EC423 Redo

材料扫描与范围

Found in EC423 Labour/:

Category	Materials used	Scope decision
Past Exams	`Past Exams/EC423_2020.pdf` to `Past Exams/EC423_2025.pdf`	全部纳入
AT lecture slides	`Lecture Slides/EC423_Autumn_Topic1.pdf` to `EC423_Autumn_Topic4.pdf`	纳入主表
WT lecture slides	`Lecture Slides/EC423_WT_Lec1.pdf` to `EC423_WT_Lec10.pdf`; `EC423_WT_Lec4_Update.pdf`	纳入主表；Place-Based Policies 采用 update 版
Practice and seminars	Problem sets, WT seminars, available solutions	用于辅助识别题型，不替代课件页码
Caveat	Some old crime questions	当前 AT/WT lecture PDFs 中没有完整 Becker crime model deck；这些题仍保留，并标注未找到对应课件页

高频考点速记

Topic	High-yield exam pattern	Core source
AT Topic 1: Labor Supply	static labor supply, Slutsky equation, income floor, NIT, UBI, SSP, IV/LATE	`Autumn Topic 1, p7-23`; `Autumn Topic 1, p62-87`; `Autumn Topic 1, p124-149`
AT Topic 2: Human Capital and Methods	Mincer, OLS bias, IV, DiD, RD, fuzzy RD, manipulation checks	`Autumn Topic 2, p9-52`; `Autumn Topic 2, p55-73`; `Autumn Topic 2, p74-101`
AT Topic 3: Immigration	immigration surplus, Roy/Borjas selection model, counterfactual earnings, IMR	`Autumn Topic 3, p15-28`; `Autumn Topic 3, p47-75`
AT Topic 4: Inequality and Technology	quantile regression, CES skill premium, relative supply/demand, robots/technology	`Autumn Topic 4, p29-45`; `Autumn Topic 4, p57-93`
WT Lecture 1-2	gender gaps, racial gaps, taste-based and statistical discrimination, employer learning	`Lecture 1, p26-55`; `Lecture 2, p57-96`
WT Lecture 4-5	Rosen-Roback, amenity valuation, MTO, place-based policies	`Lecture 4 Update, p21-76`; `Lecture 5, p37-67`
WT Lecture 6	minimum wage theory, bite/coverage, Cengiz, Dube, distributional effects	`Lecture 6, p16-34`; `Lecture 6, p67-104`; `Lecture 6, p108-121`
WT Lecture 8-9	intergenerational mobility, measurement, adoptees/siblings, displacement scarring	`Lecture 8, p15-27`; `Lecture 8, p37-58`; `Lecture 8, p78-123`; `Lecture 9, p3-44`
WT Lecture 10	unemployment insurance, Baily-Chetty, liquidity vs moral hazard	`Lecture 10, p4-19`; `Lecture 10, p46-59`

Past Exam Redo Table

题目	题号(a,b,c,d)	考点	主要解题方法	关联课件
2020 Q1 Migration/Roy model	a	迁移决策的线性近似	从 migrate iff exp(w1)-C > exp(w0) 出发，用小迁移成本近似得到 $(μ_{1} - μ_{0} - π) + (ε_{1} - ε_{0}) > 0$	`Autumn Topic 3, p56-63`
2020 Q1 Migration/Roy model	b	migration probability	定义 $ν = ε_{1} - ε_{0}$ ，标准化成 z，用 normal CDF 写迁移概率	`Autumn Topic 3, p60-66`
2020 Q1 Migration/Roy model	c	migrants 的 counterfactual home earnings	用 truncated normal/IMR 计算 $E [w_{0} ∣ I = 1]$ ，强调 selection term	`Autumn Topic 3, p64-68`
2020 Q1 Migration/Roy model	d	模型假设评价	逐条评估 earnings normality, known parameters, no nonpecuniary motives, proportional migration cost, no credit/friction	`Autumn Topic 3, p47-75`
2020 Q2 RD school grant	a	parametric RD specification	写 $Y_{i} = α + β (x_{i} - x_{0}) + ρ D_{i} + η_{i}$ ，并说明 running variable, cutoff, bandwidth, treatment definition	`Autumn Topic 2, p74-85`
2020 Q2 RD school grant	b	fewer parametric assumptions	用 local linear/nonparametric RD；说明 bandwidth 选择和 local nature	`Autumn Topic 2, p86-88`
2020 Q2 RD school grant	c	other policy at same threshold	检查其它政策或 predetermined outcomes/covariates 是否在 cutoff 同时跳跃	`Autumn Topic 2, p91-95`; `Autumn Topic 2, p108`
2020 Q2 RD school grant	d	headteacher changes after policy	区分 post-treatment mediator 和 manipulation；若 assignment 用 census pre-policy headteacher status，不一定 invalid，但解释变成 reduced form	`Autumn Topic 2, p91-101`
2020 Q2 RD school grant	e	manipulation concerns	检查 running variable density bunching, covariate balance, sorting around cutoff	`Autumn Topic 2, p91-95`; `Autumn Topic 2, p123-124`
2020 Q3 Crime and incentives	a	Becker crime model and legal wages	用 opportunity cost 写 legal wage 上升降低 crime participation；解释 table wage coefficient sign	未在当前 lecture PDFs 中找到 Becker crime 对应页
2020 Q3 Crime and incentives	b	conviction rate and deterrence	conviction probability raises expected cost of crime；看 coefficient 是否与 deterrence prediction 一致	未在当前 lecture PDFs 中找到 Becker crime 对应页
2020 Q3 Crime and incentives	c	conviction rate endogeneity	crime rates may affect policing/conviction, omitted enforcement shocks, reverse causality	未在当前 lecture PDFs 中找到 Becker crime 对应页
2020 Q3 Crime and incentives	d	sentence length as IV	写 relevance and exclusion；比较 OLS/IV 可解释 measurement error or endogeneity direction	未在当前 lecture PDFs 中找到 Becker crime/IV 应用页；IV 基础见 `Autumn Topic 1, p124-149`
2020 Q3 Crime and incentives	e	minimum wage as IV for low wage	评估 relevance 和 exclusion；最低工资可能通过 employment margin 直接影响 crime，需用 MW theory/evidence 平衡讨论	`Lecture 6, p21-34`; `Lecture 6, p50-104`
2020 Q4 Becker-Tomes IGM	a	OLS 不识别 structural beta	写 unobserved endowment correlated with parental income，OLS mixes parental investment and inherited endowment	`Lecture 8, p15-25`; `Lecture 8, p37-52`
2020 Q4 Becker-Tomes IGM	b	identical twins	用 twins 控制 genetic endowment；说明需要 identical endowment and comparable family treatment	`Lecture 8, p78-88`
2020 Q4 Becker-Tomes IGM	c	adoptees	用 adoptive parent income variation；需要 random assignment/no selective placement	`Lecture 8, p93-108`
2020 Q4 Becker-Tomes IGM	d	adoption estimates interpretation	分 biological parent effect 和 adoptive parent effect，解释 nature vs nurture/environment	`Lecture 8, p93-99`
2020 Q4 Becker-Tomes IGM	e	China shock and upward mobility	把 local trade shock 连到 parental income, school/neighborhood, job displacement scarring and local opportunity	`Lecture 8, p117-123`; `Lecture 4 Update, p7-13`; `Lecture 9, p55-59`
2021 Q1 Divorce and income	a	simultaneous causal signs	判断 income affects divorce and divorce affects income 的符号，并说明机制	`Autumn Topic 1, p100-123`; `Autumn Topic 4, p42-45`
2021 Q1 Divorce and income	b	OLS in simultaneous equations	代入联立方程，写 OLS estimand as structural effect plus simultaneity bias	`Autumn Topic 1, p114-123`
2021 Q1 Divorce and income	c	bias sign intuition	用 covariance/signs 判断 OLS bias 方向，解释 reverse causality	`Autumn Topic 1, p114-123`
2021 Q1 Divorce and income	d	first-born girl as IV	写 first stage, exclusion, independence；用于 divorce 对 income 的 causal effect	`Autumn Topic 1, p124-133`
2021 Q1 Divorce and income	e	heterogeneous IV model	需要说服 monotonicity, first-stage heterogeneity, no defiers, exclusion with heterogeneous effects	`Autumn Topic 1, p134-149`
2021 Q1 Divorce and income	f	LATE limitations	说明 IV identifies compliers not ATE/ATT；可用 multiple instruments, external validity discussion, heterogeneity analysis	`Autumn Topic 1, p144-152`
2021 Q2 UBI	a	UBI vs income floor and labor supply	画 budget constraints；分 initial hours and participation；UBI income effect vs income floor kink/notch	`Autumn Topic 1, p62-72`
2021 Q2 UBI	b	evidence for quantitative UBI effects	引 lottery wealth effects, NIT/transfer evidence, SSP caution；强调 external validity and financing	`Autumn Topic 1, p24-45`; `Autumn Topic 1, p67-87`
2021 Q2 UBI	c	work-conditioned transfer	比较 unconditional UBI and eligibility threshold；讨论 extensive margin, bunching at 10 hours, inequality	`Autumn Topic 1, p65-72`; `Autumn Topic 1, p88-92`
2021 Q3 Crime scars	a	unemployment and crime participation	用 legal opportunity cost/scarring logic；entry unemployment worsens legal labor market payoff	未在当前 lecture PDFs 中找到 Becker crime 对应页；scarring 相关见 `Lecture 9, p3-16`
2021 Q3 Crime scars	b	migration self-selection	local labor shocks may induce selective migration；selection changes treated population and biases local estimates	`Autumn Topic 3, p47-59`; `Lecture 4 Update, p5-13`
2021 Q3 Crime scars	c	dynamic crime effects	读 event-time pattern，区分 short-run and long-run scarring；联系 labor market scarring persistence	`Lecture 9, p3-16`; `Lecture 9, p26-33`
2021 Q3 Crime scars	d	criminal record persistence	early convictions create path dependence, lower legal opportunities, higher reoffending	未在当前 lecture PDFs 中找到 Becker crime 对应页；employment scarring 相关见 `Lecture 9, p3-16`
2021 Q3 Crime scars	e	racial heterogeneity and discrimination	预期 minorities experience larger labor-market shocks；不能直接把 heterogeneity 等同 discrimination，需要额外识别	`Lecture 2, p5-15`; `Lecture 2, p57-96`; `Lecture 9, p3-16`
2021 Q4 Cengiz minimum wage	a	missing/excess jobs	读 wage-bin figure，比较 below-MW missing jobs and above-MW excess jobs；对照 perfect competition	`Lecture 6, p21-34`; `Lecture 6, p84-91`
2021 Q4 Cengiz minimum wage	b	employment-wage elasticity	用 percent employment effect divided by percent wage effect；解释 near-zero elasticity	`Lecture 6, p21-34`; `Lecture 6, p86-91`
2021 Q4 Cengiz minimum wage	c	industry bite	restaurants/low-wage sectors have higher bite and stronger first stage	`Lecture 6, p16-19`; `Lecture 6, p84-91`
2021 Q4 Cengiz minimum wage	d	small shocks and adjustment margins	讨论价格、利润、productivity、firm value、reallocation, not only employment	`Lecture 6, p79-104`; `Lecture 6, p108-120`
2021 Q4 Cengiz minimum wage	e	MW and long-run racial gaps	区分 current earnings gap and persistent mobility; one-generation policy may not change rank-rank mobility	`Lecture 6, p108-120`; `Lecture 8, p37-42`; `Lecture 8, p117-123`; `Lecture 2, p5-7`
2022 Q1 MTO	a(i)	treatment 1 vs treatment 2	Restricted voucher vs unrestricted voucher ITT; interpret offer effect and take-up	`Lecture 5, p37-43`
2022 Q1 MTO	a(ii)	treatment 1 vs control	Offer of low-poverty voucher vs no offer; not pure effect of moving itself	`Lecture 5, p37-43`
2022 Q1 MTO	a(iii)	controls in randomized experiment	Controls improve precision/check balance; do not rescue identification	`Lecture 5, p40-43`
2022 Q1 MTO	b	babies born after program	警惕 endogenous fertility/post-treatment sample；建议预定义 cohorts/outcomes	`Lecture 5, p37-50`
2022 Q1 MTO	c	many outcomes	讨论 multiple testing, pre-analysis plan, outcome families, FWER	`Lecture 5, p45-50`; `Autumn Topic 2, p113-118`
2022 Q1 MTO	d	voucher offer as IV	写 first stage, exclusion, monotonicity, LATE for movers/compliers	`Lecture 5, p37-43`; `Autumn Topic 1, p134-149`
2022 Q1 MTO	e	demolition quasi-experiment	需要 demolition as-good-as-random, comparable controls, no selective mobility；比较 MTO ITT	`Lecture 5, p52-67`
2022 Q2 Degree RD	a	linear RD assumptions	说明 cutoff 附近 distinction recipients/nonrecipients comparable；linear CEF assumption	`Autumn Topic 2, p74-85`
2022 Q2 Degree RD	b(i)	potential outcome CEFs	写 $E [Y_{0} ∣ x]$ 和 $E [Y_{1} ∣ x]$ 的二阶多项式近似，允许 cutoff 两边形状不同	`Autumn Topic 2, p80-86`
2022 Q2 Degree RD	b(ii)	flexible estimating equation	用 centered running variable and interactions with treatment for second-order polynomial	`Autumn Topic 2, p80-86`
2022 Q2 Degree RD	b(iii)	advantage over linear RD	flexible CEF 减少非线性误判为 discontinuity 的风险	`Autumn Topic 2, p81-88`
2022 Q2 Degree RD	c	nonparametric RD	local linear RD around cutoff; identifies local treatment effect at cutoff	`Autumn Topic 2, p86-88`; `Autumn Topic 2, p96-101`
2022 Q2 Degree RD	d	RD threats	manipulation/bunching and covariate imbalance; check density and predetermined covariates	`Autumn Topic 2, p91-95`; `Autumn Topic 2, p123-124`
2022 Q3 Ban the Box	a	discrimination against ex-offenders	statistical discrimination: criminal record is productivity/risk signal；也可提 taste-based	`Lecture 2, p57-67`; `Lecture 2, p71-96`
2022 Q3 Ban the Box	b	DiD assumptions	BTB timing across MSAs; need parallel trends by race, no differential shocks/no anticipation	`Lecture 2, p93-96`; `Autumn Topic 2, p55-73`
2022 Q3 Ban the Box	c	interpreting Table 4	compare White/Black/Hispanic effects for young low-educated males	`Lecture 2, p93-96`
2022 Q3 Ban the Box	d	older/more educated placebo-style evidence	effects concentrated in groups employers use as proxy supports statistical not pure taste discrimination	`Lecture 2, p71-96`
2022 Q3 Ban the Box	e	BTB and reoffending	lower employment opportunity can increase offending/reoffending incentives; connect policy backfire	`Lecture 2, p93-96`
2022 Q4 Parental displacement	a	IGE beta	Define beta as intergenerational earnings elasticity; discuss lifetime income measurement	`Lecture 8, p37-58`
2022 Q4 Parental displacement	b	causal effect of plant closure	Need plant closure conditionally exogenous; compare displaced vs non-displaced with controls/no pre-trends	`Lecture 9, p3-8`; `Lecture 9, p30-33`; `Lecture 9, p55-59`
2022 Q4 Parental displacement	c	child outcomes	Interpret earnings, UI claims, social assistance effects as long-run spillovers	`Lecture 9, p55-59`; `Lecture 8, p37-42`
2022 Q4 Parental displacement	d	Becker-Tomes channels	parental income shock lowers investment, changes neighborhood/school/stress	`Lecture 8, p15-27`; `Lecture 9, p3-16`
2022 Q4 Parental displacement	e	persistence and scarring	persistent losses through employer/match capital and unemployment spells feed into children outcomes	`Lecture 9, p3-16`; `Lecture 9, p26-44`; `Lecture 9, p55-59`
2023 Q1 Migration/Roy model	a	migration decision approximation	Same Borjas/Roy derivation as 2020 Q1a	`Autumn Topic 3, p56-63`
2023 Q1 Migration/Roy model	b	counterfactual and realized migrant earnings	Use IMR selection terms for $E [w_{0} ∣ I = 1]$ and $E [w_{1} ∣ I = 1]$	`Autumn Topic 3, p64-68`
2023 Q1 Migration/Roy model	c	treatment effect of migration on migrants	Compare $E [w_{1} - w_{0} ∣ I = 1]$ to $μ_{1} - μ_{0}$ ; characterize selection terms and equality conditions	`Autumn Topic 3, p67-75`
2023 Q1 Migration/Roy model	d	ideal migration experiment	Randomly assign migration opportunity/cost; discuss ethics, compliance, external validity, SUTVA/general equilibrium	`Autumn Topic 3, p47-75`
2023 Q2 SSP	a	static labor supply prediction	Draw welfare vs SSP budget; use Slutsky and participation threshold; evaluate 0/15/30/45 hours	`Autumn Topic 1, p7-23`; `Autumn Topic 1, p75-87`
2023 Q2 SSP	b	control group hours over time	Read table time pattern; describe baseline and trend in control group	`Autumn Topic 1, p78-87`
2023 Q2 SSP	c	treatment group hours over time	Read treatment group trend and take-up/intensive/extensive margins	`Autumn Topic 1, p78-87`
2023 Q2 SSP	d	program effects over time	Compare adjusted/unadjusted differences; discuss take-up, job search, eligibility, dynamic response	`Autumn Topic 1, p78-87`
2023 Q2 SSP	e	static model vs observed evidence	Explain static model predicts incentives but not gradual take-up/time dynamics	`Autumn Topic 1, p75-87`
2023 Q3 MW and racial inequality	a	coverage, bite, racial pay gaps	Newly covered industries have higher Black share, creating stronger treatment intensity	`Lecture 6, p16-19`; `Lecture 6, p108-121`; `Lecture 2, p5-7`
2023 Q3 MW and racial inequality	b	DiD assumptions	Treatment newly covered industries, control already covered; need parallel trends by race	`Lecture 6, p54-58`; `Lecture 6, p86-90`; `Autumn Topic 2, p55-73`
2023 Q3 MW and racial inequality	c	wage effects by race	Positive earnings effects; larger Black effect follows exposure/bite	`Lecture 6, p115-120`; `Lecture 6, p108-121`; `Lecture 2, p5-7`
2023 Q3 MW and racial inequality	d	employment effects	Compare perfect competition job-loss prediction to small/zero modern employment effects	`Lecture 6, p21-34`; `Lecture 6, p84-104`
2023 Q3 MW and racial inequality	e	historical racial gap decline	MW compressed lower tail and high-exposure Black industries; later reforms may have less racial exposure	`Lecture 6, p108-121`; `Lecture 6, p115-120`; `Lecture 2, p5-7`; `Lecture 2, p57-67`
2023 Q4 Trade and crime	a	import competition and local labor markets	Explain tariff reductions by regional industrial exposure affect earnings/employment	`Lecture 4 Update, p9-13`; `Autumn Topic 4, p83-93`
2023 Q4 Trade and crime	b	DiD assumptions and labor-market results	Need regional exposure as-good-as-random conditional on controls; interpret earnings/employment panels	`Autumn Topic 2, p55-73`; `Lecture 4 Update, p9-13`
2023 Q4 Trade and crime	c	labor-market effects in Becker crime model	Lower legal earnings/employment reduce opportunity cost of crime; Becker crime model itself not in lecture deck	未在当前 lecture PDFs 中找到 Becker crime 对应页；labor shock context `Lecture 9, p3-16`
2023 Q4 Trade and crime	d	reduced-form crime estimates	Interpret RTC coefficient signs relative to predicted crime changes and policy exposure scale	未在当前 lecture PDFs 中找到 Becker crime 对应页；local shocks context `Lecture 4 Update, p9-13`
2023 Q4 Trade and crime	e	IV exclusion problem	Trade shock affects crime through public goods, revenue, inequality, migration, not only labor market outcomes	`Autumn Topic 1, p124-149`; `Lecture 4 Update, p9-18`
2024 Q1 Slutsky and lottery	a	wage decrease via Slutsky	Use Marshallian vs Hicksian leisure demand; translate leisure response into labor supply	`Autumn Topic 1, p16-23`
2024 Q1 Slutsky and lottery	b	unambiguous wage increase effect	State cases where substitution and income effects imply same direction, e.g. inferior leisure or dominated income effect	`Autumn Topic 1, p20-23`
2024 Q1 Slutsky and lottery	c	Imbens lottery figure	Discuss preexisting differences, timing, magnitude, volatility, and relation to income effect	`Autumn Topic 1, p24-29`
2024 Q1 Slutsky and lottery	d	Swedish lottery identification	Explain lottery-cell randomization, controls, randomization tests, difference from Imbens et al.	`Autumn Topic 1, p30-39`
2024 Q1 Slutsky and lottery	e	decompose earnings into hours and wages	Distinguish labor-supply hours response from wage/job-quality response	`Autumn Topic 1, p39-45`
2024 Q2 CES skill premium	a	derive skill premium and elasticity	Use marginal products under CES; derive $w_{H} / w_{L}$ and elasticity wrt $A_{H} / A_{L}$	`Autumn Topic 4, p57-67`
2024 Q2 CES skill premium	b	aggregate time-series estimation	Estimate log skill premium on log relative supply and demand trend; discuss identification and endogeneity	`Autumn Topic 4, p68-79`
2024 Q2 CES skill premium	c	firm-panel estimation	Use firm variation in technology/skill mix; discuss sorting, simultaneity, demand shocks	`Autumn Topic 4, p75-90`
2024 Q2 CES skill premium	d	low-skilled wage decline	Explain possible skill-biased tech with relative demand shifts; assess plausibility	`Autumn Topic 4, p65-79`; `Autumn Topic 4, p83-93`
2024 Q3 Upward mobility	a(i)	gender/race mobility differences	Read figure through gender/racial gaps and pre-market/post-market mechanisms	`Lecture 1, p7-25`; `Lecture 2, p5-15`; `Lecture 8, p117-123`
2024 Q3 Upward mobility	a(ii)	geography in mobility	Use Chetty-style geography and neighborhood opportunity	`Lecture 8, p117-123`; `Lecture 5, p37-50`
2024 Q3 Upward mobility	b	school quality in Becker model	School quality changes return/cost/productivity of parental investment	`Lecture 8, p15-27`; `Lecture 8, p110-114`
2024 Q3 Upward mobility	c(i)	border-pair model and Dube analogy	Cross-border pairs absorb local shocks while policy variation differs	`Lecture 6, p67-78`; `Lecture 8, p117-123`
2024 Q3 Upward mobility	c(ii)	why border pairs help	Within-pair differencing removes shared geography; remaining concern state-specific shocks	`Lecture 6, p67-78`; `Lecture 8, p117-123`
2024 Q3 Upward mobility	d	mandated teacher wages as IV	Discuss relevance, exclusion, first stage, 2SLS vs OLS	`Lecture 6, p21-34`; `Lecture 6, p67-78`; `Lecture 8, p110-114`
2024 Q3 Upward mobility	e	employment margin critique	Mandated wage may affect teacher employment/quality/composition, threatening exclusion	`Lecture 6, p21-34`; `Lecture 6, p79-104`
2024 Q4 Racial bias in policing	a	manipulation around threshold	Use bunching/density logic near 9/10 mph to classify lenient officers	`Autumn Topic 2, p91-95`; `Lecture 2, p57-92`
2024 Q4 Racial bias in policing	b(i)	Price-Wolfers similarity	Compare decision-maker assignment and same/different race treatment interaction	`Lecture 2, p38-52`; `Lecture 2, p57-67`
2024 Q4 Racial bias in policing	b(ii)	assignment assumption	Need quasi-random encounter with lenient officer conditional on controls	`Lecture 2, p38-52`; `Lecture 2, p57-67`
2024 Q4 Racial bias in policing	b(iii)	testing assignment assumption	Covariate balance: driver characteristics should not predict officer leniency	`Lecture 2, p38-52`; `Lecture 2, p57-67`
2024 Q4 Racial bias in policing	c	beta3 interpretation	beta3 shows whether white drivers receive more leniency discount than minority drivers	`Lecture 2, p57-92`
2024 Q4 Racial bias in policing	d(i)	taste vs statistical discrimination	Define prejudice/taste versus Bayesian group signal under imperfect information	`Lecture 2, p57-67`; `Lecture 2, p71-86`
2024 Q4 Racial bias in policing	d(ii)	prior tickets and statistical discrimination	If prior tickets proxy criminality, residual race gap is less consistent with statistical discrimination on that signal	`Lecture 2, p71-92`
2025 Q1 Occupational decline	a	data for long-run worker effects	Use linked admin/panel earnings and occupation data with pre-period occupation and controls	`Lecture 9, p3-8`; `Lecture 9, p28-33`
2025 Q1 Occupational decline	b	causal assumptions	Conditional exchangeability/parallel trends between declining and non-declining occupations	`Lecture 9, p30-33`
2025 Q1 Occupational decline	c	testing assumptions	Pre-trends, placebo outcomes, pre-period balance, robustness to control groups	`Lecture 9, p30-33`
2025 Q1 Occupational decline	d	decomposition	Interpret treatment effect for A, treatment effect for B, and baseline occupation differences	`Lecture 9, p28-44`
2025 Q1 Occupational decline	e	regression interpretation	State assumptions for beta to summarize causal occupational-decline loss	`Lecture 9, p28-44`
2025 Q2 UBI	a	UBI vs income floor	Draw budget constraints and compare participation/hours by initial-hours scenario	`Autumn Topic 1, p62-72`
2025 Q2 UBI	b	quantitative UBI evidence	Use lottery/NIT/SSP evidence with caution about external validity and financing	`Autumn Topic 1, p24-45`; `Autumn Topic 1, p67-87`
2025 Q2 UBI	c	work-conditioned UBI	Analyze 10-hour eligibility threshold, participation incentives, bunching, inequality	`Autumn Topic 1, p65-72`; `Autumn Topic 1, p88-92`
2025 Q3 Rosen-Roback	a	worker problem and FOCs	Set utility over composite good, land, amenity; FOC equates MRS to rent/price	`Lecture 4 Update, p23-26`; `Lecture 4 Update, p30-32`
2025 Q3 Rosen-Roback	b	indirect utility and worker equilibrium	Derive $V (w, r, s)$ ; mobile workers require equalized utility	`Lecture 4 Update, p43-45`; `Lecture 4 Update, p59-63`
2025 Q3 Rosen-Roback	c	firm unit cost and equilibrium	Derive $c (w, r, s)$ and zero-profit/free-entry condition $c = 1$	`Lecture 4 Update, p27-29`; `Lecture 4 Update, p64-75`
2025 Q3 Rosen-Roback	d(i)	regulation cost shock	Treat zero-emission machinery as negative productivity/cost shock; sign wages/rents through equilibrium conditions	`Lecture 4 Update, p47-58`; `Lecture 4 Update, p64-75`
2025 Q3 Rosen-Roback	d(ii)	firm lump-sum transfer	Distinguish lump-sum transfer from marginal productivity/amenity change; discuss capitalization	`Lecture 4 Update, p47-58`; `Lecture 4 Update, p64-75`
2025 Q3 Rosen-Roback	e	empirical amenity valuation	Estimate capitalization into wages/rents; specify data, identifying variation, welfare formula	`Lecture 4 Update, p59-75`
2025 Q3 Rosen-Roback	f	policy recommendation	Recommend policies only if amenity/productivity gains exceed costs and target groups retain welfare	`Lecture 4 Update, p16-18`; `Lecture 4 Update, p76-77`; `Lecture 5, p9-22`
2025 Q4 Statistical discrimination	a	statistical discrimination model	Write productivity prior by group, noisy signal, Bayesian updating	`Lecture 2, p71-86`
2025 Q4 Statistical discrimination	b	employer learning test	Use experience interactions; group/schooling proxy weights should change as employers learn	`Lecture 2, p90-92`
2025 Q4 Statistical discrimination	c	cautions	Omitted ability, schooling as proxy, selection, experience profile differences, what firms observe	`Lecture 2, p21-36`; `Lecture 2, p90-92`
2025 Q4 Statistical discrimination	d	ability proxy z	Estimate Altonji-Pierret style model with ability proxy and experience interactions	`Lecture 2, p90-92`
2025 Q4 Statistical discrimination	e	policy recommendation	Improve individual information signals/certification/auditing; warn removing information can backfire	`Lecture 2, p93-101`

空复盘表

题目	我当时卡在哪里	下次重做计划
2020 Q1 Migration/Roy model
2020 Q2 RD school grant
2020 Q3 Crime and incentives
2020 Q4 Becker-Tomes IGM
2021 Q1 Divorce and income
2021 Q2 UBI
2021 Q3 Crime scars
2021 Q4 Cengiz minimum wage
2022 Q1 MTO	这道题好就好在它几乎覆盖了所有本节课教的 identification Strategy: RCT 下的 OLS, IV/2SLS → LATE, 以及简单带过的 DiD。需要知道怎么去区分 ATE 和 LATE。
2022 Q2 Degree RD	考察了 RDD 的相关内容，需要回顾 `Topic 2`，而且不算难，只需要回顾课件里的内容就行，不需要有天马行空的设定。
2022 Q3 Ban the Box	经典 Statistical Discrimination 话题，
2022 Q4 Parental displacement	e 问的突破点反而是在 Lecture 9，而且提到了一个 Labour Market Scarring 这个概念完全没有想到。	重新推导一次
2023 Q1 Migration/Roy model
2023 Q2 SSP
2023 Q3 MW and racial inequality
2023 Q4 Trade and crime
2024 Q1 Slutsky and lottery
2024 Q2 CES skill premium
2024 Q3 Upward mobility
2024 Q4 Racial bias in policing
2025 Q1 Occupational decline
2025 Q2 UBI
2025 Q3 Rosen-Roback
2025 Q4 Statistical discrimination

感悟

什么情况下 ITT = TOT？
- 当我们强行要求有 compliance 的时候（perfect compliance）
- ITT 和 TOT 的数量关系：TOT = ITT/Take up rate
ATE 和 LATE 的区别？
- 同上，如果 compliance
Control Variable 的原则
- 只加 pre-treatment variable
- post-treatment variable? → Bad Control
Internal Validity 和 External Validity 取舍问题
- RCT → 强internal validity，弱external validity（self-selected sample）
- Quasi-experiment → internal validity依赖parallel trends，但population更接近general population
最新奇：Multiple Hypothesis test 带来的问题
- 存在可能性说 you find some significance by chance
  - 解决办法：Anderson summary index或Bonferroni correction. 后者比较简单，只需要拿传统的 significance level 去除以 outcome 的个数即可。
Sample Selection bias
- 当分析样本的 存在本身 被treatment影响时，randomization在子样本上失效
IV 的 Exclusion Restriction 问题
- IV要求instrument只通过endogenous variable影响outcome
- 不可检验（untestable），只能通过经济直觉论证
- Excluded category的选择影响exclusion restriction的可信度（T2 vs. Control的区别）
如何理解 Excluded Category?
- 即regression中 被省略的那个baseline组，也叫 reference group。在 sample 里，是都有的
Regression Discontinuity Design
- 主要需要掌握三种 RDD（Sharp）即 Linear, Flexible Polynomial, Nonpar，分别是递进的关系。需要说出，每种设定放松了哪个假设，代价是什么。
  - 这里有一个隐藏的点，就是我们在 Linear RD 中，其实隐含了 treatment effect 是 constant 的点。
- Threats to Identification
  - Manipulation：操控/谎报直接导致两侧 not comparable
    - Solution：McCrary density test + pre-determined covariates在cutoff处无jump
  - Confounding discontinuities: 其他treatment也在同一cutoff变化 → bundled effect
    - Solution: 检查其他 outcome 是否也一同变化 + 研究制度规则，人为扫清障碍。
- Nonparametric Method
  - Nonparametric RD的核心tension：用更窄的窗口更准确但更 noisy，用更宽的窗口更精确但可能引入 bias。
    - Bandwidth $δ$ 小 → 低bias，高variance
    - Bandwidth $δ$ 大 → 低variance，高bias
  - 解决方案：local linear regression
Potential Outcome
- 一个合并 CEF 的技巧，可以见 2022 Q2，交互项的含义指允许两侧的斜率和曲率不同
  - 为什么要 renormalize？因为让截距差直接代表我们想捕捉的效应，也就是在 cutoff 处的 treatment effect。
LATE
- 这里主要是需要和 IV 的 LATE 辨析一下，IV 的 LATE 是 local 在 complier，但是在 RDD 中，主要是 local 在这个 cutoff 附近。
F test
- F test 的作用就是去检测一组参数是否同时为 0.如果 F＞critical value，那么我们就要拒绝 $H_{0}$ 意思就是额外的参数是有意义的。反之，则建议简单的模型就够用。

August's Digital Garden

EC423 Redo

EC423 Redo

材料扫描与范围

高频考点速记

Past Exam Redo Table

空复盘表

感悟

Graph View

Table of Contents