The Signal and the Noise
Page 55
6. Bruno de Finetti, “La Prévision: Ses Lois Logiques, Ses Sources Subjectives,” Annales de l’Institut Henri Poincaré, 7 (1937).
7. Markets provide us with valuable information whether or not we participate in them directly. An economist might hate bananas, refusing to buy bananas at the supermarket at any price, but she still might be interested in knowing what they cost, to help her calculate the inflation rate. Or an orchard might be interested to know what the price was to decide whether it should plant more banana trees.
Economists call this property “price discovery,” and it’s one of the key advantages of a free-market economy; the common price for a good or service in the market provides valuable information about the supply and demand for it, and the price can rise or fall in response. If prices are set by a central authority instead, it’s much more cumbersome to determine which goods should be produced.
8. William F. Sharpe, Investments (Englewood Cliffs, NJ: Prentice-Hall, 1978).
9. Caren Chesler, “A Bettor World,” The American, American Enterprise Institute, May/June 2007. http://www.american.com/archive/2007/may-june-magazine-contents/a-bettor-world.
10. Iowa Electronic Markets. http://iemweb.biz.uiowa.edu/pricehistory/pricehistory_SelectContract.cfm?market_ID=214.
11. Per interview with Justin Wolfers and David Rothschild.
12. David Rothschild, “Forecasting Elections: Comparing Prediction Markets, Polls, and Their Biases,” Public Opinion Quarterly, 73, no. 5 (2009), pp. 895–916. http://assets.wharton.upenn.edu/~rothscdm/RothschildPOQ2009.pdf.
13. Intrade has suffered from some systematic—and perhaps predictable—biases. In particular, research by Wolfers and others has shown it suffers from something called favorite-longshot bias, which means that bettors tend to overprice low-probability events. For instance, events that Intrade might suggest had a 1 in 10 chance of occurring in fact occur only 1 in 20 times over the long run, and events that the market implied had a 1 in 30 chance of occurring might in fact happen just 1 time in 100. Correcting for this bias was enough for Intrade to outperform FiveThirtyEight in 2008, according to Rothschild’s paper, while FiveThirtyEight was quite a bit better without this adjustment.
14. Forecast error in this context would typically be measured by root-mean squared error (RMSE).
15. Andy Bauer, Robert A. Eisenbeis, Daniel F. Waggoner, and Tao Zha, “Forecast Evaluation with Cross-Sectional Data: The Blue Chip Surveys,” Economic Review, Federal Reserve Bank of Atlanta, 2003. http://www.frbatlanta.org/filelegacydocs/bauer_q203.pdf.
16. Emile Servan-Schreiber, Justin Wolfers, David M. Pennock, and Brian Galebach, “Prediction Markets: Does Money Matter?” Electronic Markets, 14, no. 3l (September 2004). http://hcs.ucla.edu/lake-arrowhead-2007/Paper6_Watkins.pdf.
17. Several models of the 2010 U.S. House elections, for instance, relied on the Gallup poll and the Gallup poll alone. Although Gallup is usually a very good polling firm, its surveys were way off that year and implied a much larger gain for the Republicans than they actually achieved—perhaps 80 or 90 seats rather than the 63 they actually gained. However, taking the polling average would have gotten you quite close to the right result.
18. “Super Tuesday 2012 on Intrade;” Intrade.com, March 8, 2012. http://www.intrade.com/v4/reports/historic/2012-03-07-super-tuesday-2012/.
19. Nate Silver, “Intrade Betting Is Suspicious,” FiveThirtyEight, September 24, 2008. http://www.fivethirtyeight.com/2008/09/intrade-betting-is-suspcious.html.
20. Nate Silver, “Evidence of Irrationality at Intrade,” “Live Coverage: Alabama and Mississippi Primaries;” FiveThirtyEight, New York Times, March 13, 2012. http://fivethirtyeight.blogs.nytimes.com/2012/03/13/live-coverage-alabama-and-mississippi-primaries/#evidence-of-irrationality-at-intrade.
21. I haven’t actually bet on Intrade, as I would consider it a conflict of interest.
22. More specifically, my hope (or belief) is that my subjective probability estimates (my Bayesian priors) might be better than Intrade’s on average. I consider it less likely that this would be true of the forecast models that we run, since those models make certain simplifying assumptions and since it is important to think carefully about whether your model is flawed or whether the consensus is flawed when there is a divergence between them. (More often than not, the model is flawed.) If you follow the FiveThirtyEight blog, you may sometimes see me writing that you should bet against the FiveThirtyEight model for this reason.
23. Eugene F. Fama, “My Life in Finance,” Annual Review of Financial Economics, 3 (2011), pp. 1–15. http://faculty.chicagobooth.edu/brian.barry/igm/fama_mylifeinfinance.pdf.
24. Eugene F. Fama, “The Behavior of Stock-Market Prices,” Journal of Business, 38, no. 1 (January 1965), pp. 34–105. http://stevereads.com/papers_to_read/the_behavior_of_stock_market_prices.pdf.
25. Ibid., p. 40.
26. Google Scholar search. http://scholar.google.com/scholar?q=BEHAVIOR++OF+STOCK-MARKET++PRICES&hl=en&btnG=Search&as_sdt=1%2C33&as_sdtp=on.
27. According to a search of the Google News archive, Fama’s name was not mentioned in the mainstream press until it appeared in a 1971 New York Times article. Marylin Bender, “Chicago School Foes to the Head of the Class,” New York Times, May 23, 1971. http://query.nytimes.com/mem/archive/pdf?res=F00614F8355F127A93C1AB178ED85F458785F9.
28. William F. Sharpe, “Mutual Fund Performance,” Journal of Business, 39, 1 (January 1966), part 2: Supplement on Security Prices, pp. 119–138. http://finance.martinsewell.com/fund-performance/Sharpe1966.pdf.
29. The sample consists of all mutual funds listed as balanced large-capitalization American equities funds by E*Trade’s mutual funds screener as of May 1, 2012, excluding index funds. Funds also had to have reported results for each of the ten years from 2002 through 2006. There may be a slight bias introduced in that some funds that performed badly from 2002 through 2011 may no longer be offered to investors.
30. Charts A, B, C, and E are fake. Chart D depicts the actual movement of the Dow over the first 1,000 days of the 1970s, and chart F depicts the actual movement of the Dow over the first 1,000 days of the 1980s. Congratulations if you guessed correctly! Send your résumé and cover letter to Mad Money, 900 Sylvan Ave., Englewood Cliffs, NJ 07632.
31. Eugene F. Fama, “Efficient Capital Markets: A Review of Theory and Empirical Work,” Journal of Finance, 25, 2 (1970), pp. 383–417.
32. Per interview with Eugene Fama.
33. Alan J. Ziobrowski, Ping Cheng, James W. Boyd, and Brigitte J. Ziobrowski, “Abnormal Returns from the Common Stock Investments of the U.S. Senate,” Journal of Financial and Quantiative Analysis, 39, no. 4 (December 2004). http://www.walkerd.people.cofc.edu/400/Sobel/P-04.%20Ziobrowski%20-%20Abnormal%20Returns%20US%20Senate.pdf.
34. Google Scholar search. http://scholar.google.com/scholar?hl=en&q=%22efficient+markets%22&as_sdt=0%2C33&as_ylo=1992&as_vis=0.
35. Google Scholar search. http://scholar.google.com/scholar?hl=en&q=%22efficient+markets+hypothesis%22&btnG=Search&as_sdt=1%2C33&as_ylo=2000&as_vis=0.
36. Google Scholar search. http://scholar.google.com/scholar?as_q=&num=10&as_epq=theory+of+evolution&as_oq=&as_eq=&as_occt=any&as_sauthors=&as_publication=&as_ylo=1992&as_yhi=&as_sdt=1&as_subj=bio&as_sdtf=&as_sdts=33&btnG=Search+Scholar&hl=en.
37. John Aidan Byrne, “Elkins/McSherry—Global Transaction Costs Decline Despite High Frequency Trading,” Institutional Investor, November 1, 2010. http://www.institutionalinvestor.com/Popups/PrintArticle.aspx?ArticleID=2705777.
38. Specifically, a linear regression of the sign of that day’s stock price (1 indicating a positive movement and –1 indicating a negative movement) on the previous day’s stock price. The trend is also highly statistically significant if you take a regression of the percentage change in the stock price on the previous day’s percentage change. Note, however, that standard forms of regression analysis assume that errors are normally distributed, whereas the stock market does not obey a normal distribution. Economi
sts like Fama think this is a problem when applying standard statistical tests to analyze patterns in stock prices.
39. Index funds would not have been widely available in 1976; the analysis assumes that the investor’s returns would track that of the Dow Jones Industrial Average.
40. This is much worse than the market-average return. Although the 2000s were a poor decade for stocks, a buy-and-hold investor would have had about $9,000 rather than $4,000 left over by the end of the period.
41. Carlota Perez, “The Double Bubble at the Turn of the Century: Technological Roots and Structural Implications,” Cambridge Journal of Economics, 33 (2009), pp. 779–805. http://www.relooney.info/Cambridge-GFC_14.pdf.
42. Based on a comparison of revenues from technology companies in the Fortune 500 to revenues for all companies in the Fortune 500 as of 2010. The companies I count as technology companies are Amazon.com, Apple, Avaya, Booz Allen Hamilton, Cisco Systems, Cognizant Technology Solutions, Computer Sciences, Corning, Dell, eBay, EMC, Google, Harris, Hewlett-Packard, IBM, Liberty Media, Microsoft, Motorola Solutions, NCR, Oracle, Pitney Bowes, Qualcomm, SAIC, Symantec, Western Digital, Xerox, and Yahoo!
43. Specifically, a 90 percent increase in value over the trailing five years, adjusted for dividends and inflation. This would correspond to a 14 percent annualized increase, twice the long-run rate of 7 percent.
44.
CASES IN WHICH THE S&P 500 INCREASED BY 90 PERCENT OVER A FIVE-YEAR PERIOD
Year(s)
Crash?
Description
1881–83
No
Although stocks lost 36 percent of their nominal value between June 1881 and January 1885, much of this is accounted for by the severe deflation that was gripping the country at the time. This should probably not be considered a crash, although returns were below-average even on an inflation-adjusted basis.
1901
Yes
Panic of 1901. Initially confined to a relatively narrow industry, railroad stocks, which happened to be very popular with small investors, the downturn eventually spread to the broader market with the S&P 500 losing more than a quarter of its value from 1901 to 1903.
1925–30
Yes
Great Depression. Stock prices declined by almost 80 percent over a three-year period from 1929–1932 amid the worst financial crisis in history.
1937
Yes
Recession of 1937–38. Stocks lost more than 40 percent of their value over a 14-month period as the economy went into a severe double-dip recession.
1954–59
No
The early 1950s were an extremely good time to buy stocks as investors benefited from America’s postwar prosperity boom. Toward the end of the 1950s was somewhat less favorable, as stock prices were choppy throughout the 1960s.
1986–87
Yes
Black Monday came on October 19, 1987, with the Dow declining by almost 23 percent in a single day. An investor who persevered through the period would have seen very good returns in the 1990s.
1989
No
An investor buying in 1989 would have experienced the robust returns of the 1990s without the trauma of Black Monday. A good time to buy stocks.
1997–2000
Yes
Dot-com crash. Someone buying the S&P 500 at the top of the market in August 2000 and selling it at the bottom in February 2003 would have lost almost half his money. An investor with a portfolio rich in technology stocks would have done even worse.
45. The original title of Dow 100,000 was apparently the more modest Dow 30,000 before its publisher realized that it was going to be one-upped. I infer this because the original description of the book at Amazon.com refers to it as “Dow 30,000,” emphasizing its prediction that the Dow would increase to 30,000 by 2010. The book also predicted that the Dow would rise to 100,000 by 2020, giving it its new title. See http://www.amazon.com/Dow-100-000-Fact-Fiction/dp/0735201374 retrieved November 25, 2011.
46. By “real return,” I refer to the share price plus dividends but less inflation. I assume that dividends are automatically re-invested in the stock index rather than held.
47. Alan Greenspan, “The Challenge of Central Banking in a Democratic Society,” Remarks at the Annual Dinner and Francis Boyern Lecture of The American Enterprise Institute for Public Policy Research, Washington, DC, December 5, 1996. http://www.federalreserve.gov/boarddocs/speeches/1996/19961205.htm.
48. Charts like those in figure 11-7 look to be extremely rich with data, but they are somewhat deceptive. They include a data point to represent every year of stock-market returns: one of the circles in the chart, for instance, represents how stocks behaved in the twenty years between 1960 and 1980. Another circle represents how they behaved from 1961 to 1981. The problem is that those periods overlap with one another and therefore are double-counting the same data. If we’re looking at how stock prices perform in twenty years at a time, we don’t really have that much data to work with. Shiller’s P/E ratio can first be calculated in 1881. Start in 1881 and count upward twenty years at a time . . . you get to 1901, 1921, 1941, 1961, 1981, and 2001. Then you run out of time after just six data points.
49. You should, of course, make as large a trade as you can if you are certain about what the stock is going to do.
50. “Henry Blodget’s Risky Bet on the Future of News,” Bloomberg Businessweek, July 8, 2010. http://www.businessweek.com/print/magazine/content/10_29/b4187058885002.htm.
51. Dan Mitchell and Scott Martin, “Amazon Up 46 Points; Report ‘Clarified,’” CNET News, December 16, 1998. http://news.cnet.com/2100-1017-219176.html.
52. Amazon.com Inc. (AMZN) Historical Prices; Yahoo! Finance. http://finance.yahoo.com/q/hp?s=AMZN&a=00&b=1&c=1997&d=11&e=25&f=2011&g=d.
53. Amazon.com shares rose to an intraday high of $302 on December 16, 1988, after beginning the day at $243, before closing at $289.
54. Out of the more than one hundred people I interviewed for this book, Blodget is one of just two or three for whom you’d be happy to publish the interview transcript almost word for word.
55. Denver Post, April 16, 1998.
56. The price of a share of Amazon is now superficially six times cheaper than it was in 1998 because of stock splits.
57. Zinta Lundborg, “Report Card: Henry Blodget,” The Street, June 27, 2000. http://www.thestreet.com/markets/analystrankings/977502.html.
58. “Vested Interest;” PBS Now; May 31, 2002. http://www.pbs.org/now/politics/wallstreet.html.
59. Securities and Exchange Commission, 450 Fifth Street, N.W. Washington, DC 20549, Plaintiff,—against—Henry Mckelvey Blodget, Defendant,” United States District Court, Southern District of New York, April 28, 2003. http://www.sec.gov/litigation/complaints/comp18115b.htm.
60. “The Securities and Exchange Commission, NASD and the New York Stock Exchange Permanently Bar Henry Blodget from the Securities Industry and Require $4 Million Payment;” U.S. Securities and Exchange Commission, April 28, 2003. http://www.sec.gov/news/press/2003-56.htm.
61. David Carr, “Not Need to Know but Nice to Know,” MediaTalk, New York Times, November 24, 2003. http://www.nytimes.com/2003/11/24/business/mediatalk-not-need-to-know-but-nice-to-know.html?ref=henryblodget.
62. Crash is defined here as a 20 percent decline in stock prices, net of dividends and inflation.
63. “Securities Industry Employment 2Q 2010;” Securities Industry and Financial Markets Association Research Report, 5, no. 13. http://www.cdfa.net/cdfa/cdfaweb.nsf/fbaad5956b2928b086256efa005c5f78/7b5325c9447d35518625777b004cfb5f/$FILE/SecuritiesIndustry_Employment_20100810_SIFMA.pdf.
64. There is also some evidence that the analysts know privately when a stock is a loser much sooner than they are willing to say publicly, and will tip their firm’s investment clients to this before they inform the general public. See for instance Jeffrey A. Buss, T. Clifton Green, and Narasimhan Jegadeesh, “Buy-Side Trades and Sell-Side Recomme
ndations: Interactions and Information Content,” Emory University, January 2010. http://www.bus.emory.edu/cgreen/docs/busse,green,jegadeesh_wp2010.pdf.
65. Sorin Sorescu and Avanidhar Subrahmanyam, “The Cross-Section of Analyst Recommendations,” Recent Work, Anderson Graduate School of Management, UC Los Angeles, January 9, 2004. http://escholarship.org/uc/item/76x8k0cc;jsessionid=5ACA605CE152E3724AB2754A1E35FC6A#page-3.
66. Floyd Norris, “Another Technology Victim; Top Soros Fund Manager Says He ‘Overplayed’ Hand,” New York Times, April 29, 2000. http://www.nytimes.com/2000/04/29/business/another-technology-victim-top-soros-fund-manager-says-he-overplayed-hand.html?pagewanted=2&src=pm.
67. John C. Bogle, “Individual Investor, R.I.P.,” Wall Street Journal, October 3, 2005.
68. Jonathan Lewellen, “Institutional Investors and the Limits of Arbitrage,” Journal of Financial Economics, 102 (2011), pp. 62–80. http://mba.tuck.dartmouth.edu/pages/faculty/jon.lewellen/docs/Institutions.pdf.
69. Individual investors, because they tend to make smaller trades, account for an even smaller fraction of the trading that takes place. One recent study (Alicia Davis Evans, “A Requiem for the Retail Investor?” Virginia Law Review, May 14, 2009. http://www.virginialawreview.org/content/pdfs/95/1105.pdf) holds that individual investors account for just 2 percent of market volume on the New York Stock Exchange.
70. Say that a trader has the opportunity to bet $1 million of his firm’s capital on a short position in the market. He figures there is a 55 percent chance the market will crash over the next year and a 45 percent chance that it will improve. This is a pretty good bet. If the analyst has calculated the odds right, the expected return from it is $100,000.
But the incentives may look nothing like this for the trader himself. There is a high probability that he will be fired if he gets the bet wrong. Short-sellers aren’t popular on Wall Street, especially when they’re underperforming their peers. He may benefit to some extent if he gets the bet right, but not enough to outweigh this. Better to wait until the crash is unavoidable and all traders fall down together.