Panel Seminar 2012 spring

パネルデータ解析セミナー

慶應義塾大学パネル調査共同研究拠点

2013 ^{年3月4日～7日}

I Stata 1

1 3

1.1 Stata . . . 3

1.2 . . . 4

1.2.1 Stata dta . . . 4

1.2.2 csv . . . 4

1.3 . . . 5

1.3.1 . . . 5

1.3.2 . . . 5

1.4 . . . 5

1.4.1 . . . 5

1.4.2 . . . 6

1.4.3 . . . 6

1.4.4 . . . 6

1.5 . . . 6

1.5.1 . . . 7

1.5.2 . . . 7

1.6 . . . 7

1.7 do log . . . 7

1.7.1 do . . . 8

1.7.2 do . . . 8

1.7.3 log . . . 9

1.8 . . . 9

1.9 . . . 10

1.10 . . . 10

1.11 : . . . 11

1.12 Stata . . . 13

2 15 2.1 . . . 15

2.2 . . . 16

2.3 : . . . 17

2.4 . . . 17

2.4.1 . . . 18

2.4.2 . . . 18

2.4.3 . . . 18

2.4.4 . . . 19

2.5 . . . 20

2.5.1 . . . 20

2.5.2 . . . 21

2.5.3 . . . 21

2.6 : . . . 22

2.7 Stata . . . 23

II KHPS 25

3.1.1 . . . 28

3.1.2 . . . 29

3.2 . . . 29

3.2.1 . . . 29

3.2.2 . . . 30

3.3 Stata . . . 31

4 33 4.1 . . . 33

4.2 . . . 33

4.3 . . . 34

4.3.1 . . . 34

4.3.2 . . . 35

4.4 . . . 35

4.4.1 OLS . . . 35

4.4.2 Postestimation . . . 37

4.5 . . . 38

4.5.1 . . . 39

4.5.2 . . . 39

4.5.3 . . . 40

4.6 Stata . . . 41

5 43 5.1 . . . 43

5.2 . . . 43

5.2.1 . . . 43

5.2.2 . . . 44

5.3 . . . 46

5.3.1 . . . 46

5.3.2 . . . 46

5.3.3 . . . 47

5.4 . . . 49

5.4.1 OLS . . . 49

5.4.2 (Heckit) . . . 50

5.5 Stata . . . 51

6 (1) 53

6.1 . . . 53

6.2 . . . 54

6.2.1 . . . 54

6.2.2 DID . . . 54

6.3 . . . 56

6.3.1 . . . 56

6.4 . . . 58

6.4.1 . . . 58

6.4.2 . . . 59

6.4.3 . . . 60

6.5 Stata . . . 61

7 (2) 63 7.1 . . . 63

7.2 . . . 63

7.3 Pooled OLS, , . . . 64

7.4 Hausman . . . 65

7.5 Stata . . . 65

8 (3) 67 8.1 DID . . . 67

8.2 . . . 67

8.2.1 . . . 67

8.2.2 . . . 68

8.3 . . . 69

8.4 . . . 69

8.5 DID . . . 70

9 (1) 73 9.1 . . . 73

9.2 . . . 74

9.3 . . . 74

9.4 . . . 75

9.5 . . . 76

10 (2) 77 10.1 . . . 77

10.1.1 . . . 77

10.1.2 . . . 79

10.2 . . . 80

10.2.1 Cox . . . 80

10.2.2 . . . 81

10.2.3 . . . 82

10.2.4 Kaplan-Meier . . . 82

10.2.5 Cox . . . 83

10.3 Stata . . . 84

(Keio Household Panel Survey, KHPS)

(1)

(2) (3) 3

(Keio Household Panel Survey, KHPS)

(a) KHPS

KHPS

Stata Stata

(b)

KHPS

(c)

Stata

1.

• (2006) .

• (2005) .

• Hsiao, C. (2002), Analysis of Panel Data, Cambridge University Press. , 2007

2.

• Colin R. McKenzie (2009) .

• (2009) .

• Wooldridge, J.M. (2001), Econometric Analysis of Cross Section and Panel Data, MIT Press.

• Cameron, A.C. and P.K. Trivedi (2005), Microeconometrics: Methods and Applications, Cam- bridge University Press.

3. Stata

• (2010) Stata .

• (2007) Stata

.

• Baum, C.F. (2006), An Introduction to Modern Econometrics Using Stata, Stata Press.

• Cameron, A.C. and P.K. Trivedi (2008), Microeconometrics Using Stata, Stata Press.

(Keio Household Panel Survey, KHPS) KHPS

(1) KHPS

(2) KHPS

KHPS

http://www.gcoe-econbus.keio.ac.jp/-keio-household-panel-survey-khps-2.html

1. 2. 3. KHPS

CSV →

H → main

PC

Stata

1

Stata

1.1 Stata

Stata 4

1.1: Stata

(1) Results (2) Command (3) Review

Command (4) Variables

Command (5)

1.2

3

1. Stata .dta

2. .csv

3.

1.2.1 Stata dta

.dta (1) (2) Command

Stata (2)

Stata 1.1 use

H data day1 test04.dta

1. Command

use H:\main\day1\test04.dta

2. Variables test04.dta

H:\main\day1\test04.dta H

→ [main] → [day1] day1

test04.dta

“u H:\main\day1\test04.dta”

1.2.2 csv

dta csv

csv CSV

Stata

1.2 insheet

H main day1 test04.csv

1. Command clear

2. Command

insheet using H:\main\day1\test04.csv

1.3

1.3.1

1.3 set memory Stata

Stata 300

set memory 300m[, permanently]

“, permanently” set memory Stata

[ ]

1.4 set maxvar Stata

Stata 3,000

set maxvar 3000[, permanently]

Stata/SE 5,000

32,767

1.3.2

1.2

cd H:\main\day1 insheet using test04.csv

1.4

1.4.1

1.5 Browse

1.6 list

v1 v3 result

list v1 v3 Stata

result “more” Enter Space

“q”

1.4.2

1.7 describe

describe describe

1.4.3

1.8 summarize

summarize v1 v3 v4 v5 4

summarize v1 v3 v4 v5

summarize

1.4.4

1.9 tabulate

tabulate v6

tabulate 2

tabulate v4 v5

1.5

insheet v1, v2, v3...

1

Variables label

1.5.1

1.10 label variable 1. v6

label variable v6 " " 2. variables

1.5.2

1.11 rename

1. v6 birthy

rename v6 birthy 2. variables

1.6

Stata 1.12 save

H main day1 test04

save H:\main\day1\test04.dta

“, replace”

“, replace”

save H:\main\day1\test04.dta, replace

1 “saveold”

saveold H:\main\day1\test04old.dta

1.7 do log

command

(interactive use) do Stata

1 CSV 1

1.2: do

do

log Stata Stata Result

log Result log

1.7.1 do

do

“.do” 1.2 do 1.2 1

* test04.dta Stata

“*” “/*” “*/”

1.7.2 do

1.13 do

do Stata 3

1 do

2 → “Do...” do

3 Command do : do

do H:\main\day1\ex1 13.do

1.3: Stata

1.7.3 log

1.14 log

log log

1 [File]-[Log]-[Begin...]

2 Command log

log using H:\main\day1\test.log

log “, append”

“, replace”

log “log close” log

log log

Stata

1.8

Stata PC

1.15

2

1 [Help]-[Stata Command...]

2 Command (cmd )

help cmd summarize

“Syntax” “Description”

“Options” “Examples”

1.16

2

1 [Help]-[Search...]

• “Search documen-

tation and FAQs” Stata

“Search net resources” web

“Search all” 2

2 Command word

search word [word... ]

• 1 “local” “net” “all”

1.9

summarize (syntax)

Stata

cmd [varlist ] [=exp ] [weight ] [if ] [in ] [, options ]

cmd varlist

varname

=exp weight if in options

•

“,”

• [ ]

•

1.10

Stata

result 1.4 “file test04.dta

not found” test04.dta

1.4: Stata

1.5:

r(601)

1.5

1.11 :

1.17 do

CSV 2 (test04.csv test05.csv) 1

do 1.6

1. CSV ( : day1 )

2. 3. 2

• “test varlist.xls”

• 2

4.

5. Stata : day1 : test04.dta, test05.dta

1.6: do

1ページ目 2ページ目

* test04.dtaとtest05.dtaを 作 成 clear

set memory 300m set maxvar 3000 set more off cd H:\main\day1

insheet using test04.csv log using test0405.log, replace des

rename v1 x1 rename v3 x3 rename v4 x4 rename v5 x5 rename v6 x6 rename v11 x11 rename v106 x106 rename v751 x306 rename v752 x321 rename v754 x323 label var x1 "id" label var x3 " ネ ル 回 数" label var x4 "配 偶 者 有 無" label var x5 "性 別" label var x6 "生 年"

label var x11 "就 学 就 業 状 況" label var x106 "最 終 学 歴" label var x306 "健 康 状 態" label var x321 "タ コ の 喫 煙"

label var x323 "タ コ 喫 煙 本 数" save test04.dta, replace clear

insheet using test05.csv des

rename v1 x1 rename v3 x3 rename v4 x4 rename v5 x5 rename v6 x6 rename v11 x11 rename v306 x306 rename v321 x321 rename v323 x323 label var x1 "id" label var x3 " ネ ル 回 数" label var x4 "配 偶 者 有 無" label var x5 "性 別" label var x6 "生 年"

label var x11 "就 学 就 業 状 況" label var x306 "健 康 状 態" label var x321 "タ コ の 喫 煙" label var x323 "タ コ 喫 煙 本 数" save test05.dta

log close clear

1.18 KHPS

KHPS

Stata main KHPS2004.csv

KHPS2009.dta do

1. CSV 2. 3.

• 3

v1 year

v2 resp

v3 hhid

• CSV

4.

5. dta ( : main : KHPS2004.dta, KHPS2005.dta,

KHPS2006.dta, KHPS2007dta, KHPS2008.dta, KHPS2009.dta)

1.12 Stata

cd

clear Stata

describe

do do

help

insheet CSV (insheet using XXX.csv)

label variable list

log log (log using XXXX.log) (log close)

rename

save dta

saveold 1

search set memory set maxvar summarize tabulate

use dta

2

2.1

2.1

x1 ID (test04.dta)

(pref2004.dta) 2

2.1 merge Stata

merge 2.1

1. day1 CSV test pref04.csv Stata (insheet

)

2. x1 sort

sort x1

3. Stata : day1

: test pref04.dta

4. (clear ) 1.17

test04.dta

5. test04.dta x1

6. test04.dta

merge x1 using H:\main\day1\test pref04.dta

7.

save H:\main\day1\test04.dta, replace

• replace

8. 2005

merge syntax

merge varlist using filename

varlist filename

2.1:

x1 x3 x4 x1 pref_id

1 1 1 1 18

2 1 2 接続 2 6

←

4004 1 2 4004 40

4005 1 1 4005 23

省略 test_pref04.dta test04.dta

省略

• ID varlist

merge merge

2 3

merge=1 test04.dta

merge=2 test pref04.dta

merge=3

1 1 1

2.2

2.2

2.1 test04.dta test05.dta

2.2 append

Stata append 2.2

1. 2004 (test04.dta)

use H:\main\day1\test04.dta

2. 2004 (test04.dta) 2005 (test05.dta)

append using H:\main\day1\test05.dta

3. : day1 : testpanel.dta

save H:\main\day1\testpanel.dta append

2004

v751 2005 v306

1.17 x306

2.2:

x1 x3 x4 x106

1 1 1 2

2 1 2 2

4005 1 1 4

x1 x3 x4

1 2 1

2 2 2

4005 2 1

↑接続

test04.dta

省略

test05.dta

省略

2004 2005

2005

2.3 :

2.3

1.18 KHPS2004.dta KHPS2009.dta 6 do

1. merge main

pref2004.csv pref2009.csv

• insheet v1 v3

hhid, pref id, pref

• merge hhid

2. append 6 : main

: KHPS seminar.dta 3. summarize describe

2.4

2.4.1

2.4 generate

generate 2.2 testpanel.dta

(year) x3

generate year=2003+x3

year (x6) (age)

generate age=year-x6

generate syntax

generate newvar =exp [if ] [in]

newvar =exp

age year v6

=exp help functions

2.4.2

2.5 replace

replace

(if) x5

1 2

1 0

x5 2 0

replace x5=0 if x5==2

2.4.3

2.6 if

• (==) =

replace x5=0 if x5==2

• (!= ~=)

replace x5=1 if x5~=1

• (>, <, >=, <=)

– old x6 1939 1

gen old=1 if x6<=1939

• (&)

– stud x11 2 7 1

gen stud=1 if x11>=2 & x11<=7

• (|)

– uni x106 4 5 1

gen uni=1 if x106==4|x106==5

• “&” “|”

– smok x3 x321 1 x3 2 x321 1 2

1

gen smok=1 if (x3==1&x321==1)|(x3==2&(x321==1|x321==2))

2.7 replace if

2.2 testpanel.dta (x321)

1 2

9

Stata

“.”

replace x321=. if x321==9 testpanel.dta

2.4.4

2.8 drop keep

drop keep old stud

drop old stud drop

keep x2 x3 smok 3

keep x3 x4 smok

*

x* x

x

keep x*

2.3: bysort

x1 x3 x106 educ

1 1 4 4

1 2 . 4

2 1 2 2

2 2 . 2

3 1 3 3

3 2 . 3

clear testpanel.dta

2.5

2.5.1

2.9 bysort bysort

bysort x5: tab x106

tabulate bysort

2004 (x106) 2005

educ 2.3

bysort x1: egen educ=max(x106)

egen generate (extentions to

generate) max( )

bysort egen educ=max(x106) x106

educ

bysort x1:

max( ) (.) 1 (x1==1)

max(x106) 4

2.5.2

2.10 xtset

xtset x1 x3, yearly

replace x106=L.x106 if x3==2 L.

2 (x3==2) x106

L. 1 (L.xit= x_i,t−1)

L2. 2 (L2.xit= x_i,t−2)

F. 1 (F.xit= xi,t+1)

F2. 2 (F 2.xit= xi,t+2)

D. (D.xit= xit− x_i,t−1)

D2. 2 (D2.xit= (xit− x_i,t−1) − (x_it−1− x_i,t−2))

2.5.3

2.11 foreach

foreach testpanel.dta

sum x323 if x3==1 ( 1 sum x323 if x3==2 2

2 2

x3==1 x3==2 foreach

foreach n of numlist 1 2{ sum x323 if x3==‘n’ }

foreach 1 2

1 “n” 1 2

2 1

‘n’

1 2

foreach syntax

foreach lname of listtype list{ commands referring to ‘lname’ }

• lname

n

• listtype numlist

• list 1

2

1(2)9 1 2 9 (1, 3, 5, 7, 9)

• ‘lname’

‘ ’

• { }

2.12

testpanel.dta x11, x106, x306 3 99

2.5 .

3

replace x11=. if x11==99 replace x106=. if x106==99 replace x306=. if x306==99

x11, x106, x306 foreach

foreach v of varlist x11 x106 x306{ replace ‘v’=. if ‘v’==99

}

2.11 listtype varlist

list

2.6 :

2.13

2.3 KHPS seminar.dta

2.1

2.1:

作成する変数 (label) 変数名 元 る変数

年齢 age year, v6

女性ダミー female v5

現在の会社で の勤務開始年 wkstart v34

現在の会社で の勤続年数 tenure wkstart, v48

仕事からの収入 income v36

2.7 Stata

append (append using filename )

merge (merge varlist using filename )

generate replace drop keep bysort xtset foreach

KHPS

3

Stata

KHPS seminar.dta 3.1

3.1

• 2004 2.9

2005

• (self) 1

0

• (emp) 1

0

• (nonreg)

1

3.1: 3.1

変 数 名 元 に る 変 数

最 終 学 歴 ダ ミ ー: 中 学 校 jrhigh v27

高 校 high v27

短 大 高 専 jrcol v27

大 学 大 学 院 col v27

そ の 他 other v27

就 業 形 態 ダ ミ ー : 自営業 self v32

被雇用 emp v32

非正規雇用 nonreg v32, v33

作 成 す る 変 数 label

3.1

3.1.1

3.2

tabulate

tab female emp

% column % row

% nofreq

missing Stata

3.3 3.2

tab female emp, col tab female emp, row tab female emp, missing

3.4 2

1. (age) 20 (agerank)

• 20 39 20 40 59 40 60 79 60

2. % %

3. (tenure) 10

• 0 9 0 10 19 10 20 29 20 30 39

30 40 40

4. (income) 200

• 200 0 200 400 200 400

600 400 600 800 600 800 1000

800 1000 1000

5.

gen agerank=.

replace agerank=20 if age>=20&age<40 replace agerank=40 if age>=40&age<60 replace agerank=60 if age>=60&age<80

3.1.2

summarize sum v4 v5 v6...

3.5

summarize (income)

sum income

sum income if year==2004 sum income if female==1

summarize Obs Mean Std.Dev

Min Max detail

3.6

3.2

Stata

3.2.1

histogram histogram varname [if ] [in ] [weight ] [, options ]

3.7 (income)

hist income if

hist income if income<=1000

3.2:

プ シ ョ ン 内 容

frequency 度 数 表 示

percent %表 示

title(title) グ ラ フ に 見 出 し を 付 け る xtitle(axix_title) 横 軸 の 見 出 し

ytitle(axix_title) 縦 軸 の 見 出 し

xlabel(a(b)c) 横 軸 の 表 示 形 式 a 目 盛 の 最 小 値 、c 最 大 値 、b 目 盛 間 隔 ylabel(a(b)c) 縦 軸 の 表 示 形 式 a 目 盛 の 最 小 値 、c 最 大 値 、b 目 盛 間 隔

start(#) 横 軸 の 最 小 値

width(#) ヒ ス ト グ ラ ム の 間 隔 addlabel 数 値 ラ ベ ル を 追 加 normal 正 規 分 布 を 追 加 で 表 示 kdensity ー ネ ル 密 度 を 追 加 で 表 示

Graph Editor

3.8 (income)

1. 1000

2. ”Figure 3-1”

3. x ”Annual labor earnings” y ”Frequency”

4. 100

hist income if income<=1000, freq title(Figure3-1) xtitle(Annual labor earnings) ytitle(frequency) width(100) addlabel

Stata

3.9

1 50 2

“Figure 3-2” x “Current job tenure” 3 5 4

5 6 x 1 50 5

hist tenure if tenure<=50, title(Figure3-2) xtitle(Current job tenure) width(5) addlabel kdens xlabel(1(5)50)

3.2.2

2 scatter scatter

scatter y varname x varname [if ] [in ] [,options ]

3.10

1. loginc

gen loginc=log(income)

2. scatter

scatter loginc tenure

title, xlabel, xtitle scatter

Graph Editor

3.11

1 “Job tenure and

annual earnings” x “Current job tenure” y “log(annual earnings)”

2 (small circle) 3 (black) 4

(small) (2) (4) graph editor

scatter loginc tenure, title(Job tenure and annual earnings)

xtitle(Current job tenure) ytitle(log(annual earnings))

graph box( ) graph pie

( ) Stata

3.3 Stata

histogram scatter

4

4.1

(1999) (2001)

KHPS 2007

4.2

KHPS (Ordinary Least Square, OLS)

yi xki

yi= α + β1x1i+ β2x2i+ · · · + ui (4.1)

ui 2 (α, β1, β2, · · · )

Stata regress

regress depvar varlist [if ] [in ] [, options ]

depvar varlist

(2007) (2008)

2

4.1:

差 = 23年

学 卒 時 調 査 時 点

18歳 41歳

1981年 2004年

gradyear = 2004 - 23

4.3

4.3.1

4.1

(grad unemp)

1. (gradage)

2. (year) (age) 1

(gradyear)

3. 2 gradyear

• 1

gen gradage=.

replace gradage=15 if jrhigh==1 replace gradage=18 if high==1 replace gradage=20 if jrcol==1 replace gradage=22 if col==1 label var gradage " "

• 2

gen gradyear=year-(age-gradage) replace gradyear=. if age<gradage label var gradyear " "

• 3

– unemp.csv CSV

Stata insheet using H:\main\day1\unemp.csv sort gradyear

save H:\main\day1\unemp.dta

– KHPS

merge gradyear using H:\main\day1\unemp.dta label var grad unemp " (%)"

4.3.2

4.2 2

(age) (tenure) 2

gen agesq=age^2 gen tenuresq=tenure^2

4.3

sum income loginc female jrhigh high jrcol col age tenure self emp nonreg grad unemp

4.4

4.4.1 OLS

4.4

(loginc) 2

2 2004

reg loginc age agesq female high-col tenure tenuresq nonreg grad unemp if year==2004&emp==1

reg loginc age agesq high-col tenure tenuresq nonreg grad unemp if year==2004&emp==1&female==0

reg loginc age agesq high-col tenure tenuresq nonreg grad unemp if year==2004&emp==1&female==1

high - col

Variable high col

4.1:

Stataの 表 記 内 容

Number of obs. サ ン プ ル サ イ ズ

R-squared 決 定 係 数

F _{F検定統計量 (H0}: 回帰係数 = 0)

Prob>F _{F検定のp値}

Adj. R-squared 自 由 度 修 正 済 み 決 定 係 数

Coef. 回 帰 係 数

Std.Err. 回 帰 係 数 の 標 準 誤 差

t 回 帰 係 数 の_t値

P>|t| 回 帰 係 数 の_p値 95% Conf. Interval 係 数 の_{95%信頼区間}

Stata 4.1

• (semi-log)

1

1

log(y) = α + βx + u (4.2)

y

y = e^α+βx+u (4.3)

x

∂y

∂x ^{= βe}

α+βx+u_{= βy (4.4)}

β

β = ^∂y/∂x

y ^(4.5)

∂y/∂x x y y

y

−0.072 1 (1%) 7.2%

• 2

y x

y = a + bx + cx² (4.6)

x

∂y

∂x ^{= b + 2cx (4.7)}

x

0.0322 2 −0.0004

4.2:

2 4.2

x 40

Stata

White (1980) robust

robust 4.5 White (1980)

reg loginc age agesq high-col female tenure tenuresq nonreg grad unemp if year==2004&emp==1,robust

4.6 e(sample)

sum income loginc female jrhigh high jrcol col age tenure self emp nonreg grad unemp if e(sample)

e(sample)

1 if e(sample) if e(sample)==1

ereturn list

4.4.2 Postestimation

predict

predict newvar [if ] [in ] [,options ]

4.2: 4.8

作 成 す る 変 数 変 数 名 元 る 変 数

業 種 ダ ミ ー: 農 林 漁 業 鉱 業 ダ ミ ind1 v30

建 設 業 ダ ミ ind2 v30

製 造 業 ダ ミ ind3 v30

卸 売 小 売 飲 食 宿 泊 業 ダ ミ ind4 v30

金 融 保 険 不 動 産 ダ ミ ind5 v30

運 輸 業 ダ ミ ind6 v30

情 報 通 信 業 ダ ミ ind7 v30

電 気 ガ ス 水 道 医 療 福 祉 業 ダ ミ ind8 v30

教 育 学 習 支 援 そ の 他 サ ビ ス 業 ダ ミ ind9 v30

公 務 及 び そ の 他 業 種 ダ ミ ind10 v30

企 業 規 模 ダ ミ ー: 企 業 規 模 1-4人 ダ ミ fmsz1 v31

企 業 規 模 5-29人 ダ ミ fmsz2 v31

企 業 規 模 30-99人 ダ ミ fmsz3 v31

企 業 規 模 100-499人 ダ ミ fmsz4 v31

企 業 規 模 500人 以 上 ダ ミ fmsz5 v31

企 業 規 模 官 公 庁 ダ ミ fmsz6 v31

労 働 組 合 加 入 ダ ミ ー union v35

有 配 偶 ダ ミ ー marr v4

newvar predict

residual help regress postestimation

4.7

predict pred loginc

label var pred loginc "loginc " predict pred res, residuals

label var pred res " "

4.8

(v30) (v31) (v4)

robust 4.2

4.5

4.5.1

0 1

Stata probit logit

4.9

(nonreg) 2

2004

probit nonreg age agesq female high-col grad unemp if year==2004&emp==1

probit nonreg age agesq high-col grad unemp if year==2004&emp==1&female==0 probit nonreg age agesq high-col grad unemp if year==2004&emp==1&female==1

4.5.2

y x y = 1

Pr(y = 1|x) = Φ (xβ) (4.8)

Φ (·) k xk

∂ Pr(y = 1|x)

∂xk

= φ (xβ) βk (4.9)

φ (·) βk k (4.9)

xk

x

¯

x ¹

dprobit ²

4.10

4.9 dprobit

dprobit nonreg age agesq female high-col grad unemp if year==2004&emp==1

1 Average Partial Effect (APE)

xⁱ APE

AP E=^∑

i^{φ(xiβ) βk}

2 1 0

Stata

*

4.5.3

(1) (2)

(3) 3

4.11

wkst 1

2 3

wkst

4.11 3

Stata mlogit 4.12

(wkst) 2

2004

mlogit wkst age agesq female high-col grad unemp if year==2004

mlogitwkst age agesq high-col grad unemp if year==2004&female==0 mlogitwkst age agesq high-col grad unemp if year==2004&female==1

4.13

mlogit mfx, predict(p outcome(# ))

# 4.12

mfx, predict(p outcome(1))

nose

4.6 Stata

ereturn list dprobit mfx mlogit predict probit

regress 2

robust (regress )

5

5.1

1

1.

2

2.

3

KHPS

KHPS

4

5.2

Oaxaca (1973)

Oaxaca (1973), Neuman and Oaxaca (2004) KHPS

5.2.1

ln Wi= β0+ β1Educi+ β2Expi+ β3T enurei+ ui (5.1)

1Mincer (1985) (2004) 100

79.4 2003 80.6 1999 66.8 2003

2OECD (2002)

3Oaxaca (1973) (1991) (2003) Johnes and Tanaka (2008) Miyoshi (2008)

4 (2008)

i ln Wi

i 1 Educ T enure

Exp

Mincer (1962)

ln W, Educ, T enure, Exp β0, β1, β2, β3

(5.1) F emale

ln Wi = βi0+ βi1Educi+ βi2Expi+ βi3T enurei+ βi4F emalei+ ui (5.2) F emale β4

Educ Exp T enure

ln Wi= β0^m+ β1^mEduci+ β2^mExpi+ β3^mT enurei+ u^m_i (5.3) ln Wi = β0^f+ β1^fEduci+ β2^fExpi+ β3^fT enurei+ u^f_i (5.4)

H0=





 β^m1

β^m2

β^m3





⁼





 β1^f

β₂^f β3^f







5

5.2.2

(5.1)

(ln W ) Stata

(5.1)

(5.2) F emale

5Oaxaca (1973)

Stata (oaxaca)

(5.1) β0, β1, β2, β3

Heckman (1979) (5.1)

Heckman (1979)

i

d^∗i = Ziα + εi

di= 1 [d^∗i > 0] (5.5)

di i di = 1 di= 0

1 [·] (index function) [·] 1

d^∗_i > 0 di= 1 i di= 0

i d^∗_i

εi

(5.4)

i Zi Zi

ln Wi^∗= Xiβ + ui

d^∗i = Ziα + εi

di= 1 [d^∗i > 0] ln Wi= di(ln W_i^∗) [ ui

εi

]

∼ N ID (0, Σ)

Σ =

[σ² ρσ ρσ 1

]

(5.6)

Xi Zi

Zi Xi

Heckman (1979)

(5.2) (5.3) (5.4)

5.3

0 3 4 6

5.3.1

KHPS KHPS

5.1

wage logwage

1. (v38) 1 (v39)

mearn

2. v46 wkhr

3. 1 2 wage

4.

wage

5. logwage

• (mearn)

gen mearn=1000*v39 if (v38==1|v38==2)&(v39~=88888&v39~=99999)

label var mearn " , "

• (wkhr)

gen wkhr=v46 if v46~=888&v46~=999 label var wkhr " "

5.3.2

(emp==1)

(v38==1) (v38==2) (v38==4)

5.2

wdummy wdummy

1 0

gen wdummy=.

replace wdummy=1 if wage~=. replace wdummy=0 if v32==8

5.3.3

(netfin) (sincome) 0 3

(noku3) 4 6 (noku6) (exp)

1 5.3

netfin sincome

•

gen sincome=v37

replace sincome=0 if v37==88888 replace sincome=. if v37==99999 label var sincome " "

•

gen saving=v457

replace saving=. if v457==999999 label var saving " " ...

5.4

p.1 2

3 noku3 2 10

3

noku3 4 6

(noku6) (noku3)

5.1: KHPS

1. noku3

gen noku3=0

2. p.1 2 3

noku3 1 (noku3=noku3+1)

replace noku3=noku3+1 if ((v8==1&marr==0)|(v8==2&marr==1))

&v9>=year-3&(v9~=9999&v9~=8888)

2 1

3. 3

replace noku3=noku3+1 if ((v10==1&marr==0)|(v10==2&marr==1))

&v11>=year-3&(v11~=9999&v11~=8888) ...

• 2 10

foreach

gen noku3=0

foreach n of numlist 8(2)24{ local k=‘n’+1

replace noku3=noku3+1 if ((v‘n’==1&marr==0)|(v‘n’==2&marr==1))

&v‘k’>=year-3&(v‘k’~=9999&v‘k’~=8888) }

local k=‘n’+1 k foreach

n 1

KHPS 5.1

18 v151 1

19 v152 1

18 v202 1 19 v203 1

KHPS 5.5

exp

2 (expsq)

exp

gen exp=0

replace exp=exp+1 if v151==1|v202==1 replace exp=exp+1 if v152==1|v203==1 ...

foreach

gen exp=0

foreach n of numlist 151(1)201{ local k=‘n’+51

replace exp=exp+1 if v‘n’==1|v‘k’==1 }

5.4

(5.2)

(5.3) (5.4) ⁶

local indep "exp expsq tenure tenuresq high jrcol col other"

‘indep’ "exp expsq tenure tenuresq high jrcol

col other" Stata 2004

5.4.1 OLS

1 regress (5.2) (5.3) (5.4)

5.6 (5.2)

(5.2) reg logwage ‘indep’ female if year==2004

6 (5.2) (5.3)

(5.3) (5.4)

5.7 (5.3) (5.4) (5.3)

reg logwage ‘indep’ if year==2004&female==0

mb mv

matrix mb= e(b) matrix mv= e(V)

matrix e(b) e(V)

(5.4)

reg logwage ‘indep’ if year==2004&female==1 matrix fb=e(b)

matrix fv=e(V)

5.8

matrix teststats=(mb-fb)*inv(mv+fv)*(mb-fb)’ matrix list teststats

(5.2) teststats

7

5.4.2 (Heckit)

(5.5) (5.5)

regress Stata

(5.5) (5.4)

5.9

local indep2 "exp expsq high jrcol col other noku3 noku6 netfin sincome" (wdummy)

7 Hausman

(5.5) heckman

5.10 (Heckit)

heckman

heckman logwage ‘indep’ female if year==2004, select(wdummy = ‘indep2’ female)

heckman logwage ‘indep’ if year==2004&female==0, sel(wdummy = ‘indep2’) matrix mb=e(b)

matrix mv=e(V)

heckman logwage ‘indep’ if year==2004&female==1, sel(wdummy = ‘indep2’) matrix fb=e(b)

matrix fv=e(V)

matrix teststats=(mb-fb)*inv(mv+fv)*(mb-fb)’ matrix list teststats

5.11 1

5.5 Stata

heckman (Heckit)

local matrix matrix list

6 (1)

6.1

1 _{6.1 x ID}

N

1, 2, · · · , T ²

3

6.1

xit i t

xit i t

1. 2.

6.1:

観 測 時 点

パ ネ ル _{ID 1 2 3} …

…

1

^…

^…

2

^…

^…

3

^…

^…

： ： ： ： ： ：

i xi₁ xi₂ xi₃

…

…

： ： ： ： ： ：

N x_N1 x_N2 x_N3

…

…

1 (longitudinal data)

2

7-1

3 (repeated

cross-section data)

6.2: N

_{= 2,693}

I II III IV V Total

I 66.5% 24.4% 5.2% 2.0% 2.0% 100.0%

II 26.7% 47.2% 19.1% 6.0% 1.0% 100.0%

III 10.4% 23.3% 42.3% 17.9% 6.0% 100.0%

IV 4.6% 6.6% 23.1% 47.8% 18.0% 100.0%

V 2.1% 2.6% 4.9% 20.7% 69.7% 100.0%

20.0% 20.2% 19.8% 19.7% 20.3% 100.0%

所得五分位 2004年

Total

所得五分位

2005年

注: KHPS2004およびKHPS2005より作成。

6.2

6.2.1

20%

2 2

6.2 KHPS

6.2 2004 2005

4 ₂₀₀₄

100

20% 6.2

2004 2/3 2005

฀ _10%

6.2.2 DID

DID (Difference-in-Differences, )

(y)

t = 1 t = 2

4 0.5

KHPS (2007)

y^T_t (t = 1, 2)

y^T1 = µ^T+ φ1

y^T2 = µ^T+ φ2+ δ (6.1)

(6.1) µ^T

φt δ ⁵

∆y^T ≡ y^T2 − y1^T = δ + (φ2− φ1) (6.2)

(6.2) Before-After comparison φ1= φ2

δ φ1 ̸= φ2 (6.2)

(φ2> φ1)

(δ = 0) (6.2)

(6.1) 2

(6.1)

y1^C= µ^C+ φ1

y2^C= µ^C+ φ2 (6.3)

(1)

(µ^T ̸= µ^C) (2)

δ = 0 2

(6.2)

∆y^C ≡ y^C2 − y^C1 = (φ2− φ1) (6.4)

DID (6.2)

(6.4)

∆y^T − ∆y^C= [δ + (φ2− φ1)] − (φ2− φ1) = δ (6.5)

i t

yit

yit= xitβ + γ1d^T + γ2d2+ δd^T · d2+ εit (6.6)

xit etc. d^T i

d2 (t = 2) 1 β, γ1, γ2, δ

5

εit (6.6) 2 δ (6.5) DID

(6.6) DID

2

DID

6 (6.6) DID

(E [d^Tεit] = 0)

(6.6) δ

7 _DID

KHPS DID 8

6.3

6.3.1

OJT (on the job training)

(OJTit) (wit)

N (i = 1, 2, · · · , N ) T (t = 1, 2, · · · , T )

log (wit) = βOJTit+ δi+ εit (6.7)

δi εit δi

δi

OJT

OJT β ⁸

(6.7) 2 Pooled OLS

uit≡ δi+ εit 2 β β^ˆOLS

βˆOLS =^{Cov (OJTit, log (wit))} Var (OJTit)

= ^{Cov (OJTit, βOJTit+ uit)} Var (OJTit)

= β +^{Cov (OJTit, uit)}

Var (OJTit) ^(6.8)

6 DID Bertrand et al. (2004)

7

(regression discontinuity)

8 β

(N ) β

6.1: OLS β

(6.8) β^ˆOLS OJTit uit (Cov (OJTit, uit) = 0)

β β^ˆOLS

δi

OJT OJTit δi (uit)

(Cov (OJTit, uit) > 0) OJT (β > 0)

OLS

6.1 δi OJTit δi

2 OJTit OLS OJT

δi

OLS OJTit

(Fixed-Effects model, FE) (6.7)

T

∑

t=1

log (wit) =β

T

∑

t=1

OJTit+

T

∑

t=1

δi+

T

∑

t=1

εit (6.9)

T

log (wi) = βOJTi+ δi+ ¯εi (6.10)

log (wi) = ¹ T

T

∑

t=1

log (wit) , OJTi= ¹ T

T

∑

t=1

OJTit, ε¯i= ¹ T

T

∑

t=1

εit (6.11)

6.2:

δi δi

(6.10) (6.7)

log (wit) − log (wi) = β(OJTit− OJTi) + (εit− ¯εi)

log (wit)^∗= βOJTit^∗+ ε^∗it (6.12) (6.11) (6.7)

(within estimator) ⁹ (6.11)

δi OJTit OLS

(Random-Effects model, RE)

OLS δi

β KHPS

OLS, FE, RE 7

6.4

6.4.1

KHPS

6.2 6.2 hhid ID

year 2004 2009 6

9 1 (First-Difference estimator, FD)

(Least-Squares Dummy Variable estimator, LSDV)

6.3: KHPS (xtdes)

6.1

1. KHPS seminar.dta

use H:\main\KHPS seminar.dta

2.

drop if resp==0

3. xtset

xtset hhid year, yearly

xtset hhid ID year

yearly

6.4.2

6.2 xtdes

xtdes

6.3 5 57% 2 6

17% 11% 6% 5% 4%

5

6.4.3

1

x y

gen y=L.x

“L.” x (L.xit= x_i,t−1)

Stata

L. 1 (L.xit= x_i,t−1)

L2. 2 (L2.xit= x_i,t−2)

F. 1 (F.xit= xi,t+1)

F2. 2 (F 2.xit= xi,t+2)

D. (D.xit= xit− x_i,t−1)

D2. 2 (D2.xit= (xit− x_i,t−1) − (x_it−1− x_i,t−2)) 6.3

(v27) 2004 2005

2005

1. 2005 x27 2004

replace v27=L.v27 if year==2005

2. 2006 2005

replace v27=L.v27 if year==2006

3. 2009

4. : jrhigh / : high / : jrcol / : col /

: other

1 3 2005 2009

foreach n of numlist 2005(1)2009{ replace v27=L.v27 if year==‘n’ }

6.4

(marr) (female) (age)

6.5

xtsum

xtsum jrhigh high jrcol col other marr female age

6.5 Stata

xtdes xtset xtsum

7 (2)

7.1

6.3

1

1

2

2

6.3

log (wit) = α + βtenureit+ xitγ + δi+ εit (7.1)

log (wit) tenureit δi i

xit

7.2

(logwage)

(emp=1)

7.1:

変 数 名 内 容

年 齢 (age) 対 象 者 の 年 齢

性 別 (female) 対 象 者 の 性 別 ダ ミ ー 女 性 = 1

最 終 学 歴 (jrhigh-other) 対 象 者 の 最 終 学 歴 ダ ミ ー 中 卒 が 基 準 非 正 規 雇 用 ダ ミ ー (nonreg) 非 正 規 雇 用 ダ ミ ー 正 規 雇 用 が 基 準

企 業 規 模 (fmsz1-fmsz6) 企 業 規 模 ダ ミ ー fmsz1: 1～4人 / fmsz2: 5～29人 / fmsz3: 30～99人 /fmsz4: 100～499人 /fmsz5: 500

人 上 / fmsz6: 官 公 庁 fmsz1が 基 準

地 域 ダ ミ ー (rg1–rg8) 全 国8地 域 ダ ミ ー 北 海 道 が 基 準

市 郡 規 模 ダ ミ ー (bigcity / city) 市 郡 規 模 政 指 定 都 市 = 1 / そ の 他 の 市 =1 調 査 年 度 ダ ミ ー (year2004–year2008) 調 査 年 度 ダ ミ ー 2004年 が 基 準

1 (2006)

7.1

(rg1-rg8) (bigcity, city, town)

(year2004-year2009)

7.2 xtsum

xtsum wage logwage tenure age female jrhigh-other nonreg fmsz1-fmsz6 rg1-rg8 bigcity-town year2004-year2009 if emp==1

7.3 Pooled OLS, ,

7.3 OLS

OLS regress

reg logwage tenure age female high-other nonreg fmsz2-fmsz6 rg2-rg8 bigcity city year2005-year2009 if emp==1, robust 6.3

xtreg

xtreg depvar indepvars [in ] [if ] [, (re/fe) i(idvar ) robust]

• re fe

• i(idvar ) ID xtset

• robust White

7.4 7.3

xtreg logwage tenure age female high-other nonreg fmsz2-fmsz6 rg2-rg8 bigcity city year2005-year2009 if emp==1, re

7.5 (within estimator)

7.3

xtreg logwage tenure nonreg fmsz2-fmsz6 rg2-rg8 bigcity city year2005-year2009 if emp==1, fe 6.3

(6.12)

7.4 Hausman

δi

Hausman

βF E βRE Hausman

H0: βRE= βF E

(βF E− βRE) V_H⁻¹(βF E− βRE)^′∼ χ²(k)

VH 2 k 2

Hausman

7.6 Hausman

7.4 7.5 Hausman Stata hausman

1. 7.4 estimates store

re

estimates store re

2. 7.5 estimates store

fe

estimates store fe

3. Hausman

hausman fe re

hausman syntax

hausman name-consistent [name-efficient ] [,options ] name-consistent

name-efficient δi

7.7

7.5 Stata

estimates store

hausman Hausman

xtreg re: fe:

8 (3)

8.1 DID

DID

1

1,000

8.1

70 76 38 + 38

2004 (a)

2

3

8.2

DID

KHPS 2

8.2.1

7.3 DID (6.1)

8.1:

2004年 2005年

ト ト メ ン ト グ プ ： 有 配 偶 者 影 響 あ り 影 響 な し

コ ン ト ロ グ プ ： 無 配 偶 者 影 響 な し 影 響 な し

1 McKenzie (2006)

2 8.1 (a)

3

McKenzie (2006)

8.1: 2004

(6.3) 8.1

1 DID

2 KHPS

2004 2003

KHPS 1 2003

8.1 2004 1

8.1

4

McKenzie (2006)

8.2.2

KHPS 2004

2004

5

8.2

4 2003 DID

5 2005

8.2:

2004年 2005年

ト ト メ ン ト グ プ ：

有 配 偶 者 制 度 変 更 を 知 な か っ た

影 響 あ り 影 響 な し

コ ン ト ロ グ プ ：

無 配 偶 者 ＋ 有 配 偶 者 制 度 変 更 を 知 っ て い た

影 響 な し 影 響 な し

8.3:

変 数 名 内 容

年 齢 (age) 対 象 者 の 年 齢

配 偶 関 係 (marr) 対 象 者 の 配 偶 関 係 ダ ミ ー 有 配 偶 = 1

最 終 学 歴 (jrhigh-other) 対 象 者 の 最 終 学 歴 ダ ミ ー 中 卒 が 基 準

被 雇 用 就 業 (emp) 被 雇 用 就 業 ダ ミ ー 被 雇 用 者 = 1

非 正 規 雇 用 (nonreg) 非 正 規 雇 用 ダ ミ ー 非 正 規 雇 用 = 1

世 帯 員 数 (hhnum) 本 人 を 含 む 世 帯 員 数

配 偶 者 の 年 間 所 得 (sincome) 配 偶 者 の 年 間 所 得 無 配 偶 者 ゼ ロ

3 の 子 供 の 数 (noku3) 3 未 満 の 未 就 学 児 童 数

4 6 未 満 の 子 ど も の 数 (noku6) 4 6 未 満 の 子 ど も の 数

市 郡 規 模 ダ ミ ー (bigcity / city) 市 郡 規 模 政 指 定 都 市 = 1 / そ の 他 の 市 =1

8.3

Hit= α + β0d^T_i + β1d⁰⁴_t + β2d^T_i × d⁰⁴_t + xitγ + εit (8.1)

Hit (wkhr) d^Ti 1

d⁰⁴_t 2004 1 xit

8.3

d^T_i × d⁰⁴_t 2004 1

β2 8.1 8.2

β2< 0

2004 2005 2 2004

1

8.4

8.1 2006

keep if year==2004|year==2005

8.2

8.2 2

1. 2004 (treat1)

gen treat1=marr if year==2004

replace treat1=L.treat1 if year==2005

label var treat1 "Treatment 1: 2004 "

2. 2004 (treat2)

gen unaware=0

replace unaware=1 if v44==2

replace unaware=. if v44==8|v44==9|v44==. replace unaware=L.unaware if year==2005

label var unaware " = 1 "

gen treat2=unaware*treat1

label var treat2 "Treatment 2: 2004 "

8.3 2004

8.2 2004 (d^T_i × d⁰⁴t

) gen treat1 04=treat1*year2004

label var treat1 04 "Treatment 1× 2004 " gen treat2 04=treat2*year2004

label var treat2 04 "Treatment 2× 2004 "

8.4

v7 (hhnum)

8.5 2004

2004 1

gen work04=1 if year==2004&wkhr>0&wkhr~=. replace work04=L.work04 if year==2005

label var work04 "2004 1 "

8.5 DID

8.6

(6.5) tab treat1 year if female==1&work04==1, sum(wkhr)

8.7 2004

treat1 (8.1)

reg wkhr treat1 treat1 04 year2004 age marr high-other emp nonreg hhnum sincome noku3 noku6 bigcity city if female==1&work04==1,robust

8.8

treat2 8.6 8.7

8.9

8.2 DID

8.2

9 (1)

9.1

9 10 ¹

2

9.1

9.1

25

10 9.2

9.2 10 27

25

9.1:

年 齢 配 偶 関 係 初 婚 年 齢

67 既 婚 19

34 既 婚 31

24 未 婚 .

50 既 婚 22

47 既 婚 25

43 既 婚 29

27 未 婚 .

66 離 死 別 21

27 未 婚 .

48 既 婚 28

1 (event history analysis) (hazard analysis) (survival

analysis) (duration analysis)

2 (2001-02) Yamaguchi (1991)

9.2: 10 年 齢 配 偶 関 係 初 婚 年 齢

67 既 婚 19

34 既 婚 31

24 既 婚 34

50 既 婚 22

47 既 婚 25

43 既 婚 29

27 既 婚 32

66 離 死 別 21

27 既 婚 29

48 既 婚 28

9.2

3

9.1

9.3

9.1 9.1

KHPS

2004 2009

(A) (F) (A)

(B)

(C) (B)

(D)

(E)

(F)

3

9.1:

イ ベ ン ト 発 生 観 察 打 ち 切

△

○

×

ス 期 間 開 始

左 ト ン ー シ ョ ン

左 セ ン サ ン

右 セ ン サ ン

右 セ ン サ ン

観 察 さ ス 期 間

観 察 さ な い ス 期 間

観 察 開 始 観 察 打 ち 切

(A)

(B)

(C)

(D)

(E)

(F)

△

△ ×

△ ○

○

△

○

△

○

？

9.4

t = 1, 2, · · · h (t)

t t

h (t) = Pr (T = t|T ≥ t) (9.1)

T S (t) t

S (t) = Pr (T > t) (9.2)

S (0) = 1

S (t + 1) = S (t) · (1 − h (t + 1)) (9.3)

t + 1 S (t + 1) t

(S (t)) t + 1 (1 − h (t + 1))

9.3 100

Et t Rt t

100 t = 1 40

(E1= 40) t = 1 60 (= 100 − 40)

(R1= 60) t = 2 30 (E2= 30) t = 2

30 (R2= 100 − E1− E2= 100 − 40 − 30 = 30)

h (t) S (t) t = 1

R1= 60 S (1) = 60/100 = 0.6

9.3:

t =

_{0 1 2 3 4}

E_t

― _{40 30 20 10}

Rt

_{100 60 30 10 0}

h

_{(t )} ― _{0.4 0.5 1/3 1}

S

_{(t ) 1 0.6 0.3 0.1 0}

リ ス ク 期 間

t = 1 t = 0

100 t = 1 h (1) = 40/100 = 0.4

t = 2 S (2) = R2/100 = 30/100 = 0.3 h (2) = E2/R1= 0.5

(9.3) S (0) × (1 − h (1)) = 1 × (1 − 0.4) = S (1) S (1) × (1 − h (2)) = 0.6 ×

(1 − 0.5) = 0.3 N

S (t) = ^Rt N ⁼

N −^∑t_k=1Ek

N ^(9.4)

h (t) = ^Et

R_t−1 ^(9.5)

S (t) · (1 − h (t + 1)) = ^Rt N ^·

(

1 −^Et+1 Rt

)

= ^{Rt− Et+1}

N ⁼

N −^∑t+1_k=1Ek

N = S (t + 1) (9.6)

9.5

h (t)

h (t; xit) = f (xitβ) (9.7)

t = T

Li= S (t − 1) · h (t) = {_{T −1}

∏

k=1

[1 − f (xikβ)] }

· f (xiTβ) (9.8)

Li= S(T ) =

T

∏

k=1

[1 − f (xikβ)] (9.9)

10 (2)

10.1

2

h (t; xit) = ¹

1 + exp [− (λ (t) + xitβ)] ^(10.1)

xit β λ (t)

xit β

1 _{x 1 exp (β)}

KHPS

10.1.1

10.1

1 1

1

1 (wkchng)

1.

xtset hhid year, yearly

2. (wkstate)

gen wkstate=F.v48

label var wkstate " 1 "

1

h(t; x^it) 1 − h (t; x^it)

3. 1 (wkchng) gen wkchng=0

replace wkchng=1 if wkstate==4|wkstate==6 replace wkchng=. if wkstate==99|wkstate==. replace wkchng=. if emp==0

10.2 (t)

1.

• 1 9

t0109 1 9 t1014 10 14 t1519 15 19 t2029 20 29

t3060 30

2. (logt)

1.

foreach n of numlist 1(1)9{ gen t‘n’=0

replace t‘n’=1 if tenure==‘n’ replace t‘n’=. if tenure==. }

gen t1014=0

replace t1014=1 if tenure>=10&tenure<15 replace t1014=. if tenure==.

gen t1519=0

replace t1519=1 if tenure>=15&tenure<20 replace t1519=. if tenure==.

gen t2029=0

replace t2029=1 if tenure>=20&tenure<30 replace t2029=. if tenure==.

gen t3060=0

replace t3060=1 if tenure>=30&tenure<100 replace t3060=. if tenure==.

2.

gen logt=log(tenure)

(unemp)

10.3

unemp gen unemp=.

replace unemp=4.71 if year==2004 replace unemp=4.42 if year==2005 replace unemp=4.13 if year==2006 replace unemp=3.85 if year==2007 replace unemp=3.98 if year==2008 replace unemp=5.08 if year==2009 label var unemp " "

10.1.2

10.4

3 logit

robust White

or β exp (β)

1.

logit wkchng t2-t3060 age agesq female high jrcol col other nonreg fmsz2-fmsz6 unemp, robust or

2. 2

logit wkchng tenure tenuresq age agesq female high jrcol col other nonreg fmsz2-fmsz6 unemp, robust or

3.

logit wkchng logt age agesq female high jrcol col other nonreg fmsz2-fmsz6 unemp, robust or

10.5

10.4 mfx

1.

logit wkchng t2-t3060 age agesq female high jrcol col other nonreg fmsz2-fmsz6 unemp, robust

mfx

2. 2

logit wkchng tenure tenuresq age agesq female high jrcol col other nonreg fmsz2-fmsz6 unemp, robust

mfx

3.

logit wkchng logt age agesq female high jrcol col other nonreg fmsz2-fmsz6 unemp, robust

mfx

10.2

9

KHPS

10.2.1 Cox

Cox Cox

h (t; xit) = λ0(t) φ (xit, β) (10.2)

λ0(t) φ (xit, β) xit

λ0 φ λ0(t)

2 φ (·) φ (xit, β) = exp (xitβ)

2 Cox (1972, 1975) Cameron and Trivedi (2005, p.594)

Stata stphplot stcoxkm

10.2.2

2005 2005

keep if year==2005

10.6 (duration)

1. duration

18 16

gen duration=.

replace duration=v462-18 if female==0&(v461==2|v461==8)&v462~=99 replace duration=v462-16 if female==1&(v461==2|v461==8)&v462~=99

2. duration 2005

replace duration=age-18 if female==0&v461==1 replace duration=age-16 if female==1&v461==1 3.

label var duration " "

10.7 (evermarr)

Stata

1 0

gen evermarr=.

replace evermarr=1 if (v461==2|v461==8)&v462~=99 replace evermarr=0 if v461==1

label var evermarr " = 1 "

10.8 (bc40, bc50, bc60, bc70) 4

1950 50 60 70 4

1.

gen bc=.

replace bc=40 if v6>=1930&v6<1950 replace bc=50 if v6>=1950&v6<1960 replace bc=60 if v6>=1960&v6<1970 replace bc=70 if v6>=1970&v6<1990

2.

foreach n of numlist 40(10)70{ gen bc‘n’=0

replace bc‘n’=1 if bc==‘n’ }

10.2.3

stset

10.9

stset duration, failure(evermarr)

stset syntax

stset timevar, failure(failvar )

timevar failvar

help stset

10.2.4 Kaplan-Meier

Cox Kaplan-Meier

t t

9.3 S (t)

Kaplan-Meier

Nt: t Et: t

t t h (t)

h (t) = ^Et Nt

(10.3)

t 1 − h (t)

t S (t − 1)

S (t) = [1 − h (t)] S (t − 1) (10.4)

t − 1 S (t − 1) = [1 − h (t − 1)] S (t − 2)

S (t) =

t

∏

k=1

[1 − h (k)] =

t

∏

k=1

( 1 − ^Ek

Nk

)

(10.5)

Kaplan-Meier

10.10 Kaplan-Meier

stset Kaplan-Meier

sts graph

by

sts graph, by(female)

sts graph, by(bc)

10.11

10.2.5 Cox

Cox stcox

10.12 Cox

Cox

stcox female high jrcol col other bc50-bc70

stcox exp (β)

1

1

β nohr

10.13

stcox high jrcol col other bc50-bc70 if female== 0 stcox high jrcol col other bc50-bc70 if female== 1

10.14

10.3 Stata

logit

stcox Cox

sts graph Kaplan-Meier stset

[1] (2007) KHPS 3

III pp.101-129.

[2] (1999)

pp.13-42.

[3] (2008)

[4] (2005) .

[5] (2009) .

[6] (2001) 490 .

[7] (2004) .

[8] (2008)

77 .

[9] C.R. McKenzie (2006)

II pp.129-151.

[10] (2007) Stata

.

[11] (2003)

No. 158

[12] (1991)

[13] (2006) .

[14] Colin R. McKenzie (2009) .

[15] (2007) KUMQRP Discussion Paper Series, DP2007-33.

[16] (2006)

II pp.153-167.

[17] (2001-02) (1)-(15) 52(9)-53(6).

[18] Baum, C.F. (2006), An Introduction to Modern Econometrics Using Stata, Stata Press.

[19] Bertrand, M, E. Duflo, and S. Mullainathan (2004) “How Much Should We Trust Differences-in- Differences Estimates,” Quarterly Journal of Economics, 119(1), pp.249-275.

[20] Cameron, A.C. and P.K. Trivedi (2005), Microeconometrics: Methods and Applications, Cambridge University Press.

[21] Cameron, A.C. and P.K. Trivedi (2008), Microeconometrics Using Stata, Stata Press.

[22] Cox, D.R. (1972) “Regression Models and Life Tables (with Discussion),” Journal of the Royal Statistical Society, B, 34, pp.187–220.

[23] Cox, D.R. (1975) “Partial Likelihood,” Biometrika, 62, pp.269–276.

[24] Heckman, J.J. (1979) “Sample Selection Bias as A Specification Error,” Econometrica, 47(1), pp.153- 62.

[25] Hsiao, C. (2002), Analysis of Panel Data, Cambridge University Press. , 2007

[26] Johnes, G. and Y. Tanaka (2008) “Changes in Gender Wage Discrimination in the 1990s: A Tale of Three Very Different Economies,” Japan and the World Economy, 20(1), pp.97-113.

[27] Mincer, J. (1962) “On-the-Job Training: Costs, Returns, and Some Implications,” Journal of Polit- ical Economy, 70, pp.50-79.

[28] Mincer, J. (1985) “Inter-Country Comparisons of Labor Force Trends and of Related Developments: An Overview,” Journal of Labor Economics, 3(1), pp.S1-S32.

[29] Miyoshi, K. (2008) “Male-Female Wage Differentials in Japan,” Japan and the World Economy, 20(4), pp.479-496.

[30] Neuman, S. and R.L. Oaxaca (2004) “Wage Decompositions with Selectivity-Corrected Wage Equa- tions: A Methodological Note,” Journal of Economic Inequality, 2, pp.3-10.

[31] Oaxaca, R.L. (1973) “Male-Female Wage Differentials in Urban Labor Markets,” International Eco- nomic Review, 14, pp.693-709.

[32] OECD ed. (2002) Employment Outlook, Paris: Organization for Economic Co-operation and Devel- opment.

[33] Wooldridge, J.M. (2001), Econometric Analysis of Cross Section and Panel Data, MIT Press. [34] Yamaguchi, K. (1991) Event History Analysis, Sage Publications, London.