Lecture Outline: 1. Mo1va1on: Why study economic policy uncertainty and its measurement? 2. Measuring Economic Policy Uncertainty (EPU)

Transcription

1 Professor Steven Davis, November 2014 Lecture Outline: 1. Mo1va1on: Why study economic policy uncertainty and its measurement? 2. Measuring Economic Policy Uncertainty (EPU) a. U.S. EPU index b. EPU indexes for other countries 3. Measurement Concerns and Index Evalua1on and Refinement Methods 1

2 Lecture Outline (continued): 4. A Different Text Source: FOMC Beige Books 5. Evidence on the Sources of EPU a. From newspapers b. From Beige Books 6. A Longer Term Perspective on EPU 7. Some Evidence on the Effects of EPU 8. Another Approach: Policy news and big stock market moves 2

3 Lecture Outline (continued): For additional information on our Economic Policy Uncertainty Indexes: Guide for Human Audit of Newspaper-Based Economic Policy Uncertainty at Coding_Guide.pdf In addition to these lecture slides, review the Guide to get a sense of how to take a systematic approach to developing, evaluating and refining a text-based index. 3

4 1. Motivation: Why policy uncertainty and its measurement? 4

5 Wall Street Journal OPINION January 4, 2010 Uncertainty and the Slow Recovery A recession is a terrible time to make major changes in the economic rules of the game. Article Comments (98) By GARY S. BECKER, STEVEN J. DAVIS AND KEVIN M. MURPHY We argued that policy was high and harmful to the recovery. Largely impressionistic and indirect evidence How to make progress??? à Research project with Baker and Bloom 5

6 Our Ini1al Goal: Assess the Policy Uncertainty View Policy Uncertainty View 1. Policy- related economic uncertainty is at historically high levels in recent years in the United States and in Europe. 2. High levels of policy uncertainty cause businesses and households to cutback or defer spending, investment and hiring undermining macroeconomic performance and slowing recovery from recent financial crises. 6

7 Broadening Our Goals 1. Construct new measures of economic policy uncertainty (EPU) for many countries. Thus far: Monthly indexes for USA, Canada, UK, Germany, France, Italy, Spain, China, India, Japan, Russia. Daily measure for USA 2. Develop methods to evaluate and refine our EPU indexes, especially the news- based component 3. Assess the effects of policy uncertainty on firm- level and macroeconomic performance 4. Study economic, poli1cal and social forces that underlie policy uncertainty 7

8 How Could Policy Uncertainty Hold Back the Economy? Poten1al Mechanisms (Not an Exhaus1ve List) 1. More precau1onary savings and deleveraging by households 2. When investment and hiring decisions are costly to reverse, greater uncertainty depresses and delays investment and hiring (Bernanke, 1983) 3. More costly debt and equity finance (Gilchrist et al., 2014, and Pastor and Veronesi, 2012, 2013) 4. Higher markups, intensifying monopoly distor1ons (Fernandez- Villaverde et al., 2013) 5. Managerial risk aversion (Panousi and Panikolaou, 2012) 6. Intensifica1on of agency problems, reducing the value of new and exis1ng employment, business and financial rela1onships (Narita, 2011)

9 2. Measuring Economic Policy Uncertainty (EPU) 9

10 Measuring Economic Policy Uncertainty Scott R. Baker (Stanford) Nick Bloom (Stanford, CEP & NBER) Steven J. Davis (Chicago Booth, AEI & NBER) Working paper, latest draft at

11 What Do We Want our Measures to Capture? All of the following: Uncertainty about who will make economic policy decisions e.g., who will win the next elections? Uncertainty about what economic policy actions decision makers will undertake, and when. Uncertainty about the economic effects of policy actions past, present and future actions Economic uncertainty induced by policy inaction Economic uncertainty related to national security concerns and other policy matters that are not mainly economic in character 11

12 Construc1ng Our Newspaper- Based Index of EPU for the U.S. Search digital archives of 10 major newspapers for ar1cles with terms related to EPU For each newspaper: Get monthly EPU ar1cle counts Scale by count of all ar1cles in same month Normalize scaled count so that SD=1 for à Newspaper- level EPU Index Sum across the newspaper- level indexes by month to get the U.S. news- based EPU index 12

13 Construc1ng Our Newspaper- Based EPU Index for the U.S. Text String Search Criteria: EU: {economic OR economy} AND {uncertain OR uncertainty} EPU: AND {regulation OR deficit OR federal reserve OR congress OR legislation OR white house } 13

14 Construc1ng Our Newspaper- Based EPU Index for the U.S. Text String Search Criteria: EU: {economic OR economy} AND {uncertain OR uncertainty} EPU: AND {regulation OR deficit OR federal reserve OR congress OR legislation OR white house } Later: How did we select the policy terms for our EPU search filter? 14

15 Construc1ng Our Newspaper- Based EPU Index for the U.S. Newspapers: Boston Globe Chicago Tribune Dallas Morning News Los Angeles Times Miami Herald New York Times SF Chronicle USA Today Wall Street Journal Washington Post Note: We use Access World News Newsbank Service when constructing a daily EPU Index, because the daily index requires a higher density of news sources. 15

16 Newspaper-based index of U.S. economic policy uncertainty: Jan 1985-Aug Black Monday Gulf War I Clinton- Election Bush Election Russian Crisis/LTCM 9/11 Gulf War II Lehman and TARP Stimulus Debate Debt Ceiling; Euro Debt Euro Crisis and 2010 Midterms Fiscal Cliff Shutdown Source: Measuring Economic Policy Uncertainty by Scott R. Baker, Nicholas Bloom and Steven J. Davis, all data at Data normalized to 100 prior to 2010.

17 Rise in mainly reflects concerns about tax policy and health policy More on the policy areas that underlie high policy uncertainty in recent years later in the lecture. Note: Analysis uses Newsbank coverage of around 1000 US national and local newspapers See Table 1 in the Baker, Bloom and Davis (2013) for a more detailed analysis.

18 U.S. EPU Index Back to 1900 Based on 6 Papers January 1900 to December 2012 Policy Uncertainty Index Gold Standard Act Great Depression, New Deal and FDR Assassination of McKinley Versailles conference Start of WW I McNary Haughen farm bill Great Depression Relapse Truman- Dewey election Black Monday OPEC II OPEC I Watergate Berlin Conference Gulf War I Lehman and TARP 9/11 and Gulf War II Asian Fin. Crisis Debt Ceiling Notes: Source: Measuring Economic Policy Uncertainty by Scott R. Baker, Nicholas Bloom and Steven J. Davis at Counts of articles about economic policy uncertainty in six major newspapers (WP, BG, LAT, WSJ, CHT, NY Times) scaled by number of articles about economics and business.

19 European Economic Policy Uncertainty Index Policy Uncertainty Index Asian Crisis 9/11 Nice Treaty Referendum Treaty of Accession/ 2 nd Gulf War European Cons1tu1on Rejec1on/Summit Northern Rock Takeover Lehman Bros. Italy Ra1ng Cut Greek Bailout Request, Ra1ng Cuts Papandreou call for referendum Source: Based on 10 papers (El Pais, El Mundo, Corriere della Sera, La Repubblica, Le Monde, Le Figaro, the Financial Times, Times, Handelsblatt, FAZ.)

20 India Economic Policy Uncertainty Index India Based Policy Uncertainty Index Congress Party wins Na1onal Elec1on India- US Nuclear Deal Bear Sterns Lehman Bros Price Hikes Exchange Rate Fluctua1ons and Worry Lokpal Bill Source: Data from 7 Indian newspapers (Economic Times, Times of India, Hindustan Times, Hindu, Statesman, Indian Express, and Financial Express)

21 Source: Data until August Based on newspaper articles from the South China Morning Post. China Economic Policy Uncertainty Index China Based Policy Uncertainty Index /11 China Defla1on and Deficit Rising Interest Rates Eurozone Fears and Protec1onism Infla1on and Export Pressure China S1mulus Poli1cal Transi1on and new Na1onal Congress

22 Russian Economic Policy Uncertainty Index (beta) Ac1ng PM Gaidar resigns Cons1tu1onal Crisis Kizlyar hostage crisis; PM Chubais resigns First Chechen War Russian military exits Chechnya Russian financial crisis Second Chechen War Pu1n becomes PM Orange Revolu1on in Ukraine Timoshenko resigns; Terror aoack in Nalchik Parliament dismissed In Ukraine Medveded elec1on Lehman Brothers Failure Duma elec1ons and protests against elec1on fraud Terror aoacks in Nalchik & Stavropol Pu1n elec1on Kiev Euromaidan; Crimea annexa1on Taper Tantrum Ukraine Conflict Source: Data from Kommersant daily newspaper ( ) 22

23 250" 200" 150" 100" LDP"wins"fewer"" than""half"of"" House"seats;"" G7"Summit"" in"tokyo" New"Prime" Minister" Kaifu;" Structural" reform" inicacve"" " Gulf"" War" Japan%Economic%Policy%Uncertainty%Index,%June%1988%to%September%2014% LDP"regains" Prime" Ministership" Electoral" Reform" Collapse"of"Yamaichi"SecuriCes" Government"" debt"and"" tax"policy" concerns"" New"Prime"" Minister"" Obuchi" " House"" Dissolves"" &"New" General" ElecCon" " Prime"Minister"" Mori"announces" intencon"to" step""down" " Invasion" of"iraq" House"Dissolves" &"General" ElecCon" " Prime" Minister" Fukuda" Resigns;" Lehman" Brothers" Failure" New"Prime" Minister"Noda" BOJ"Introduces" Comprehensive" Monetary"Easing" " House"" Dissolves"" &"New" General" ElecCon" " 50" 0" Source: Scoo R. Baker, Nicholas Bloom and Steven J. Davis, Measuring Economic Policy Uncertainty, Jun)88" Dec)88" Jun)89" Dec)89" Jun)90" Dec)90" Jun)91" Dec)91" Jun)92" Dec)92" Jun)93" Dec)93" Jun)94" Dec)94" Jun)95" Dec)95" Jun)96" Dec)96" Jun)97" Dec)97" Jun)98" Dec)98" Jun)99" Dec)99" Jun)00" Dec)00" Jun)01" Dec)01" Jun)02" Dec)02" Jun)03" Dec)03" Jun)04" Dec)04" Jun)05" Dec)05" Jun)06" Dec)06" Jun)07" Dec)07" Jun)08" Dec)08" Jun)09" Dec)09" Jun)10" Dec)10" Jun)11" Dec)11" Jun)12" Dec)12" Jun)13" Dec)13" Jun)14"

24 3. Measurement Concerns and Index Evalua1on Suitability Concerns A Proof of Concept: Newspaper- based VIX analog Poli1cal Slant Accuracy Concerns Designing and Fielding an Audit Study Using the Audit to Evaluate and Refine the Index Assessing Error Proper1es in the Time Series Market Evidence Another text source: Next sec1on of lecture 25

25 Two Types of Measurement Concerns Suitability: Whether an accurate count for news ar1cles about a par1cular type of uncertainty provides a good indicator for that type of uncertainty. Accuracy: Whether specific text- string search criteria accurately iden1fy the set of ar1cles that discuss a certain type of uncertainty, e.g., policy- related economic uncertainty. 26

26 Assessing Suitability Concern Idea: Apply news- based approach to a concept of uncertainty for which we have external, market- based evidence. Implementa1on: Compare VIX measure of uncertainty about future equity returns to a news- based index of equity market uncertainty, with search terms as follows: {economic OR economy} AND {uncertain OR uncertainty} AND { stock price OR equity price OR stock market } 27

27 Figure 7: News-based index of equity market uncertainty compared to marketbased VIX, January 1990 to December Correlation=0.733 Monthly Average of Daily VIX Values Monthly News-Based Index of Equity Market Uncertainty Notes: The news-based index of equity market uncertainty is based on the count of articles that reference economy or economic, and uncertain or uncertainty and one of stock price, equity price, or stock market in 10 major U.S. newspapers, scaled by the number of articles in each month and paper. The news-based index and the VIX are normalized to a mean of 100 over sthe period.

28 Newspaper-based index of equity market uncertainty compared to market-based VIX (30 day), January 1990 to August (?) 2014 Correlation of monthly time series = 0.75 Notes: The newspaper index is based on the count of articles that reference economy or economic, and uncertain or uncertainty and one of stock price, equity price, or stock market in 10 major U.S. newspapers, scaled by the number of articles in each month and paper. The 29 newspaper-based index and the VIX are normalized to a mean of 100 over the period.

29 Another Check: news based indices for tracking unemployment also seem to work reasonably well Unemployment news search Correlation= Unemployment rate Notes: Index of Unemployment News composed of quarterly news articles containing terms like unemployment, layoffs, or job loss (scaled by the smoothed total number of articles) in 5 newspapers (WP, BG, LAT, WSJ and CHT). Data normalized to 100 from Jan 1900-Dec Unemployment data is overall seasonally adjusted unemployment rate taken from the BLS.

30 Media Slant in EPU News Coverage? Present but Modest in Size XSSX Slant Media slant measure from Gentzkow and Shapiro (2010). 31

31 The impact of media slant on our EPU index is modest Bush I Clinton Bush II Obama 5 most Republican papers 5 most Democrat papers Papers sorted using the media slant measure from Gentzkow and Shapiro (2010).

32 Assessing Accuracy Concerns Working under our direction, several RA s read 9,000+ newspaper articles, following a 63-page audit guide to code articles: economic uncertainty =0/1, economic policy uncertainty =0/1, and more 33

33 Accuracy Concerns, 2 Conduct a human audit of articles that satisfy automated search criteria for EU Use the audit to: i. Assess error properties of EPU counts ii. Refine the automated search criteria iii. Iterate over steps i and ii to minimize accuracy errors (but without repeating human audit) iv. Model the properties of residual errors, if necessary v. Use human reading to extract additional information that is difficult to uncover by automated search methods 34

34 Accuracy Concerns, 3 Designing the Audit Template 1. Design a provisional audit template 2. Read some articles, record results 3. Assess adequacy of audit template 4. Refine audit template 5. Iterate over steps above until audit template is satisfactory Note: The authors handled the design of the provisional audit template, refined the audit process based on their readings of sampled articles, and repeated this process several times. They then took the audit template to a pre-audit phase, which led to small additional modifications to the audit template. Based on an initial scaled-up audit, the authors revised the audit guide and auditor training procedures. 35

35 Accuracy Concerns, 4 Constructing the Audit Universes Define the newspaper-specific Audit Universe: All articles coded EU=1 by automated search. By construction, this approach omits articles falsely coded as EU=0 by our automated search criteria. Provisional assumption, testable: In practice, we find that the EU false positive rate is around 20 percent in U.S. newspapers. 36

36 Accuracy Concerns, 5 Creating the Audit Samples 1. Two-stage sample design for each paper: First, sample randomly over months (w/o replacement). Second, sample randomly over articles within the month. This design automatically adjusts for changes over time in newspaper size, # of articles, and level of economic uncertainty. 2. Assign samples to auditors, with about 1/4 th of articles assigned to 2 auditors. 3. Randomize the order in which a given auditor reviews and codes assigned articles Ensures that auditor learning does not bias an assessment of differences across papers or over time. 37

37 Human Readings and Automated Computer Methods Yield Similar News- Based EPU Indexes, 1985Q1 to 2012Q Using the Audit Sample and our preferred policy term set. Computer Human Note: Based on random samples of 45 articles per quarter (fewer prior to 1993). For these articles, we calculate quarterly EPU rates based on human readings (red line) and based on automated computer methods (blue line).

38 Humans and computers also give similar results over the long-span period from 1900 to Correlation=0.837 Computer Human year Based on audit sample of 3727 articles (average of 34 per year) in the LA Times and NY Times, the two papers we could audit over entire period from 1900 to 2012.

39 Selec1ng a Preferred Term Set Consider permuta1ons of policy- relevant terms: e.g., regula1on, budget, spending, policy, deficit, tax, federal reserve, white house, house of representa1ves, government, congress, senate, president, and legisla1on. Interpre1ng the human coding as truth, select the term set that minimizes the sum of false posi1ve and false nega1ve error rates We also tried mul1- word terms such as government policy. 40

40 Error rates for 28,000 permutations of 14 policy terms in a human audit sample of 3,500 randomly selected articles Permutations of 14 policy terms: regulation, budget, deficit, tax, federal reserve, government, congress, senate, president, legislation, government spending, federal spending, etc. Our preferred policy term set: {congress, deficit, federal reserve, legislation, regulation, white house } False positives and false negatives expressed as a fraction of true EPU i.e., as a fraction of false negatives + true positives.

41 Other Audit Results, Correla1on of news- based EPU error rate and real GDP growth rate = (quarterly data) Correla1on of news- based EPU error rate and true EPU = (quarterly data) Only 2% of EPU=1 ar1cles are mainly about low or declining uncertainty. Among EPU=1 ar1cles in the Audit Sample: 69% discuss uncertainty about what or when 40% discuss uncertainty about effects 21% discuss uncertainty about who nearly doubles in presiden1al elec1on years 42

42 The human-computer differences in EPU are nearly uncorrelated with real GDP growth rates Correlation=0.071 Yearly GDP growth Computer human EPU % diff Yearly GDP growth Yearly economic policy uncertainty index based on human audit of 3727 articles in the LA Times and New York Times (the two papers we could audit from 1900 to 2012) in a 3-year moving average to yield an average of 121 articles per year.

43 Market Evidence Rapid market adoption suggests that our EPU measures have informational value I) Goldman Sachs, Citibank, JP Morgan, Blackrock, Wells Fargo, IMF, Federal Reserve, ECB and other commercial and policy institutions are using our data or designing their own versions. See for examples. II) Bloomberg, FRED, Reuters and Haver Analytics carry our data and make them available to their financial and policy users

44 4. A Different Text Source: FOMC Beige Book 45

45 Figure 9: The frequency of uncertainty and policy-related uncertainty discussions in FOMC Beige Books rose sharply after Uncertainty 1983Q4 2012Q4 Policy Uncertainty Note: Plots the frequency of the word uncertain in each quarter of the Federal Open Market Committees (FOMC) Beige Book. The Beige Book is an overview of economic conditions of about 15,000 words in length prepared two weeks before each FOMC meeting. The count of Policy Uncertainty uses a human audit to attribute each mention of the word uncertain to a policy context (e.g. uncertainty about fiscal policy) or a non-policy context (e.g. uncertainty about GDP growth). See the paper for full details. 46

46 5. Evidence on the Sources of EPU: -- Newspapers -- Beige Books 47

47 What drives recent rise in U.S. economic policy uncertainty? Note: Analysis uses Newsbank coverage of around 1000 US national and local newspapers See Table 1 in the Baker, Bloom and Davis (2013) for a more detailed analysis.

48 What drives recent rise in U.S. economic policy uncertainty? Newspaper articles point mainly to concerns about fiscal and healthcare policies Note: Analysis uses Newsbank coverage of around 1000 US national and local newspapers See Table 1 in the Baker, Bloom and Davis (2013) for a more detailed analysis.

49 Beige Books Tell a Similar Story Counts By Category and Selected Time Period, 1983Q3 to 2013Q Q Q1 Gulf War I 1993 Q Q3 Clinton Tax Reforms 2001 Q Q2 9/11 AJacks 2002 Q Q2 Gulf War II 2004 Q Q4 Bush/Kerry ElecNon 2008 Q Q4 Lehman's and recession 2010 Q Q1 Debt- ceiling crisis 1983 Q Q1 Overall Average Overall Economic Uncertainty Economic Policy Uncertainty All Fiscal Maoers Taxes Only Spending Only Monetary Policy Health Care Na1onal Security and War Financial Regula1on Sovereign debt, currency crisis U.S. Elec1ons and Leadership Changes Other Specified Policy Maoers Poli1cs, Unspecified Sum of Policy & Poli1cs Categories

50 But also Highlight Concerns Related to Financial Regula1on, Sovereign Debt, and Poli1cs 1990 Q Q1 Gulf War I 1993 Q Q3 Clinton Tax Reforms 2001 Q Q2 9/11 AJacks 2002 Q Q2 Gulf War II 2004 Q Q4 Bush/Kerry ElecNon 2008 Q Q4 Lehman's and recession 2010 Q Q1 Debt- ceiling crisis 1983 Q Q1 Overall Average Overall Economic Uncertainty Economic Policy Uncertainty All Fiscal Maoers Taxes Only Spending Only Monetary Policy Health Care Na1onal Security and War Financial Regula1on Sovereign debt, currency crisis U.S. Elec1ons and Leadership Changes Other Specified Policy Maoers Poli1cs, Unspecified Sum of Policy & Poli1cs Categories

51 And NEVER Breathe a Word about about Monetary Policy Uncertainty! 1990 Q Q1 Gulf War I 1993 Q Q3 Clinton Tax Reforms 2001 Q Q2 9/11 AJacks 2002 Q Q2 Gulf War II 2004 Q Q4 Bush/Kerry ElecNon 2008 Q Q4 Lehman's and recession 2010 Q Q1 Debt- ceiling crisis 1983 Q Q1 Overall Average Overall Economic Uncertainty Economic Policy Uncertainty All Fiscal Maoers Taxes Only Spending Only Monetary Policy Health Care Na1onal Security and War Financial Regula1on Sovereign debt, currency crisis U.S. Elec1ons and Leadership Changes Other Specified Policy Maoers Poli1cs, Unspecified Sum of Policy & Poli1cs Categories

52 More on The Intensity and ComposiNon of Policy- Related Economic Uncertainty over Time Drilling into the news ar1cles about EPU, we find: Big role for uncertainty related to na1onal security issues around the 1me of Gulf War I and in the wake of Uncertainty related to taxes and government spending policies are the biggest factors responsible for historically high levels of policy uncertainty in Although less pronounced, we also find elevated levels of uncertainty in in several other policy categories: en1tlement programs, healthcare, and regula1on. Many of the same categories are elevated in

53 Fiscal policy matters and health care are the most important sources of policy uncertainty in , according to our news-based analysis. Note: This analysis uses Newsbank data to obtain a greater density of news articles.

54 Big 5 Sources of Economic Policy Uncertainty, 1985Q1 to 2013Q3 160 Monetary 140 Fiscal Health Security Regula1on Jan- 85 Jul- 85 Jan- 86 Jul- 86 Jan- 87 Jul- 87 Jan- 88 Jul- 88 Jan- 89 Jul- 89 Jan- 90 Jul- 90 Jan- 91 Jul- 91 Jan- 92 Jul- 92 Jan- 93 Jul- 93 Jan- 94 Jul- 94 Jan- 95 Jul- 95 Jan- 96 Jul- 96 Jan- 97 Jul- 97 Jan- 98 Jul- 98 Jan- 99 Jul- 99 Jan- 00 Jul- 00 Jan- 01 Jul- 01 Jan- 02 Jul- 02 Jan- 03 Jul- 03 Jan- 04 Jul- 04 Jan- 05 Jul- 05 Jan- 06 Jul- 06 Jan- 07 Jul- 07 Jan- 08 Jul- 08 Jan- 09 Jul- 09 Jan- 10 Jul- 10 Jan- 11 Jul- 11 Jan- 12 Jul- 12 Jan- 13 Jul- 13 Note: This chart is a quarterly version of Table 1 in Measuring Economic Policy Uncertainty by Baker, Bloom and Davis. It shows the 5 most important sources of economic policy uncertainty based on frequency counts of newspaper articles.

55 Fiscal Cliff Sequester Downloaded from PolicyUncertainty.com On 18 October 2013 Government Shutdown And Debt- Ceiling Fears

56 Debt Ceiling and Government Shutdown in U.S. Newspapers, Jan to Sep Fraction of Articles In U.S. Newspapers Summer 2011 Debt-Ceiling Dispute 1995 Shutdown This chart reports frequency counts of articles containing debt ceiling or government shutdown, expressed as a fraction of all articles in U.S. newspapers covered by Access World News Newsbank Service (1000+ newspapers). Fiscal Cliff

57 Figure 3: Federal Tax Code Expirations Index, Federal Tax Code Expiration Index Additional evidence that tax policy uncertainty was high in recent years Notes: Based on Congressional Budget Office data on projected revenue effects of federal tax code provisions set to expire in the current calendar year and next ten years. For a given year, the index value is calculated as the discounted sum of projected revenue effects associated with expiring tax code provisions, using a discount factor of 0.5^T applied to future revenue effects for T=0,1, 10 years. Index normalized to a mean of 100 before 2010.

58 Figure 3: Federal Tax Code Expirations Index, Federal Tax Code Expiration Index Undiscounted projected 10-year revenue impact of scheduled tax code expirations: Before 2003 < $250 billion : $3-5 trillion 2013: Huge drop due to Fiscal Cliff resolution Notes: Based on Congressional Budget Office data on projected revenue effects of federal tax code provisions set to expire in the current calendar year and next ten years. For a given year, the index value is calculated as the discounted sum of projected revenue effects associated with expiring tax code provisions, using a discount factor of 0.5^T applied to future revenue effects for T=0,1, 10 years. Index normalized to a mean of 100 before 2010.

59 6. U.S. Economic Policy Uncertainty: A Longer Term Perspective 60

60 Data back to 1900 notable secular rise since 1960s (Jan 1900 Dec 2012) Policy Uncertainty Index Gold Standard Act Great Depression, New Deal and FDR Assassination of McKinley Versailles conference Start of WW I McNary Haughen farm bill Great Depression Relapse Truman- Dewey election Black Monday OPEC II OPEC I Watergate Berlin Conference Gulf War I Lehman and TARP 9/11 and Gulf War II Asian Fin. Crisis Debt Ceiling Notes: Index of Policy-Related Economic Uncertainty composed of quarterly news articles containing uncertain or uncertainty, economic or economy, and policy relevant terms (scaled by the smoothed total number of articles) in 6 papers (WP, BG, LAT, WSJ, CHT, NY Times). Data normalized to 100 from

61 Two possible types of explanations for the secular rise in U.S. EPU since the 1960s Rising size of government: - Rising share of government expenditure/gdp - Rising scale and complexity of of government regulation Increased political polarization: - Increasing polarization of congress by various metrics - Polarization of electorate rose, too, but much less so than polarization of Congress For more on these topics, see Baker et al. (AEA Papers & Proceedings, May 2014) and Marina Azzimonti s working paper on Partisan Conflict.

62 Figure 15: Policy uncertainty and government expenditure share of GDP Policy Uncertainty Government expenditure as share of GDP Notes: Index of Policy-Related Economic Uncertainty composed of quarterly news articles containing uncertain or uncertainty, economic or economy, and policy relevant terms (scaled by the smoothed total number of articles) in 6 newspapers (WP, BG, LAT, NYT, WSJ and CHT). Data normalized to 100 from Jan Dec Government expenditure is total federal, state, and local expenditures over GDP, annually.

63 But UK Policy Uncertainty index is flat despite rising spending Salisbury resignation Salisbury loses NCV Boer war Start of WWI Government spending, % of GDP Irish war Atlee- Churchill election Start of WWII 3-day week and 2 national elections Pound devalues 30% MacMillan resigns Policy Uncertainty IMF crisis Falklands war Thatcher forced to resign ERM exit, Major election Notes: Index of Policy-Related Economic Uncertainty composed of quarterly news articles containing uncertain or uncertainty, economic or economy, and policy relevant terms (scaled by the smoothed total number of articles) in the Guardian and the Observer.

64 The U.S. Has Become More Politically Segregated With Respect to Where People Choose to Live

65 Nevertheless, most districts have remained fairly centrist, while congress has polarized dramatically Notes: Source..

66

67 Political Polarization in the U.S. Congress Has Greatly Intensified in Recent Decades 68

68 Congress became increasingly polarized based on data for both voting and funding Notes: DW-Nominate index is the difference between the mean Republican and mean Democratic DW-Nominate roll-call ideal points generated by McCarty, Poole, and Rosenthal. Bonica (2013) has created ideological scores for each member of Congress based on scaling of campaign contribution records. We display the difference between the means of Republicans and Democrats in Congress.

69 Other explanations for polarization, all of which are hard to square with important aspects of the data Income polarization: but does not match centrist voting Media polarization: but does not match centrist voting Primaries: intuitively appealing, but introduction of primaries not correlated with polarization (Hirano et al. 2010) Campaign finance, individual contributors appear more polarized than Political Action Committees (Barber, 2013)

70 7. Evidence on the Effects of Economic Policy Uncertainty 71

71 How Could Policy Uncertainty Hold Back the Economy? Poten1al Mechanisms (Not an Exhaus1ve List) 1. More precau1onary savings and deleveraging by households 2. When investment and hiring decisions are costly to reverse, greater uncertainty depresses and delays investment and hiring (Bernanke, 1983) 3. More costly debt and equity finance (Gilchrist et al., 2010, and Pastor and Veronesi, 2012, 2013) 4. Higher markups, intensifying monopoly distor1ons (Fernandez- Villaverde et al., 2011) 5. Managerial risk aversion (Panousi and Panikolaou, 2011) 6. Intensifica1on of agency problems, reducing the value of new and exis1ng employment, business and financial rela1onships (Narita, 2011)

72 Assessing the Effects of EPU: Summary of our Work 1. Micro approach: Exploit differences in exposure to government purchases contracts to estimate the effects of EPU on firm-level investment, employment growth and implied equity return volatility (working through government spending channel). 2. Macro approach: Include our EPU measure in an otherwise standard VAR model of aggregate dynamics. Estimate the effects of EPU shocks on aggregate output, investment and employment. 73

73 Micro evidence: exploiting differences in exposure to government contract awards Source: Federal Registry of Contracts (1999 to 2013)

74 Cross-Firm Effects of EPU Changes on Investment Rates and Employment Growth Using annual firm-level data from

75 Key variable of interest is the interaction between aggregate EPU change and firm-level exposure to government contract awards. Firm-level exposure measures are time invariant. We calculate exposure as weighted SIC-level government contract intensity, with weights given by firm s own industry distribution of sales. 76

76 Controls include interactions of firm-level exposure with current change and forecasted future change in government purchases. And interaction with change in VIX, plus firm and time fixed effects. 77

77 Magnitude of Firm-Level Effects For a firm with average investment rate and average contract exposure, doubling EPU results in estimated investment fall of only 0.08 percentage points (~1.2% fall). For firms in the 90 th percentile of exposure rates, the impact is much larger, with predicted investment drops of percentage points, depending on specification and baseline investment. 78

78 Microdata: Firms with greater government exposure show higher uncertainty when EPU is high

79 Figure 12: Estimated Industrial Production and Employment after a Policy Uncertainty Shock Industrial Production Impact (% deviation) Employment Impact (millions) Notes: This shows the impulse response function for Industrial Production and employment to an 102 unit increase in the policy-related uncertainty index, the increase from 2006 (the year before the current crisis) to The central (black) solid line is the mean estimate while the dashed (red) outer lines are the onestandard-error bands. Estimated using a monthly Cholesky Vector Auto Regression (VAR) on the EPU index, log(s&p 500 index), federal reserve funds rate, log employment, log industrial production and linear time trend. Fit to data from 1985 to Months after the economic policy uncertainty shock

80 Figure 13: Robustness of Estimates to Different VAR Specifications Industrial Production Impact (% deviation) Adding VIX and putting it first Three months of lags Baseline Uncertainty index has equal weight on measures Nine months of lags Bivariate (uncertainty and log industrial production) Months after the policy uncertainty shock Notes: This shows the impulse response function for GDP and employment to an 102 unit increase in the policy-related uncertainty index. Estimated using a monthly Cholesky Vector Auto Regression (VAR) of the uncertainty index, log(s&p 500 index), federal reserve funds rate, log employment, log industrial production and time trend unless otherwise specified. Data from 1985 to 2011.

81 Assessing the Effects of EPU: Selected Other Work 1. Geography of Great Recession: Shoag and Veuger (2013) find larger unemployment rate rises from 2006 to 2009 in states with (a) greater increases in state-level uncertainty and (b) institutions less well suited for mediating the effects of a general rise in uncertainty. 2. International Spillovers: The IMF s World Economic Outlook (2013) finds that increases in U.S. and European EPU reduce growth in other regions of the world, with bigger spillover effects from U.S. EPU. 3. Firm-Level Studies: Julian and Yook (2010) and Gulen and Ion (2013) find that political and policy uncertainty retard business investment. Durnev (2010) finds a lower sensitivity of investment to stock prices in election years. 82

82 Policy uncertainty covaries positively with stock market risk (as stocks move together more), especially since 9/11 Our Index of U.S. Economic Policy Uncertainty vs. Average Pairwise Stock Return Correlation Reproduced from Political Uncertainty and Risk Premia, by Lubos Pastor and Pietro Veronesi, University of Chicago, 20 November 2011.

83 8. Another Approach: Policy Developments and Big Stock Market Moves 84

84 Ques1ons and Objec1ves What triggers jumps in na1onal equity markets? What role for policy- related developments as a source of equity market vola1lity? Develop another empirical method for assessing the Policy Uncertainty View And mechanisms through which policy uncertainty affects macroeconomic performance. 85

85 Empirical Method Use news to develop systema1c evidence about reasons for equity market jumps Choose jump threshold to get daily market moves big enough to aoract aoen1on of newspapers Codify explana1ons for jumps offered in next- day news ar1cles Apply to countries in recent decades Extend back to the 1930s or earlier in U.S. and U.K. to provide historical perspec1ve 86

86 How We Analyze Equity Market Jumps 1. Set daily jump threshold 2. Pull dates with market moves > threshold 3. Use newspaper ar1cles to characterize jumps A. Go to online newspaper archive B. Enter newspaper, date range (next day) and search criteria (e.g., stock market ) C. Select ar1cle 4. Read ar1cle. 5. Record the reason for the jump, its geographic source, confidence level, ease of coding, etc. 87

87 Coding the Articles (and the Jumps) Jump by Reason Template Policy Categories Government spending Taxes Non- Policy Categories Monetary policy & central banking Commodi1es Trade & exchange rate policy Regula1on (other than above) Sovereign military & security ac1ons Macroeconomic news & outlook Corporate earnings & profits Unknown/no explana1on Foreign Stock Markets Terrorist aoacks & large- scale violence by non- state actors Other policy (specify) Other non- policy (specify) 88

88 Selec1ng and Coding the Ar1cles We develop a template and Data Construc1on Guide for our team of RAs. The Guide: Explains how to find and select newspaper ar1cles How to read the ar1cles How to code explana1ons for equity market jumps offered in next- day news accounts Defines categories for jumps by reason, and gives examples for each category FAQs that arose as we and RAs worked through the news accounts of equity market jumps 89

89 Step 5 Details: Coding the Ar1cles Category defini1ons Example ar1cles for each category with guidance on how to code them Hard calls and boundary cases The next 16 slides below reproduce a few category defini<ons, example ar<cles, and boundary cases from the Data Construc<on Guide. They give an indica<on of how we provide guidance to the RAs for coding the newspaper ar<cles. 90

90 Government Spending News reports, forecasts, or concerns about government spending and its consequences, including spending matters related to stimulus programs, publicly funded pensions, social security, health care, etc. 91

91 Government Spending 1 This article is coded as government spending because the first reason listed for the stock market plunge is the rejection of the government s bailout plan. The bailout plan itself involves the government spending money to help the economy, and even though it is a rejection of the plan, it is still coded as government spending. 92

92 Taxes News reports, concerns or events related to current, planned, or potential tax changes (e.g., income tax hikes, payroll tax cuts, corporate tax reform, sales tax change, etc.) and their consequences. 93

93 Taxes 2 This article is coded as Taxes because it claims directly that if anything could be cited as a reason it would be the tax bill that was passed. The confidence would be medium to high because the article spends some time discussing the tax bill and claims that the bill was almost certainly the reason, saying if any one reason could be cited it would be that one. 94

94 Monetary Policy and Central Banking Actions, possible actions, and concerns related to the conduct and policies of the central bank or other chief monetary authority. Such actions and policies pertain to interest rate changes and monetary policy announcements, inflation control, liquidity injections by the monetary authority, changes in reserve requirements or other bank regulations used by the monetary authority to exercise control over monetary conditions, lender-of-last resort actions, and extraordinary actions by the monetary authority in response to bank runs, systemic financial crisis and threats to the payments system. 95

95 Monetary Policy and Central Banking 2 This article is coded as Monetary Policy because it cites the reason for the market rally as a statement from the Fed that they will pay less attention to weekly swings in the monetary supply, a change in their policy. The confidence and ease of coding would also be high because the article clearly claims the Fed statement is the reason for the jump. 96

96 Elections and Political Transitions News, events and concerns related to elections, election outcomes, assassinations of political leaders, coups, revolutions, and other political leadership transitions. 97

97 Elections and Political Transitions 1 This article is coded as Elections and Political Transitions because it cites Obama s election as the reason for the market movement. It receives a low confidence ranking, because it claims that only some commentators came to this conclusion, rather than declaring the reason with greater assurance. 98

98 Sovereign Military and Security Actions Reports and concerns about military actions by sovereign actors including war, invasion, blockade, saber rattling, and large-scale violent suppression of domestic insurrections. Policy responses to terrorist actions that involve largescale use of military resources also fall into this category. 99

99 Sovereign Military and Security Actions 2 This article would be coded as Sovereign Action because it claims that stocks plunged due to the military setbacks in Iraq. Since these are military actions sanctioned by the US government, it is Sovereign Action rather than Non-Sovereign. It would receive a 2 or 3 confidence because it declares that the stocks plunged after the setbacks and correlates projected falls to future losses, but it merely states that the stocks plunged after the actions, not because of them. 100

100 Macroeconomic News and Outlook News relating to macroeconomic forecasts or reports such as inflation, housing prices, unemployment or employment, personal income, industrial production, manufacturing activity, etc. Also included in this category: News about financial crisis developments that does not fall into another category such as Monetary Policy and Central Banking. Trade and exchange rate news NOT attributed to policy (e.g., news about trade deficits or currency movements) 101

101 Macroeconomic News and Outlook 3 This article claims that the reason for the market move was a fear of a double-dip recession, a change in the Macroeconomic Outlook. Therefore the article would be coded as Macroeconomic News and Outlook. The confidence would be high because the article clearly declares that the fear of recession was the cause for the movement. 102

102 Distinguishing Monetary Policy & Central Banking from Macroeconomic News & Outlook 1 Some news articles that discuss market reactions to macro developments also discuss the Fed s normal response to the macro development. Generally, we code an article as Macro News & Outlook if it attributes the market move to news about the macro economy. We code it as Monetary Policy & Central Banking if the article attributes the market move to (a) news about how the Fed responds to a given macro development or (b) news about unexpected consequences of Fed actions. It is helpful to approach this classification issue from a Taylor Rule perspective. Consider the following cases: 103

103 Distinguishing Monetary Policy & Central Banking from Macroeconomic News & Outlook 2 1. Macro news: The market moves because it anticipates or speculates (or sees) that the Fed will respond in its usual manner to news about the macro economy. That is, the market anticipates or speculates that the Fed will respond to macro developments according to a Taylor Rule or other well-defined, well-understood description of the Fed's interest-rate setting behavior. 2. Monetary policy: The market moves because of a surprise change in the policy interest rate -- i.e., a surprise conditional on the state of the macro economy. From a Taylor Rule perspective, we can think of this change as a new value for the innovation term in the Taylor rule. 104

104 Distinguishing Monetary Policy & Central Banking from Macroeconomic News & Outlook 3 3. Monetary policy: The market moves because of an actual or potential change in the Fed s policy rule. From a Taylor Rule perspective, this event corresponds to an actual or potential change in the form of the Taylor Rule or a change in specific parameter values. A concrete example would be a big market response to proposals to increase the target interest rate. 4. Monetary policy: The market moves because of news that leads to a revised views or concerns about the consequences of the Fed's actual or anticipated actions. Articles in category 1 get coded as Macro News & Outlook. Articles in categories 2, 3 and 4 get coded as Monetary Policy & Central Banking 105

105 Foreign stock markets News reports that attribute the large domestic market move directly to foreign stock-market moves without any further explanation. For example, an article blaming stock price moves on Wall Street without any further reason 106

106 Deep Insights from News ArNcles Example: "Mostly, the selling of stocks appeared to come from tired holders. As reported with authority by the New York Times the day after a 3% drop in the S&P 500 on 22 March

107 Deep Insights from News ArNcles Example: "Mostly, the selling of stocks appeared to come from tired holders. As reported with authority by the New York Times the day after a 3% drop in the S&P 500 on 22 March OK, sometimes the proffered explanation is nonsensical or vacuous. Usually, though, the explanation has content and is not silly or obviously wrong. Are the explanations for a given jump reasonably consistent across newspapers and coders? Yes, we checked that for U.S. newspapers from 1980 to

108 High Rates of Agreement about Jump Reason across Newspapers and Readers: United States, Share of articles in most common category Policy/non-policy 14 categories Jump threshold: +/-2.5% in S&P 500 # of Equity Market Jumps: 280 # of Newspapers: 5 # of Readers per Newspaper per Jump: 2 à 10 reads per jump, 8.8 reads per jump in practice Notes: The chart plots average rates of agreement about the reason for daily equity market jumps across 5 newspapers, with two readers per paper per jump. Thus, we have a maximum of 10 (5 times 2) readings for each of 280 jumps of greater than +/-2.5% in the S&P 500 Index. In practice, the average number of readings per jump is 8.8, because we do not always find a next-day news article about the jump in every newspaper. The newspapers are the Boston Globe, Los Angeles Times, New York Times. Wall Street Journal and the Washington Post. For the 16 individual categories, we report the fraction of readings attributed to the most commonly attributed reason for the jump event. The plots interpolate across years with no jumps.

109 Detailed Summary of Data for Japan 110

110 Reproduced from What Triggers Equity Market Jumps? by Baker, Bloom and Davis. Work in progress.

111 U.S. Equity Market Jumps Per Year,

112 U.S. Equity Market Jumps Per Year,

113 Time Period Jumps Per Year AJribuNons by Geographic Source Geogra phic Source Country of Equity Market Jump USA UK Germany Australia China (HK) Ireland South Africa to 2007 US Europe < Asia , for US, UK US Europe None None Asia to 2009 US Europe Asia to 2011 US Europe Asia 0.5 None 2.0 None