What triggers stock market jumps?

Transcription

1 What triggers stock market jumps? Scott R. Baker (Kellogg, Northwestern) Nick Bloom (Stanford) Steven J. Davis (Chicago Booth) Marco Sammon (Kellogg, Northwestern) January 7 th, 2018

2 Why does the stock market jump? Number of papers analyzed stock volatility e.g. Shiller (1981) or Roll (1988) finding it is hard to fully justify Closer to our work are paper like Cutler, Poterba and Summers (1988) who linked big-jumps and news stories

3

4

5 Why does the stock market jump? Number of papers analyzed stock volatility e.g. Shiller (1981) or Roll (1988) finding it is hard to fully justify Closer to our work are paper like Cutler, Poterba and Summers (1988) who linking big-jumps and news stories Recently a number of papers have been using text search to connect stock jumps to news (e.g. Boudoukh et al 2012)

6 Results Overview Study newspapers day after over /- 2.5% jumps across 15 countries back to 1980 (1900 if possible) with 22 person RA team Main findings: 1.Policy is important: 40% US jumps attributed to policy: more than macro (27%) or corporate earnings (11%) (26% globally) 2.Policy jumps more positive: 57% policy-jumps positive in US vs 42% for non-policy (56% vs 43% globally) 3.US dominates globally: Outside US newspapers attribute 34% of jumps to US above 11% US GDP share and this is rising 4.[Fiscal policy jumps particularly negative for future stock returns, volatility and GDP growth]

7 Data: Drivers of Stock Market Jumps US Results: Types of Jumps International Results Implications of Different Jumps

8 Defining Jumps US jump definition: Days where the CRSP Value-Weighted Index (VWI) has an absolute return of at least 2.5% Use the following day s newspaper to categorize jumps To maximize accuracy: All US jumps analyzed by 2+ RAs differences discussed Post WWII jumps analyzed in WSJ, NYT, BG, WP & LAT Created 101 page audit guide and RA training program

9 Example 1 (9/29/2008, -8.7%): Government Spending This article is coded as Government Spending (Policy) 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. Geographic source would be the US.

10 Example 2 (8/5/ %): Macro News 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. Geographic source would be the US.

11 Example 3 (3/24/1980, -2.6%): Marco News This article is coded as Macroeconomic News and Outlook because it claims that the market dropped because of rising interest rates. However, it does not refer to the Fed as being responsible for these rates, and hints at the rates just shifting with the economy, and so is not Monetary Policy. The article would also have a high confidence and ease of coding because the article is very clear about the rates being responsible for the movement.

12 Example 4 (2/28/ %): Foreign Stock Markets This article is coded Foreign Stock Markets because it describes the market movements in Shanghai as being responsible for changing the market. The article would have a high confidence level because it clearly describes Shanghai s markets as being responsible, and refers to this reasoning later in the article as well. Geographic source would be the China

13 Example 5 (4/10/2001, +2.7%): Unknown This article is coded as Unknown because the author clearly states that there was no single catalyst, signifying that no one knows why the market moved.

14 How reliable are these jump codings? Two potential concerns about the method: 1. RAs reading same paper may code jumps differently 2. Results may depend on the newspaper consulted To evaluate these concerns we: 1. Use multiple RAs for same paper - calculate agreement 2. Use multiple papers - Boston Globe, LA Times, NY Times, WSJ, Washington Post - calculate agreement

15 Test 1: Agreement Across Coders (same paper) P(Agree) Policy Agreement 16 Category Agreement average 89% average 75% Notes: Agreement is the share of codings (at the coder-level) that agree, averages are shown for each year.

16 Test 2: Agreement Across Papers (different coders) P(Agree) average 79% average 49% Categories Policy Notes: There are 9 coders, across 5 newspapers each day. Agreement is the share of pairs that agree, out of 36 possible pairs. Dots represent an annual average of the share of pairs which agree.

17 Data: Drivers of Stock Market Jumps US Results: Types of Jumps International Results Implications of Different Jumps

18 US Jumps by Year (policy 40%) Downward Jumps Upward Jumps Panic 1901 Panic Banking Panic of 1907 WWI Depression Lowest Growth (1932) 1929 Crash Second Downturn (1937) WWII Black Oil Shock Monday Tech Boom/ Bust Global Financial Crisis Unknown + No Article Policy Non-Policy Notes: Each bar shows the number of positive or negative jumps in the year. Shadings indicate the number of jumps triggered by Policy and Non-Policy news. The residual category reflects jumps attributed to unknown causes by the newspaper article and instances in which we could not identify an article that discusses the jump (this only occurs for jumps occurring prior to 1926).

19 Policy share increases with higher jump threshold (policy matters even more for really big jumps) Jump Threshold # Jumps Policy Share Post-War Policy Share 2.5% % 36.7% 3.00% % 40.4% 3.50% % 43.3% 4.00% % 41.4% 4.50% % 41.4% 5.00% % 41.8% Notes: Only using US data, 1926-present. Post-war is (inclusive).

20 Category shares: 522 Prewar & 392 Postwar jumps Macroeconomic News and Outlook Unknown Corporate Earnings and Profit Monetary Policy and Central Banking [P] Sovereign Military and Security Actions [P] Government Spending [P] Commodities Regulation [P] Other NonPolicy Other Policy [P] Elections and Political Transitions [P] Taxes [P] Trade and Exchange Rate Policy [P] Foreign Stock Markets NonSovereign Military and Actions/Terrorism 11 6* * 83* 83* 68 59* 246* Pre-War Post-War Notes: Categories with a 5% significant difference in pre-war vs. post-war share of total jumps marked by an *. Ex: Macro s share increased from 23% prewar to 33% postwar. [P] denotes policy categories Labels denote total number of jumps in each category over all years. Post-war is (inclusive).

21 Categories for 437 Positive & 479 Negative jumps Macroeconomic News and Outlook Unknown Corporate Earnings and Profit Monetary Policy and Central Banking [P] Sovereign Military and Security Actions [P] Government Spending [P] Commodities Regulation [P] Other NonPolicy Other Policy [P] Elections and Political Transitions [P] Taxes [P] Trade and Exchange Rate Policy [P] Foreign Stock Markets NonSovereign Military and Actions/Terrorism * 39 31* * 59* 68* * 83* Positive 246* Negative Notes: Categories with a 5% significant difference in share of positive vs. negative jumps marked by an *. Ex: Macro makes up 30% of negative jumps, and 24% of positive jumps. [P] denotes policy categories. Labels denote total number of jumps in each category.

22 Data: Drivers of Stock Market Jumps US Results: Types of Jumps International Results Implications of Different Jumps

23 Global sample selects countries with active stock-markets and good on-line press archives Country Start Sources Jump Threshold United States 1885 Wall Street Journal 2.50% United Kingdom 1930 Financial Times (UK Edition) 2.50% Australia 1985 Australian Financial Times 2.50% Canada 1980 The Globe and Mail 2.00% China (Hong Kong) 1988 South China Morning Post 3.80% China (Shanghai) 1994 Shanghai Securities Journal 4.00% Germany 1985 Handelsblat, FAZ 2.50% Greece 1989 Kathimerini, To Vima 4.00% Ireland 1987 The Irish Times 2.50% Japan 1981 Yomiuri and Asahi 3.00% New Zealand 1996 New Zealand Herald 2.50% Saudi Arabia 1994 Al Riyadh 2.50% Singapore 1980 Business Times and Straits Times 2.50% South Africa 1986 Business Day 2.50% South Korea 1980 Chosun Ilbo 2.50% Jump threshold was chosen such that jumps were approximately 1% of trading days

24 Count by year for the UK (policy 33%) Downward Jumps Upward Jumps Great Depression WWII Suez Crisis Recession and 1976 IMF crisis Sterling Crisis Black Monday Tech crash Global Financial Crisis Unknown + No Article Policy Non-Policy Notes: Each bar represents the number of jumps from each category within a given year

25 Yearly Count - All Countries (policy 26%) Downward Jumps Upward Jumps Early 1990 s Recession Asian/Russian Financial Crisis Global Financial Crisis European Debt Crisis Unknown + No Article Policy Non-Policy Notes: Each bar is the average number of jumps per year across the following countries: Australia, Canada, China (HK), China (Shanghai), Germany, Greece, Ireland, Japan, New Zealand, Saudi Arabia, Singapore, South Africa, South Korea and UK

26 Policy Split by Country share of known is 41% Shanghai US Japan Korea South_Africa Ireland Canada UK Greece Germany Australia Singapore Hong_Kong Saudi_Arabia New_Zealand Policy No Article + Unknown Non-Policy Notes: Each bar is the share of jumps by category within each country. All years available for each country are used.

27 Policy share increases with higher jump threshold (policy matters even more for really big jumps) Increase Initial Threshold By # Jumps Policy Share 0.0% % 0.50% % 1.00% % 1.50% % 2.00% % 2.50% % Notes: Left panel direct replication of US threshold size and policy share table. Right panel Adding to country-level thresholds to account for different levels of volatilities across countries. Includes Australia, Canada, China (HK), China (Shanghai), Germany, Greece, Ireland, Japan, New Zealand, Saudi Arabia, Singapore, South Africa, South Korea and UK

28 Positive/Negative by Policy & Non-Policy: As in the US Policy jumps typically more positive Policy Non-Policy t-stats Country Share Pos. Share Neg. Share Pos. Share Neg. Policy Positive Non-Policy Positive Australia 55.6% 44.4% 38.8% 61.2% Canada 51.5% 48.5% 40.1% 59.9% Germany 54.1% 45.9% 41.2% 58.8% Greece 60.2% 39.8% 49.6% 50.4% Hong_Kong 64.6% 35.4% 41.5% 58.5% Ireland 57.0% 43.0% 41.7% 58.3% Japan 61.4% 38.6% 42.1% 57.9% Korea 61.5% 38.5% 47.9% 52.1% New_Zealand 66.7% 33.3% 29.3% 70.7% Saudi_Arabia 54.3% 45.7% 38.9% 61.1% Shanghai 50.5% 49.5% 43.3% 56.7% Singapore 55.1% 44.9% 46.8% 53.2% South_Africa 51.1% 48.9% 45.8% 54.2% UK 49.5% 50.5% 45.2% 54.8% US 56.5% 43.5% 41.7% 58.3% All 55.9% 44.1% 43.2% 56.8% Notes: Share of positive and negative jumps by country. Includes all available years for each country. A positive t-stat in Policy Positive denotes policy is more likely to be positive. A negative t-stat in Non-Policy Positive denotes non-policy is more likely to be negative.

29 US news seems to drive a rising share of stockmarket jumps in other countries average 34% US Share Early 1980 s recession 1987 Crash Gulf War I Tech Boom/ Bust Asian/LTCM crises Eurozone Crisis Global Financial Crisis Trend significant at the 5% level Notes: Share of US source of stock-market jumps averaged over all 14 non-us countries (by year) Australia, Canada, China (HK), China (Shanghai), Germany, Greece, Ireland, Japan, New Zealand, Saudi Arabia, Singapore, South Africa, South Korea and UK.

30 US share large compared to GDP share Early 1980 s recession 1987 Crash Gulf War I Asian/LTCM crises Tech Boom/ Bust U.S. Jump Share Global Financial Crisis Eurozone Crisis U.S. GDP Share Notes: Share of US source of stock-market jumps averaged over non-us countries by year: Australia, Canada, China (HK), China (Shanghai), Germany, Greece, Ireland, Japan, New Zealand, Saudi Arabia, Singapore, South Africa, South Korea and UK. Dot size is proportional to the average number of jumps by country/year. GDP share is Gross domestic product based on purchasing-power-parity (PPP) share of world total from the IMF

31 US jumps share is also higher than US trade-share North America: US North America: Canada Asia: Singapore Other: South_Africa Other: Australia Europe: Ireland Asia: Japan Europe: Germany Asia: Korea Asia: Hong_Kong Europe: UK Other: New_Zealand Europe: Greece Other: Saudi_Arabia Asia: Shanghai US Asia Europe Other Notes: Shows share of jumps by geographic origin (bars). Crosses shows the average trade (gross exports + gross imports) from the US as a share of domestic GDP between 1999 and 2015.

32 Data: Drivers of Stock Market Jumps US Results: Types of Jumps International Results Implications of Different Jumps

33 Fiscal policy associated with lower future returns Non-Policy Categories Policy Categories 22-Day Cumulative Returns (1) (2) (3) (4) (5) Return *** ** ** *** (0.052) (0.092) (0.098) (0.099) (0.098) {Return<0} x Return *** ** ** *** (0.139) (0.162) (0.162) (0.160) Jump (0.004) Policy (0.006) Non-Policy (0.006) Commodities (0.023) Corporate Earnings (0.012) Macro News (0.009) Other Non-Policy (0.009) Monetary Policy * (0.010) Fiscal Policy *** (0.013) Military + Other Policy * (0.008) Obs 24,016 24,016 24,016 24,016 24,016 R-Squared Notes: Daily regressions, sample is

34 Fiscal policy also associated with higher volatility Non-Policy Categories Policy Categories 22-Day Realized Volatility (1) (2) (3) (4) (5) Return *** *** *** *** *** (0.008) (0.009) (0.011) (0.011) (0.011) {Return<0} x Return *** *** *** *** (0.017) (0.020) (0.020) (0.020) Jump *** Policy * (0.001) Non-Policy *** (0.001) Commodities *** (0.002) Corporate Earnings * (0.001) Macro News *** (0.001) Other Non-Policy *** (0.001) Monetary Policy * (0.001) Fiscal Policy *** (0.002) Military + Other Policy (0.001) Obs 24,016 24,016 24,016 24,016 24,016 R-Squared Notes: Daily regressions, sample is

35 GDP VAR after a jump.2 basic, jumps, gdpc1 0 GDP growth rate step Quarters Notes: Sample is , using quarterly real GDP from FRED. Time is in quarters. All variables normalized to zero mean, and unit standard deviation.

36 GDP response by category.02 Impulse: _Commodities.02 Impulse: _Corporate.02 Impulse: _Macro Impulse: _Monetary.02 Impulse: _Fiscal.3 Impulse: _OtherPolicy Notes: Sample is , using quarterly real GDP from FRED. Time is in quarters. All variables normalized to zero mean, and unit standard deviation.

37 Conclusion 1. 40% US jumps attributed to policy: more than macro (27%) or corporate earnings (11%) news (26% globally attributed to policy) 2. Policy jumps more likely to be positive 57% policy-jumps positive in US vs 42% for non-policy jumps (56% vs 43% globally) 3. Outside the US, newspapers attribute 34% of jumps to US events above 11% US GDP share and this is rising 4. [Fiscal policy jumps negative for future stock returns, volatility and GDP growth] Next steps: 1. Expand country coverage 2. Look at other assets e.g. bonds

38 Back-Up Slides

39 GDP Response: -0.4% in response to Policy Jumps in General Associated with Response to a Policy and Non-Policy Jump.2 policy, policy, gdpc1.2 nonpolicy, nonpolicy, gdpc Notes: Sample is , using quarterly real GDP from FRED Cholesky VAR with jump then 100*log(GDP). Reports impulse of one jump (policy or non-policy).

40 Daily Returns

41 Daily Realized Volatility Plot of cumulative realized volatility (daily squared returns) in the 200 days following a jump

42 US Share of Global GDP Notes: Source: World Bank

43 No difference in average jump size when breaking each category by positive/negative jumps Avg. Jump Positive Jumps Negative Jumps t-stat Diff Category Size # Jumps Avg. Ret. # Jumps Avg. Ret. MacroNews 3.55% % % MonetaryPolicyCentralBanking 3.91% % % 1.28 Unknown 3.85% % % GovernmentSpending 4.51% % % 0.97 CorporateEarningsandProfit 3.60% % % 1.44 SovMilitary 3.70% % % Commodities 4.32% % % 0.70 Regulation 4.07% % % 0.91 ElectionsandPoliticalTransitions 3.87% % % OtherPolicy 4.47% % % 0.02 Taxes 3.81% % % 0.16 TradeandExchangeRatePolicy 3.52% % % OtherNonPolicy 4.59% % % 1.37 NonSovMilitaryTerror 3.88% % % ForeignStockMarkets 4.57% % % - US data, Last column is a t-test for a difference in absolute return size between positive and negative jumps

44 Independence of Newspapers A critique of our method is that the newspapers we consulted are under government control, which could bias the jump attributions away from policy categories Potential concerns: Shanghai Securities Journal is run by Xinhua, which is the official news agency of the PRC Al Riyadh is run by a private company, but the World Association of Newspapers claims it is under the control of Crown Prince Salman Although both of our newspapers for Singapore, the Business Times and the Straits Times are run by Singapore Press Holdings, a public company, there are laws in Singapore which limit the freedom of the press

45 Jump Comparison S&P 500 vs CRSP VW Number of Jumps by Year CRSP VW Index S&P 500 Index There is almost no difference in the number of jumps in the S&P 500 and the CRSP VW Index

46 Notes on Other Non-Policy There are only 8 other non-policy jumps in the post-war period The two largest other non-policy moves that may be driving results for reversals and volatility are: May 28, 1962 which was the worst one day drop since the great depression (-7%). There was a rebound of +6% the following day October 27, 1997 which was the mini-crash related to the Asian financial crisis, when market dropped almost 7% The flash crash had other non-policy as the secondary category, as it occurred the same day as the Greek approval of the IMF bailout

47 Does ordering matter? Bivariate VAR impulse response with number of jumps and industrial production Response of IP to 1 additional jump.5 order(ip #Jumps) order1, _count, indpro order(#jumps IP ) order2, _count, indpro step 95% CI orthogonalized irf step 95% CI orthogonalized irf Graphs by irfname, impulse variable, and response variable Graphs by irfname, impulse variable, and response variable