been difficult to find much
support for SS empirically - Magee (1980) looked at lobbying behavior, found
support only for SF - Dismissed as a misunderstanding of SS, which is
really about long run behavior
41Empirical Evidence
- More recent evidence suggests that industry
affiliation and factor abundance play a role
(Beaulieu and Magee 2004) - Also some evidence that voting behavior in US and
Canada consistent with SS (Beaulieu 1998, 2000
Balistreri 1997 Slaughter 1998)
42Prediction of our Model
- Relatively abundant factors employed in import
sectors should favor protection if turnover is
low, free trade if turnover is high - Relatively scarce factors employed in export
sectors should favor free trade if turnover is
low, protection if turnover is high
43Data
- Look at PAC contributions in US aimed at
influencing NAFTA and GATT - These lobbies typically give to both parties (for
access), will give more to one to influence
outcome and policy - Theory says Base decision on factor abundance
(SS effect), industry affiliation (SF effect) and
turnover rates
44Data
- DHS Turnover data (robust to other measures)
use average turnover rate for 1988-1992 - Get PAC contributions from Federal Election
Committee look at share going to supporters of
NAFTA, approval of GATT (Uruguay Round), and both
45Data
- PACs can be linked to an industry based on data
from the Center for Responsible Politics and from
info on the web - PACs classified as representing import-competing
or exporting industry based on net trade position
from NBER trade data
46Predictions
- High Turnover Industries
- Capital and labor should differ in preferences
toward trade policy - Low Turnover Industries
- Capital and labor should have the same
preferences toward trade policy
47Fraction of 1991-92 PAC contributions given to
free trade proponents
48Fraction of 1991-92 PAC contributions given to
free trade proponents
49Extensions
- Can add controls, look at differences in
differences, add fixed effects, results do not go
away, often get stronger - Bottom Line There is (strong?) support for the
theory that the structure of the labor market
effects the link between trade and wages
50Important Extensions
- With the model established, it can be used to
look at a whole host of important issues that
could not be addressed in full employment models - For example
- Costs of adjustment (taking spells of
unemployment and retraining into account) - What is the best way to compensate those harmed
by trade liberalization