* 复现任务：Angrist 与 Krueger (1991)，Table IV 第1列 OLS 和第2列 2SLS。
* 输入：从作者 MIT 存档自行取得 NEW7080.dta，放在本练习目录 data/ 下。
* 输出：replication.log、estimates.csv；第一阶段作为补充诊断写入日志。
* 样本：1970年普查，1920–1929年出生男性；保留出生年固定效应。
* 原列使用常规标准误，不擅改稳健或聚类口径；未覆盖其他列/整篇论文。
* 在新练习会话和空目录运行：clear 会清内存，replace 会覆写同名结果。
version 17
set more off
set linesize 255
capture log close
log using "replication.log", text replace
use "data/NEW7080.dta", clear
rename v4 education
rename v9 log_weekly_wage
rename v16 census
rename v18 quarter_birth
rename v27 year_birth
keep if inrange(year_birth, 1920, 1929)
assert census == 70
assert _N == 247199
assert inrange(quarter_birth, 1, 4)
assert !missing(education, log_weekly_wage, year_birth)

file open results using "estimates.csv", write text replace
file write results "model,n,coefficient,standard_error,ci_lower,ci_upper" _n
regress log_weekly_wage education i.year_birth
matrix OLS = r(table)
file write results "ols," %12.0f (e(N)) "," %21.15g (_b[education]) "," %21.15g (_se[education]) "," %21.15g (OLS[5,1]) "," %21.15g (OLS[6,1]) _n
ivregress 2sls log_weekly_wage i.year_birth (education = i.quarter_birth i.quarter_birth#i.year_birth)
matrix IV = r(table)
file write results "2sls," %12.0f (e(N)) "," %21.15g (_b[education]) "," %21.15g (_se[education]) "," %21.15g (IV[5,1]) "," %21.15g (IV[6,1]) _n
estat firststage
file close results
display "AK91_TABLE_IV_TWO_COLUMNS_COMPLETED"
log close