LANGUAGE TRENDS IN PUBLIC ECONOMICS

Transcription

1 LANGUAGE TRENDS IN PUBLIC ECONOMICS Henrik Jacobsen Kleven Princeton University July 2018

2 What Are We Talking About In Public Economics? Understanding new directions in research based on word and language trends Textual analysis of public economics papers since 1975 NBER working papers Focus on PE-tagged papers (4676 papers) Analyze full texts Caveat: NBER is a selected sample of the profession, and the nature of the selection has changed over time 2 / 42

3 Broad Themes 3 / 42

4 We Talk Less About Taxes Than We Used To Do 100% 90% 80% Regulation, Public Goods, etc Political Economy Fraction of Words 70% 60% 50% 40% 30% 20% Taxes Transfers Education 10% 0% Poterba PE Director Note: The graph shows the frequency of words within different topics as a fraction of all words across topics. See here for a list of words included in each category. The graph shows 3-year moving averages. Year 4 / 42

5 When We Do Talk About Taxes, Which Taxes Do We Talk About? Fraction of Words 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Capital Tax Labor & Income Tax Other Taxes Property Tax Consumption Tax Note: The graph shows the frequency of different words as a fraction of all words in the tax topic. See here for a list of words included in each category. The graph shows 5-year moving averages. Year 5 / 42

6 We Talk Mostly About The US, But Less So Over Time 100% 90% 80% United Kingdom Rest of EU Other Fraction of Words 70% 60% 50% 40% 30% 20% 10% 0% United States Year Note: The graph shows the frequency of each country as a fraction of all countries (among papers that mention any country). See here for details. The graph shows 5-year moving averages. 6 / 42

7 When We Don t Talk About The US, Who Do We Talk About? 100% Fraction of Words 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% United Kingdom France Germany Other India China Scandinavia Year Note: The graph shows the frequency of each country as a fraction of all countries excluding the US (among papers that mention any country). See here for details. The graph shows 5-year moving averages. 7 / 42

8 Methods 8 / 42

9 The Identification Revolution 50% 45% 40% Fraction of Papers 35% 30% 25% 20% 15% 10% 5% 0% Year Note: The graph shows the fraction of papers that mention the word identification in the context of empirical identification. See here for details. The graph shows a 5-year moving average. The Graphical Revolution 9 / 42

10 The Rise of Experiments 25% 20% Fraction of Papers 15% 10% 5% 0% Year Natural & Quasi Experiments Lab Experiments RCTs Note: The graph shows the fraction of papers that refer to each type of experiment. See here for a list of terms. The graph shows 5-year moving averages. Natural vs Quasi-Experiments 10 / 42

11 The Rise Of Quasi-Experiments 25% 20% Fraction of Papers 15% 10% 5% 0% Year Diff in Diff Regression Discontinuity Bunching Event Study Note: The graph shows the fraction of papers that refer to each type of quasi-experiment. See here for a list of terms. The graph shows 3-year moving averages. 11 / 42

12 The Rise Of Administrative Data 35% 30% Fraction of Papers 25% 20% 15% 10% 5% 0% Year Note: The graph shows the fraction of papers that mention the term administrative data or any of its synonyms. See here for a list of synonyms. The graph shows a 3-year moving average. 12 / 42

13 Big Data & Machine Learning 1.4% 1.2% Fraction of Papers 1.0% 0.8% 0.6% 0.4% 0.2% 0.0% Big Data Year Machine Learning Note: The graph shows the fraction of papers that mention the given term. See here for details. The graph shows 5-year moving averages. 13 / 42

14 The Rise Of Behavioral Economics 45% 40% 35% Fraction of Papers 30% 25% 20% 15% 10% 5% 0% Note: The graph shows the fraction of papers that mention any word unambiguously related to the topic of behavioral economics. See here for a list of words The graph shows a 5-year moving average. Year 14 / 42

15 100% When We Talk About Behavioral Economics, What Do We Talk About? Fraction of Words 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Other Social Preferences Reference Dep Nudges Present Bias Optimization Frictions Year Note: The graph shows the frequency of words within different topics as a fraction of all words across topics (among papers that mention any word unambigously related to the topic of behavioral economics). See here for a list of words included in each category. The graph shows 5-year moving averages. 15 / 42

16 Specific Terms 16 / 42

17 The Elasticity of Taxable Income 4.5% 4.0% 3.5% Fraction of Papers 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% Year Note: The graph shows the fraction of papers that mention any variation of the term elasticity of taxable income. See here for a list of terms. The graph shows a 5-year moving average. 17 / 42

18 Sufficient Statistics 8% 7% 6% Fraction of Papers 5% 4% 3% 2% 1% 0% Year Note: The graph shows the fraction of papers that use the term sufficient statistics in the context of welfare analysis. See here for details. The graph shows a 5-year moving average. 18 / 42

19 Top 1% 8% 7% 6% Fraction of Papers 5% 4% 3% 2% 1% 0% Note: The graph shows the fraction of papers that mention any variation of the term top 1%. See here for a list of terms. The graph shows a 5-year moving average. Year 19 / 42

20 Quasi-Linear 6% 5% Fraction of Papers 4% 3% 2% 1% 0% Year Note: The graph shows the fraction of papers that mention the term quasi-linear. See here for details. The graph shows a 5-year moving average. 20 / 42

21 General Equilibrium 45% 40% 35% Fraction of Papers 30% 25% 20% 15% 10% 5% 0% Note: The graph shows the fraction of papers that mention the term general equilibrium. See here for details. The graph shows a 5-year moving average. Year 21 / 42

22 What Does It All Mean? The language of (public) economics has changed enormously over the last four decades The striking changes in language reflect changes in how we conduct research Does it also reflect real progress in learning? The possibilities for this type of language analysis are endless: Battle of the Sexes Financial Crisis 22 / 42

23 Appendix 23 / 42

24 The Graphical Revolution Average Word Count Note: The graph shows the average number of mentions of the term graph or any unambiguously related word. See here for a list of related words. The graph shows a 5-year moving average. Year Back 24 / 42

25 Natural vs Quasi-Experiments 100% 90% Fraction of Words 80% 70% 60% 50% 40% 30% Natural Experiments 20% 10% Quasi Experiments 0% Year Note: The graph shows the relative frequency of the terms quasi-experiment and natural experiment among papers that mention either term. See here for details. The graph shows a 5-year moving average. Back 25 / 42

26 The Battle Of The Sexes: Pronouns 100% 90% 80% She Fraction of Words 70% 60% 50% 40% 30% 20% 10% 0% He Year Note: The graph shows the relative frequency of male and female personal pronouns among papers that use a personal pronoun at least once. See here for details. The graph shows a 3-year moving average. Back 26 / 42

27 Financial Crisis & Recession 25% Black Monday 2008 Financial Crisis 20% Fraction of Papers 15% 10% 5% 0% Financial Crisis Year Recession Note: The graph shows the fraction of papers that mention the given term. See here for details. The graph shows 5-year moving averages. Back 27 / 42

28 Category Taxes Education Transfers Welfare Programs Social Insurance Political Economy Regulation Public Goods Words tax, taxes, taxation education, school(s), schooling Welfare Programs + Social Insurance (see below for word lists in each category) Aid to Families with Dependent Children, AFDC, Children s Health Insurance Program, CHIP, SCHIP, Food Stamp(s), Medicaid, public housing, Supplemental Nutrition Assistance Program, SNAP, Supplemental Security Income, SSI, Temporary Assistance for Needy Families, TANF, transfer(s), welfare benefit(s), welfare policy(ies), welfare program(s), welfare programme(s) disability insurance, health insurance, medicare, social insurance, unemployment insurance autocracy, democracy, dictatorship, election(s), electoral, legislator(s), legislature(s), parliament(s), parliamentary, politician(s), political, politics, president(s), presidential, referendum, referenda, vote(s), voter(s), voting regulation(s), regulatory public good(s) Back 28 / 42

29 Category Labor & Income Income Tax Labor Income Tax Payroll Tax Words Income Tax + Labor Income Tax + Payroll Tax (see below for word lists in each category) income tax* labor income tax*, labour income tax*, labor tax*, labour tax*, personal income tax* payroll tax*, social security contribution*, social security tax* Capital Capital Tax + Dividend Tax + Capital Gains Tax + Corporate Income Tax + Wealth Tax & Estate Tax (see below for word lists in each category) Capital Tax Dividend Tax Capital Gains Tax Corporate Income Tax Wealth Tax & Estate Tax Consumption Tax Property Tax Other Tax capital income tax*, capital tax* dividend income tax*, dividend tax* capital gains tax*, capital-gains tax* corporate income tax*, corporate tax*, corporation tax*, business income tax*, business tax* estate tax*, inheritance tax*, wealth tax* commodity tax*, consumption tax*, excise tax*, excises*, expenditure tax*, indirect tax*, sales tax*, value added tax*, value-added tax*, VAT* housing tax*, land tax*, property tax* All other mentions of the word tax The symbol * matches any character string following the search string. Back 29 / 42

30 Category United States United Kingdom Rest of EU Other Words United States, US, U.S., USA, U.S.A., American, North America United Kingdom, Britain, England, UK, U.K., GB, G.B., British, Northern Ireland, Northern Irish Scotland, Scottish, Wales, Welsh Austria, Austrian, Belgium, Belgian, Denmark, Danish, Finland, Finnish, France, French, Germany, German, Greece, Greek, Ireland, Irish, Italy, Italian, Luxembourg, Luxembourgish, Netherlands, Dutch, Portugal, Portuguese, Spain, Spanish, Sweden, Swedish All remaining countries and respective adjectivals. Back 30 / 42

31 Category United Kingdom Germany France Scandinavia China India Other Words United Kingdom, Britain, England, UK, U.K., GB, G.B., British, Northern Ireland, Northern Irish Scotland, Scottish, Wales, Welsh Germany, German France, French Denmark, Danish, Finland, Finnish, Iceland, Icelandic, Norway, Norwegian, Sweden, Swedish China, Chinese India, Indian All remaining countries and respective adjectivals. Back 31 / 42

32 The Identification Revolution (Explanatory Note) To be included in the paper count, a paper must either mention any of these phrases or contain one or more matching sentences A matching sentence must contain the following in the given order: 1. A variation of the word identification 2. Followed by 0-2 wildcard words 3. Followed by a variation of any of these words: effect, response, impact, elasticity, parameter, coefficient Match: We use an IV strategy to identify the desired elasticity. Not a match: Voter ID laws help election officials identify voters in different districts. Back 32 / 42

33 Phrases causal identification, causally identified, condition for identification, condition for identifying, conditions for identification, conditions for identifying, empirical identification, identification analysis, identification approach, identification argument, identification assumption, identification condition, identification framework, identification issue, identification problem, identification properties, identification result, identification scheme, identification strategy, identification test, identifying assumption, identifying condition, identifying variation, issue identifying, issue with identification, issues identifying, issues with identification, non parametric identification, non parametrically identified, nonparametric identification, non-parametric identification, nonparametrically identified, non-parametrically identified, over identified, over identifying, over-identified, over-identifying, partial identification, partially identified, point identification, point identified, point identifying, point-identification, point-identified, point-identifying, problem identifying, problem with identification, problems identifying, problems with identification, set identification, set identified, set identifying, set-identification, set-identified, set-identifying, threat for identification, threat to identification, threats for identification, threats to identification, under identified, under identifying, under-identified, under-identifying, weak identification Back 33 / 42

34 Category Natural & Quasi-Experiments Lab Experiments RCTs Words natural experiment*, natural-experiment*, quasi experiment*, quasi-experiment*, quasiexperiment* laboratory experiment*, lab experiment* random* (control* field*) (trial* experiment*) The symbol * matches any character string following the search string. The pattern (word1 word2) matches either word1 or word2. Back 34 / 42

35 Category Diff-in-Diff Regression Discontinuity Bunching Event Study Words difference-in-difference*, differences-in-difference*, difference in difference*, differences in difference*, diff-in-diff*, d-in-d*, DiD regression discontinuit*, regression kink*, RKD bunching event study(ies) The symbol * matches any character string following the search string. Back 35 / 42

36 Back Category Administrative Data Big Data Machine Learning Words admin data*, admin record*, admin register*, admin registries*, admin registry*, administrative data*, administrative record*, administrative register*, administrative registries*, administrative registry* big data, big-data machine learning The symbol * matches any character string following the search string. 36 / 42

37 Category Behavioral Economics Social Preferences Optimization Frictions Present Bias Words Social Preferences + Optimization Frictions + Present Bias + Nudges + Reference Dependence + Other (see below for word lists in each category) altruism*, inequality aversion*, intrinsic motivation*, intrinsically motivated*, joy of giving*, joy-of-giving*, reciprocity*, social fairness*, social incentive*, social motivation*, social norm*, social preference*, social pressure*, social recognition*, warm glow*, warm-glow* attention cost*, inattention*, inattentive*, information friction*, limited attention*, misperception*, mis-perception*, optimisation error*, optimisation friction*, optimization error*, optimization friction*, salience* hyperbolic discounting*, present bias*, present-bias*, quasihyperbolic discounting*, quasi-hyperbolic discounting*, self control*, self-control* The symbol * matches any character string following the search string. Back Continued / 42

38 Category Nudges Reference Dependence Other Biased Beliefs Words automatic enrollment*, automatic enrolment*, default bias*, default effect*, nudge*, nudging*, optimal default*, status quo bias* endowment effect*, focal point*, loss aversion*, prospect theory*, reference dependence*, reference-dependent* Biased Beliefs + [ behavioral bias*, behavioral economic*, behavioral model*, behavioural bias*, behavioural economic*, behavioural model*, bounded rationality*, boundedly rational*, bounded-rationality*, denial of death*, denial-of-death*, inconsistent choice*, internalit*, irrational*, paternalism*, psychology* ] biased beliefs*, incorrect beliefs*, over confidence*, over-confidence*, projection bias* The symbol * matches any character string following the search string. Back Previous / 42

39 Category Elasticity of Taxable Income Top 1% Quasi-Linear General Equilibrium Words elasticities of taxable income*, elasticity of taxable income*, responsiveness of taxable income*, taxable income elasticit*, taxable income respons* top 1%*, top 1 percent*, top 1p* quasi linear*, quasi-linear*, quasilinear* general equilibrium*, general-equilibrium* The symbol * matches any character string following the search string. ETI Top 1% Quasi-Linear General Equilibrium 39 / 42

40 Sufficient Statistics (Explanatory Note) To be included in the paper count, a paper must contain one or more matching sentences A matching sentence must contain at least one word from each of the following sets of words: Back 1. [ parameter, elasticity, response, responsiveness, statistic, parameters, elasticities, responses, statistics ] 2. [ sufficient, sufficiency ] 3. [ efficiency, welfare, deadweight loss, excess burden, optimal ] Match: We show that this elasticity is a sufficient statistic for determining the optimal level of unemployment benefits. Not a match: Previous authors have challenged the assertion that Tobin s Q is a sufficient statistic for determining the investment decisions of firms. 40 / 42

41 Category Graphical Natural Experiments Quasi-Experiments Financial Crisis Recession Words graph* natural experiment*, natural-experiment* quasi experiment*, quasi-experiment*, quasiexperiment* financial cris* recession* The symbol * matches any character string following the search string. Graphical Natural vs Quasi-Experiments Financial Crisis & Recession 41 / 42

42 Category Male Personal Pronouns Female Personal Pronouns Words he, him, his she, her, hers Back 42 / 42