Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
Online appendixes for “Regional Effects of Trade Reform: What is theCorrect Measure of Liberalization?” by Brian K. Kovak
Specific Factors Model
A1. Factor prices
This section closely follows Jones (1975), but deviates from that paper’s resultby allowing the amount of labor available to the regional economy to vary. Con-sider a particular region, r, suppressing that subscript on all terms. Industriesare indexed by i = 1...N . L is the total amount of labor and Ti is the amount ofindustry i-specific factor available in the region. aLi and aT i are the respectivequantities of labor and specific factor used in producing one unit of industry ioutput. Letting Yi be the output in each industry, the factor market clearingconditions are
(A1) aT iYi = Ti ∀i,
(A2)∑i
aLiYi = L.
Under perfect competition, the output price equals the factor payments, where wis the wage and Ri is the specific factor price.
(A3) aLiw + aT iRi = Pi ∀i
Let hats represent proportional changes, and consider the effect of price changesPi. θi is the cost share of the specific factor in industry i.
(A4) (1− θi)w + θiRi = Pi ∀i,
which follows from the envelope theorem result that unit cost minimization implies
(A5) (1− θi)aLi + θiaT i = 0 ∀i.
Differentiate (A1), keeping in mind that Ti is fixed in all industries.
(A6) Yi = −aT i ∀i
Similarly, differentiate (A2), let λi = LiL be the fraction of regional labor utilized
in industry i, and substitute in (A6) to yield
(A7)∑i
λi(aLi − aT i) = L.
21
By the definition of the elasticity of substitution between Ti and Li in production,
(A8) aT i − aLi = σi(w − Ri) ∀i.
Substituting this into (A7) yields
(A9)∑i
λiσi(Ri − w) = L.
Equations (A4) and (A9) can be written in matrix form as follows.
(A10)
θ1 0 . . . 0 1− θ1
0 θ2 . . . 0 1− θ2...
. . ....
...0 0 . . . θN 1− θN
λ1σ1 λ2σ2 . . . λNσN −∑
i λiσi
R1
R2...
RNw
=
P1
P2...
PNL
Rewrite this expression as follows for convenience of notation.
(A11)
[Θ θLλ′ −
∑i λiσi
] [Rw
]=
[P
L
]Solve for w using Cramer’s rule and the rule for the determinant of partitionedmatrices.
(A12) w =L− λ′Θ−1P
−∑
i λiσi − λ′Θ−1θL
Note that the inverse of the diagonal matrix Θ is a diagonal matrix of 1θi
’s. Thisyields the effect of goods price changes and changes in regional labor on regionalwages:
(A13) w =−L∑i′ λi′
σi′θi′
+∑i
βiPi
(A14) where βi =λiσiθi∑
i′ λi′σi′θi′
This expression with L = 0 yields (1). Changes in specific factor prices can becalculated from wage changes by rearranging (A4).
(A15) Ri =Pi − (1− θi)w
θi22
Plugging in (A13) and collecting terms yields the effect of goods price changesand changes in regional labor on specific factor price changes.
(A16) Ri =(1− θi)θi
L∑i′ λi′
σi′θi′
+
(βi +
1
θi(1− βi)
)Pi −
(1− θi)θi
∑k 6=i
βkPk
Setting L = 0 in (A13) and (A16) yields the equivalent expressions in Jones(1975).
A2. Graphical representation of the model
The equilibrium adjustment mechanisms at work in the model are demon-strated graphically in Figure A1, which represents a two-region (r = 1, 2) andtwo-industry (i = A,B) version of the model.29 Region 1 is relatively well en-dowed with the industry A specific factor. In each panel, the x-axis representsthe total amount of labor in the country to be allocated across the two industriesin the two regions, and the y-axis measures the wage in each region. Focusingon the left portion of panel (a), the curve labeled PAF
AL is the marginal value
product of labor in industry A, and the curve labeled PBFBL is the marginal value
product of labor in industry B, measuring the amount of labor in industry Bfrom right to left. Given labor mobility across sectors, the intersection of the twomarginal value product curves determines the equilibrium wage, and the alloca-tion of labor in region 1 between industries A and B, as indicated on the x-axis.The right portion of panel (a) is interpreted similarly for region 2. For visualclarity, the figures were generated assuming equal wages across regions beforeany price changes. This assumption is not necessary for any of the theoreticalresults presented in the paper.
Panel (a) of Figure A1 shows an equilibrium in which wages are equalized acrossregions. Since region 1 is relatively well endowed with industry A specific factor,it allocates a greater share of its labor to industry A. Panel (b) shows the effectof a 50% decrease in the price of good A, so the good A marginal value productcurve in both regions moves down halfway toward the x-axis. Consistent with(1), the impact of this price decline is greater in region 1, which allocated a largerfraction of labor to industry A than did region 2. Thus, region 1’s wage falls morethan region 2’s wage. Now workers in region 1 have an incentive to migrate toregion 2. For each worker that migrates, the central vertical axis moves one unitto the left, indicating that there are fewer laborers in region 1 and more in region2. As the central axis shifts left, so do the two marginal value product curvesthat are measured with respect to that axis. This shift raises the wage in region1 and lowers the wage in region 2. Panel (c) shows the resulting equilibrium
29Figure A1 was generated under the following conditions. Production is Cobb-Douglas with specific-factor cost share equal to 0.5 in both industries. L = 10, T1A = 1, T1B = 0.4, T2A = 0.4, and T2B = 1.Initially, PA = PB = 1, and after the price change, PA = 0.5.
23
assuming costless migration and equalized wage across regions. Again, none ofthe theoretical results in the paper require this assumption.
A3. Nontraded goods prices
As above, consider a particular region, omitting the r subscript on all terms.Industries are indexed by i = 1...N . The final industry, indexed N , is nontraded,while other industries (i 6= N) are traded. The addition of a nontraded indus-try does not alter the results of the previous section, but makes it necessary todescribe regional consumers’ preferences to fix the nontraded good’s equilibriumprice.
Assume a representative consumer with Cobb-Douglas preferences over goodsfrom each industry. This implies the following relationship between quantitydemanded by consumers, Y c
i , total consumer income, m, and price, Pi.
(A17) Y ci = m− Pi
Consumers own all factors and receive all revenue generated in the economy;m =
∑i PiY
pi where Y p
i is the amount of good i produced in equilibrium. Apply-ing the envelope theorem to the revenue function for the regional economy, onecan show30
(A18) m = ηLL+∑i
ϕiPi.
where ηL is the share of total factor payments accounted for by wages and ϕiis the share of regional production value accounted for by industry i. Plugging(A18) into (A17) for the nontraded industry and rearranging terms,
(A19) PN −∑i 6=N
ϕi∑i′ 6=N ϕi′
Pi −ηL
1− ϕNL =
−1
1− ϕNY cN .
Holding labor fixed, this expression shows that consumption shifts away from thenontraded good if the nontraded price increases relative to a weighted averageof traded goods prices, with weights based on each good’s share of traded sectoroutput value.
A similar expression can be derived for the production side of the model. Wageequals the value of marginal product, so w = PN+FNL . The specific factor is fixed,
so Y pN = (1 − θN )LN . Combining these two observations with Euler’s theorem
30The revenue maximization problem is specified as follows.
r(P1...PN , L) = maxLi
∑i
PiFi(Li, Ti) s.t.
∑i
Li ≤ L
24
and the definition of the elasticity of substitution as in footnote 6 yields
(A20) w = PN −θN
σN (1− θN )Y pN
Substitute in the expression for w in (A13) and rearrange.
(A21) PN −∑i 6=N
βi∑i′ 6=N βi′
Pi +1
(1− βN )∑
i′ λi′σi′θi′
L =θN
(1− βN )σN (1− θN )Y pN
Holding labor fixed, this expression shows that production shifts toward the non-traded good if the nontraded price increases relative to a weighted average oftraded goods prices, with weights based on the industry’s size and labor demandelasticity captured in βi.
Equations (A19) and (A21) relate very closely to the intuition described inSection I.B for the case of only one traded good. Consumption shifts away fromthe nontraded good if its price rises relative to average traded goods prices whileproduction shifts toward it. Since regional consumption and production of thenontraded good must be equal, with a single traded good the nontraded pricemust exactly track the traded good price. Combining (A19) and (A21) showsthat in the many-traded-good case, the nontraded price change is a weightedaverage of traded goods price changes.
(A22) PN =
ηL − σNθN
(1−θN )∑i λi
σiθi∑
i′ 6=N
[σNθN
(1− θN )βi′ + ϕi′] L+
∑i 6=N
ξiPi
(A23) where ξi =
σNθN
(1− θN )βi + ϕi∑i′ 6=N
[σNθN
(1− θN )βi′ + ϕi′]
A4. Restriction to Drop the Nontraded Sector from Weighted Averages
Under Cobb-Douglas production with equal factor shares across industries, θi =θ, and σi = 1 ∀i. This implies that βi = λi, the fraction of labor in each industry.Similarly,
(A24) ϕi ≡PiY
pi∑
i′ Pi′Ypi′
=λiηL1− θ
= λi.
since ηL = 1− θ when θi = θ. When βi = ϕi, the weights ξi in (A22) are
(A25) ξi =βi∑
i′ 6=N βi′
25
Plug these weights into (3) and then plugging that into (1) yields a result equiv-alent to dropping the nontraded sector from the weighted average in (1) and(2).
(A26) w =
∑i 6=N βiPi∑i′ 6=N β
′i
A5. Wage Impact of Changes in Regional Labor
This section shows that an increase in regional labor decreases the regionalwage. Substituting (A22) into (A13), holding traded goods prices fixed (Pi =0 ∀i 6= N), and rearranging yields
(A27) w =
−σNθN
(1− θN )− (1− ϕN ) + λNσNθNηL
(∑
i λiσiθi
)[σNθN
(1− θN )(1− βN ) + (1− ϕN )] L.
Since the denominator is strictly positive, the sign of the relationship between Land w is determined by the numerator. An increase in regional labor will decreasethe wage if an only if
(A28) (1− ϕN ) >σNθN
(λNηL − (1− θN )).
Using the fact that ηL = (∑
i λi(1−θi)−1)−1 the previous expression is equivalentto
(A29) (1− ϕN ) >σNθN
(1− θN )
(λN (1− θN )−1∑i λi(1− θi)−1
− 1
).
The left hand side is positive while the right hand side is negative. Thus, theinequality always holds, and an increase in regional labor will always lower theregional wage.
26
Figure A1. Graphical Representation of Specific Factors Model of Regional Economies
(a) Initial Equilibrium
(b) Response to a Decrease in PA – Prohibiting Migration
(c) Response to a Decrease in PA – Allowing Migration
27
Data Appendix
B1. Industry Crosswalk
National accounts data from IBGE and trade policy data from (Kume, Pianiand de Souza 2003) are available by industry using the the Nıvel 50 and Nıvel80 classifications. The 1991 Census reports individuals’ industry of employmentusing the atividade classification system, while the 2000 Census reports industriesbased on the newer Classificacao Nacional de Atividades Economicas - Domiciliar(CNAE-Dom). Table B1 shows the industry definition used in this paper and itsconcordance with the various other industry definitions in the underlying datasources. The concordance for the 1991 Census is based on a crosswalk betweenthe national accounts and atividade industrial codes published by IBGE. Theconcordance for the 2000 Census is based on a crosswalk produced by IBGE’sComissao Nacional de Classificacao, available at http://www.ibge.gov.br/concla/.
B2. Trade Policy Data
Nıvel 50 trade policy data come from Kume, Piani and de Souza (2003). De-pending on the time period, they aggregated tariffs on 8,750 - 13,767 individualgoods first using unweighted averages to aggregate individual goods up to theNıvel 80 level, and then using value added weights to aggregate from Nıvel 80 toNıvel 50.31 In order to maintain this weighting scheme, I weight by value addedwhen aggregating from Nıvel 50 to the final classification listed in Table B1. Forreference, Figure B1 and Figure B2 show the evolution of nominal tariffs and effec-tive rates of protection in the ten largest sectors by value added. Along with thegeneral reduction in the level of protection, the dispersion in protection was alsogreatly reduced, consistent with the goal of aligning domestic production incen-tives with world prices. It is clear that the move from a high-level, high-dispersiontariff structure to a low-level low-dispersion tariff distribution generated substan-tial variation in protection changes across industries; industries with initially highlevels of protection experienced the largest cuts, while those with initially lowerlevels experienced smaller cuts.
Recall from Section III that the 1987-1990 period exhibited substantial tariff re-dundancy due to the presence of selective import bans and special import regimesthat exempted many imports from the full tariff charge. The tariff changes duringthis period did not generate a change in the protection faced by producers, butrather reflected a process called called tarifacao, or “tariffization,” in which theimport bans and special import regimes were replaced by tariffs providing equiv-alent levels of protection (Kume, Piani and de Souza 2003). The nominal tariffincreases in 1990 shown in Figure B1 reflect this tariffization process in whichtariffs were raised to compensate for the protective effect of the import bans thatwere abolished in March 1990 (Carvalho 1992).
31Email correspondence with Honorio Kume, March 12, 2008.
28
B3. Cross-Sectional Wage Regressions
In order to calculate the regional wage change for each microregion, I estimatestandard wage regressions separately in 1991 and 2000. Wages are calculated asan individual’s monthly earnings / 4.33 divided by weekly hours in their mainjob (results using all jobs are nearly identical). I regress the log wage on age,age-squared, a female indicator, an inner-city indicator, four race indicators, amarital status indicator, fixed effects for each year of education from 0 to 17+,21 industry fixed effects, and 494 microregion fixed effects. By running these re-gressions separately by survey year, I control for changes in regional demographiccharacteristics and for changes in the national returns to those characteristics.The results of these regressions are reported in Table B2. All terms are highlystatistically significant and of the expected sign. The regional fixed effects arenormalized relative to the national average log wage, and their standard errorsare calculated using the process described in Haisken-DeNew and Schmidt (1997).The regional wage change is then calculated as the difference in these normalizedregional fixed effects between 1991 and 2000:
(B1) d lnwr = (lnw2r000− lnw1
r991)− (lnw2000r − lnw1991
r )
The regional wage changes are shown in Figure 2. As mentioned in the maintext, some sparsely populated regions have very large measured regional wagechanges. To demonstrate that these results are driven by the data and not someartifact of the wage regressions, Figure B3 shows a similar map plotting regionalwage changes that were generated without any demographic or industry controls.They represent the change in the mean log wage in each region. The amount ofvariation across regions is similar, and these unconditional regional wage changesare highly correlated with the conditional versions used in the analysis, with acorrelation coefficient of 0.93.
B4. Region-Level Tariff Change Elements
The fixed factor share of input costs is measured as one minus the labor shareof value added in national accounts data. The resulting estimates of θi are plottedin Figure B4. The distribution of laborers across industries in each region (λri)are calculated in the 1991 Census using the sample of employed individuals, notenrolled in school, aged 18-55. The tariff changes in 5 are calculated from thenominal tariff data in Kume, Piani and de Souza (2003) and are shown in FigureB5.
29
FigureB1.NominalTariffTim
eline
Text
iles
Auto
, Tra
nspo
rt, V
ehic
les
Food
Pro
cess
ing
Non
met
allic
Min
eral
Man
f
Elec
tric
, Ele
ctro
nic
Equi
p.
Mac
hine
ry, E
quip
men
t M
etal
s Ag
ricul
ture
Ch
emic
als
Petr
oleu
m R
efin
ing
0 10
20
30
40
50
60
70
80
90
1984
19
85
1986
19
87
1988
19
89
1990
19
91
1992
19
93
1994
19
95
1996
19
97
1998
nominal tariff (percent)
30
FigureB2.EffectiveRateofProtectionTim
eline
Agric
ultu
re
Food
Pro
cess
ing
Met
als
Petr
oleu
m R
efin
ing
Mac
hine
ry, E
quip
men
t
Elec
tric
, Ele
ctro
nic
Equi
p.
Chem
ical
s
Auto
, Tra
nspo
rt, V
ehic
les
Text
iles
Non
met
allic
Min
eral
Man
f
0 20
40
60
80
100
120
140
160
1984
19
85
1986
19
87
1988
19
89
1990
19
91
1992
19
93
1994
19
95
1996
19
97
1998
effective rate of protection (percent)
31
Figure B3. Unconditional Regional Wage Changes
!!
!
!
!
!
!
!
!
!
!
BelémBelém
RecifeRecife
ManausManaus
CuritibaCuritiba
BrasíliaBrasília
SalvadorSalvador
FortalezaFortaleza
São PauloSão Paulo
Porto AlegrePorto Alegre
Belo HorizonteBelo Horizonte-15% to -5%-5% to 5%5% to 15%15% to 49%
-45% to -15%
Proportional wage change by microregion - normalized change in microregionfixed effects without demographic controls
32
Figure B4. Fixed Factor Share of Input Costs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Non
trad
ed
Agric
ultu
re
Food
Pro
cess
ing
Met
als
Petr
oleu
m R
efin
ing
Mac
hine
ry, E
quip
men
t
Elec
tric
, Ele
ctro
nic
Equi
p.
Chem
ical
s
Auto
, Tra
nspo
rt, V
ehic
les
Text
iles
Non
met
allic
Min
eral
Man
uf
Pape
r, Pu
blish
ing,
Prin
ting
Petr
oleu
m, G
as, C
oal
Appa
rel
Woo
d, F
urni
ture
, Pea
t
Plas
tics
Phar
ma.
, Per
fum
es, D
eter
gent
s
Oth
er M
anuf
.
Min
eral
Min
ing
Foot
wea
r, Le
athe
r
Rubb
er
Fixe
d Fa
ctor
Sha
re o
f Inp
ut C
osts
om
ittin
g th
e co
st o
f int
erm
edia
te in
puts
Source: IBGE National Accounts data 1990 Sorted by industry value added in 1990 (largest to smallest)
33
Figure B5. Tariff Changes
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
Agric
ultu
re
Food
Pro
cess
ing
Met
als
Petr
oleu
m R
efin
ing
Mac
hine
ry, E
quip
men
t
Elec
tric
, Ele
ctro
nic
Equi
p.
Chem
ical
s
Auto
, Tra
nspo
rt, V
ehic
les
Text
iles
Non
met
allic
Min
eral
Man
uf
Pape
r, Pu
blish
ing,
Prin
ting
Petr
oleu
m, G
as, C
oal
Appa
rel
Woo
d, F
urni
ture
, Pea
t
Plas
tics
Phar
ma.
, Per
fum
es, D
eter
gent
s
Oth
er M
anuf
.
Min
eral
Min
ing
Foot
wea
r, Le
athe
r
Rubb
er
Chan
ge in
log
(1 +
tarif
f rat
e)
Source: Author's calculations based on Kume et al. (2003) Sorted by industry value added in 1990 (largest to smallest)
34
TableB1—
Indust
ryAggregationand
Concordance
Fina
l Ind
ustr
ySe
ctor
Nam
eN
ível
50
Nív
el 8
019
91 C
ensu
s (at
ivid
ade)
2000
Cen
sus (
CNAE
-Dom
)1
Agr
icul
ture
110
1-19
901
1-03
7, 0
41, 0
42, 5
8111
01-1
118,
120
1-12
09, 1
300,
140
1, 1
402,
200
1, 2
002,
50
01, 5
002
2M
iner
al M
inin
g (e
xcep
t com
bust
ible
s)2
201-
202
050,
053
-059
1200
0, 1
3001
, 130
02, 1
4001
-140
043
Petro
leum
and
Gas
Ext
ract
ion
and
Coa
l Min
ing
330
1-30
205
1-05
210
000,
110
004
Non
met
allic
Min
eral
Goo
ds M
anuf
actu
ring
440
110
026
010,
260
91, 2
6092
5Ir
on a
nd S
teel
, Non
ferr
ous,
and
Oth
er M
etal
Pro
duct
ion
and
Proc
essi
ng5-
750
1-70
111
027
001-
2700
3, 2
8001
, 280
028
Mac
hine
ry, E
quip
men
t, C
omm
erci
al In
stal
latio
n M
anuf
actu
ring,
and
Tra
ctor
Man
ufac
turin
g8
801-
802
120
2900
110
Elec
trica
l, El
ectro
nic,
and
Com
mun
icat
ion
Equi
pmen
t and
Com
pone
nts M
anuf
actu
ring
10-1
110
01-1
101
130
2900
2, 3
0000
, 310
01, 3
1002
, 320
00, 3
3003
12A
utom
obile
, Tra
nspo
rtatio
n, a
nd V
ehic
le P
arts
Man
ufac
turin
g12
-13
1201
-130
114
034
001-
3400
3, 3
5010
, 350
20, 3
5030
, 350
9014
Woo
d Pr
oduc
ts, F
urni
ture
Man
ufac
turin
g, a
nd P
eat P
rodu
ctio
n14
1401
150,
151
, 160
2000
0, 3
6010
15Pa
per M
anuf
actu
ring,
Pub
lishi
ng, a
nd P
rintin
g15
1501
170,
290
2100
1, 2
1002
, 220
0016
Rub
ber P
rodu
ct M
anuf
actu
ring
1616
0118
025
010
17C
hem
ical
Pro
duct
Man
ufac
turin
g17
,19
1701
-170
2, 1
901-
1903
20
023
010,
230
30, 2
3400
, 240
10, 2
4090
18Pe
trole
um R
efin
ing
and
Petro
chem
ical
Man
ufac
turin
g18
1801
-180
620
1, 2
02, 3
52, 4
7723
020
20Ph
arm
aceu
tical
Pro
duct
s, Pe
rfum
es a
nd D
eter
gent
s Man
ufac
turin
g20
2001
210,
220
2402
0, 2
4030
21Pl
astic
s Pro
duct
s Man
ufac
turin
g21
2101
230
2502
022
Text
iles M
anuf
actu
ring
2222
01-2
205
240,
241
1700
1, 1
7002
23A
ppar
el a
nd A
ppar
el A
cces
sorie
s Man
ufac
turin
g23
2301
250,
532
1800
1, 1
8002
24Fo
otw
ear a
nd L
eath
er a
nd H
ide
Prod
ucts
Man
ufac
turin
g24
2401
190,
251
1901
1, 1
9012
, 190
2025
Food
Pro
cess
ing
(Cof
fee,
Pla
nt P
rodu
cts,
Mea
t, D
airy
, Sug
ar, O
ils, B
ever
ages
, and
Oth
er)
25-3
125
01-3
102
260,
261
, 270
, 280
1501
0, 1
5021
, 150
22, 1
5030
, 150
41-1
5043
, 150
50,
1600
032
Mis
cella
neou
s Oth
er P
rodu
cts M
anuf
actu
ring
3232
0130
033
001,
330
02, 3
3004
, 330
05, 3
6090
, 370
0099
Non
trade
d G
oods
and
Ser
vice
s33
-43
3301
-430
134
0, 3
51-3
54, 4
10-4
24, 4
51-4
53, 4
61, 4
62, 4
64,
471-
476,
481
, 482
, 511
, 512
, 521
-525
, 531
, 533
, 54
1-54
5, 5
51, 5
52, 5
71-5
78, 5
82-5
89, 6
10-6
19,
621-
624,
631
, 632
, 711
-717
, 721
-727
, 901
, 902
1500
, 400
10, 4
0020
, 410
00, 4
5001
-450
05, 5
0010
, 50
020,
500
30, 5
0040
, 500
50, 5
3010
, 530
20, 5
3030
, 53
041,
530
42, 5
3050
, 530
61-5
3068
, 530
70, 5
3080
, 53
090,
531
01, 5
3102
, 531
11-5
3113
, 550
10, 5
5020
, 55
030,
600
10, 6
0020
, 600
31, 6
0032
, 600
40, 6
0091
,60
092,
610
00, 6
2000
, 630
10, 6
3021
, 630
22, 6
3030
, 64
010,
640
20, 6
5000
, 660
00, 6
7010
, 670
20, 7
0001
, 70
002,
710
10, 7
1020
, 710
30, 7
2010
, 720
20, 7
3000
, 74
011,
740
12, 7
4021
, 740
22, 7
4030
, 740
40, 7
4050
, 74
060,
740
90, 7
5011
-750
17, 7
5020
, 800
11, 8
0012
, 80
090,
850
11-8
5013
, 850
20, 8
5030
, 900
00, 9
1010
, 91
020,
910
91, 9
1092
, 920
11-9
2015
, 920
20, 9
2030
, 92
040,
930
10, 9
3020
, 930
30, 9
3091
, 930
92, 9
5000
, 99
000
35
Table B2—Cross-Sectional Wage Regressions - 1991 and 2000 Census
Year 1991 2000
Age 0.057 0.061(0.000)** (0.000)**
Age2 / 1000 -0.575 -0.601(0.004)** (0.004)**
Female -0.392 -0.335(0.001)** (0.001)**
Inner City 0.116 0.101(0.001)** (0.001)**
RaceBrown (parda) -0.136 -0.131
(0.001)** (0.001)**Black -0.200 -0.173
(0.002)** (0.001)**Asian 0.154 0.122
(0.006)** (0.006)**Indigenous -0.176 -0.112
(0.010)** (0.006)**
Married 0.186 0.163(0.001)** (0.001)**
Fixed EffectsYears of Education (18) X XIndustry (21) X XMicroregion (494) X X
Observations 4,721,996 5,135,618R-squared 0.518 0.500
Robust standard errors in parentheses+ significant at 10%; * significant at 5%; ** significant at 1%Omitted category: unmarried white male with zero years of education, outside inner city,
working in agriculture
dependent variable: log wage = ln((monthly earnings / 4.33) / weekly hours) at main job
0
0.5
1
1.5
2
2.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Educ
atio
n fix
ed e
ffect
est
imat
e(r
elat
ive
to n
o ed
ucat
ion)
Years of education
1991
2000
36