Unexpectedly Intriguing!
July 29, 2010

Since January 2008, we've made some 75 predictions where we can determine whether or not they were right or wrong, spanning everything from where stock prices might go next to how many wins or losses certain major league baseball teams would have at the end of the regular season in 2010. But we have an ongoing nagging question: are we really any good at it?

Regular readers of Political Calculations know we take the crystal ball business pretty seriously, providing quarterly updates for all the predictions we've made. Better still, we apply the absolutely unforgiving plus-minus statistic from hockey and basketball, where we gain a point if we're right, lose a point when we're wrong, and score a zero for when the outcome of a prediction cannot yet be determined, or in the case where we make multiple predictions that ultimately cancel each other out.

Ultimately, the plus-minus score counts the number of times our predictions were right more than they were wrong. If we're no better at predicting the future than a coin toss, our plus-minus score will gravitate toward zero. If we're better at predicting the future than that, our score will rise over time. And if we're really bad at prognostication, our score will fall over time.

We say unforgiving because we either hit our defined prediction target, or we don't. Near misses don't count. Which is sad for us because we've had more than a few....

Feh. Enough pity. Let's talk about how we've done since our last update.

We went back over our full track record and found a counting error in our plus-minus score through April 2010, which reduced our score from +24 to +23. We then also noticed that we hadn't closed out the second part of a prediction we made back in May 2009, which dropped our plus-minus score to +22, since it worked out to be a split decision.

So how have we done once we incorporate all the latest predictions we've made since April 2010? For the 75 predictions where we've been able to determine their outcome, we've scored 49 as correct and 26 as incorrect, giving us a plus-minus score of +23. As an alternative way of measuring how effective we are at predicting the future, we can say that things have turned out the way we anticipated they would some 65.3% of the time.

We're kind of curious to see how that might compare with others. We'll have to do some research to find out if anyone keeps track of these things, or if like TV weathermen, it just doesn't matter how often you're wrong!

The table below shows the predictions that we either are still waiting to find out if they're correct or not or have been able to make a determination of whether they're correct or not.

Political Calculations' Plus-Minus Score Update, 30 July 2010
Date Prediction Outcome +/- Score
26 May 2009 We pick up on an unlikely correlation between the timing of plunges in the stock market and the results of our preferred method for determining the probability of recession in the U.S. We predict stocks will dive in value around 16-23 June 2009 and 10-16 September 2009. One down, one to go. The timing for the June drop actually coincided with the work week, beginning on Monday, 15 June 2009 and ending a week later. We'll have to see where things stand in September to see if we scored a two-fer here! Update: It was only a one-fer, as stock rose from 10 September 2009 through 16 September 2009. The split decision knocks our score to zero for the two predictions. +0
13 August 2009 We make fun of 47 economists for predicting that the U.S. recession would end in the third quarter, based on the "Cash for Clunkers" program. We point out that the dividend futures data for the S&P 500 has been saying the recession would be over in 2009Q3 for months, long before C4C even became legislation! Too soon to tell. This prediction looks pretty likely, but we'll have to wait for the NBER to get around to declaring the date of the end of the recession. Update: Looking more and more likely. +0
21 December 2009 Using incomplete data for the month of December 2009, economy would dip in the second quarter of 2010, with a slow recovery afterward. We anticipate that meaningful growth in the number of jobs would likely begin with the third and fourth quarters of 2010. We anticipate that the NBER will declare the recession they found to have begun in December 2007 to have ended in the third quarter of 2009, but we make a case for 2010Q2 as a more realistic alternate. Too soon to tell. It will be a while before we get a full confirmation for these predictions. On the potential plus side for us, different branches of the Federal Reserve have used their own models for predicting what the NBER will do to find that July 2009 is the month they will most likely declare to be the ending date for the recession. +0
31 March 2010 Using the finalized data for the fourth quarter of 2009, we project GDP for the first quarter of 2010 will be within 2% of \$13,276.5 billion chained 2005 U.S. dollars! We even give nearly 70% odds of GDP falling between \$13,136 and \$13,417 in the first quarter! We won't know the answer until the BEA releases the finalized GDP data at the end of June 2010. Update: At the end of June 2010, real GDP was announced to be within 0.3% of our target value. Score! +1
4 May 2010 Not having the data we needed to project stock prices, we do something radical - we solve the inverse problem, taking the data we have (stock prices) and using it to project the values we don't have (expected dividends for an unknown future quarter), then using that projected data to predict the data we want (stock prices). We predict stock prices in May 2010 will range between 1148 and 1200 for the S&P 500. Sure, we know we're much more likely to have big errors between our forecast and reality, but how wrong can we be? As it happens, stock prices came in about 2.0% off the low end of our range, as the S&P 500 averaged 1125.06 during the month of May. If only this were one of our GDP forecasts, where having our target come within 2% of the finalized data qualifies as a correct call.... -1
21 May 2010 With over a third of the month's trading days left to go, but more importantly, before it happened, we observed that the future expectations for dividends per share appeared to be deteriorating, indicating that the average of the S&P in May would come in below the low end of our projected range. It's surprising how much timing matters where predictions are concerned. Here, what matters is that we posted our revised forecast *before* it came to pass. After all, if you don't have enough time to make decisions or take actions based on a forecast, what's the point? +1
1 June 2010 Still not having the data we need to accurately project where stock prices will go, we turn to our backwards math again, this time anticipating that the S&P 500 will average between 1007 and 1071 during the month of June 2010. Arggh! The S&P 500 averaged 1083.36 in June 2010, just over 1.1% above the high end of our predicted range. -1
14 June 2010 Another mid-month course correction! Only this time, because we suddenly could see farther into the future with just a portion of the data for which we'd been starving in the previous two months (it still doesn't look ahead far enough to be what we really need - the dividend futures data we have available to us only goes through mid-2011, while we believe the market is reacting to what it expects will be the situation in the second half or end of 2011!) But with that new dividend futures data that only goes through June 2011, we were able to nail down the average target range for the S&P 500 in June 2010 to fall between 1074 and 1118. Since we were watching a lot of the FIFA World Cup during June, let's just say "GOOOOOOOOOOOOOL!" +1
30 June 2010 We predict that the S&P 500 will range between 1014 and 1052 during July 2010. (Yes, still using our inverse method.) There's no question that prediction was wrong for the month of July 2010. Interestingly though, it was correct for the first 7 trading days of the month. More interestingly, our model correctly anticipated where stock prices would go next (1091-1122), but we assumed it wouldn't get there until we were well into the month of August. Our best guess is that the market is a bit ahead of where our model suggests it should be, but we'll have more on that soon. Along with a pretty scary prediction for where it might go next. -1
1 July 2010 We forecast real GDP for the second quarter of 2010 will be within 2% of \$13,374.8 billion constant 2005 U.S. dollars (which is how the BEA likes to report it these days!) The advance report for U.S. GDP in the second quarter of 2010 put the inflation-adjusted number at \$13,194.1 \$13,216.5 billion (constant 2005 U.S. dollars), which means our target value overestimated where GDP would be by 1.4% 1.2%. That's within our +/- 2% target range, so we score a hit! +1
14 July 2010 We make two predictions for the 2010 major league baseball season: The Arizona Diamondbacks will lose around 100 games and the New York Yankees will end the season with about 105 wins. We'll call these predictions correct if we're within +/- one game of the final standings. What can we say? We do some predictions just for fun! We'll find out in October.... +0

### Previously on Political Calculations

The following links will take you to our previous prediction outcome reports, which we've presented below in the order they've appeared here approximately every three months beginning with April 2009. You can get the most recent status updates by clicking the "track record" tag at the bottom of the post.

Labels: ,

is good for you

Welcome to the blogosphere's toolchest! Here, unlike other blogs dedicated to analyzing current events, we create easy-to-use, simple tools to do the math related to them so you can get in on the action too! If you would like to learn more about these tools, or if you would like to contribute ideas to develop for this blog, please e-mail us at:

ironman at politicalcalculations.com

Recent Posts

Applications

This year, we'll be experimenting with a number of apps to bring more of a current events focus to Political Calculations - we're test driving the app(s) below!

Most Popular Posts
Quick Index

Site Data

Visitors since December 6, 2004:

#### JavaScript

The tools on this site are built using JavaScript. If you would like to learn more, one of the best free resources on the web is available at W3Schools.com.