Markets and Pundits Have a Data-Point Fixation
Data is the raw material we use as the basis for analysis.
Investors demand it. Baseball fans love it. Quants live for it. Pollsters depend on it. Data is the difference between anecdote and evidence, between opinion and facts, between life and death (ask a surgeon or airline pilot).
Data drives the economic world around us. It is how we understand complex, abstract things. You look at it to see how well your portfolio is doing (pretty good!), or to understand the state of the economy (better than expected!). Businesses use data to understand where they are growing. Investors use it to figure out if something is worth owning. It is how we judge athletic performance, measure academic achievement, evaluate companies, compare technologies and evaluate scientific progress.
Yet people often seem to overlook the weaknesses in data. I see this often in the financial community and the use of data series, whether it’s the jobs numbers, housing prices or the consumer price index.
The key here is the word “series” and how we perceive time. People tend to experience time as here and now, rather than as a continuum. The future is some distant, far-off event; the past is ancient history. Neither perspective is the correct way to give context to data in a time series.
This is especially true for traders, who experience time in a tick-by-tick change in their profit and loss position. Why long-term economic data is so important to them is in the short-term shifts in volatility, not the actual data. This is a revealing dichotomy.
Consider how often the average trader, economist, investor or analyst tends to see any data as a single event. They forget the series, the continuum mentioned above. Economic releases, earnings, valuations are best thought of as a film, not a photograph.
Our language reveals that we don't usually think about it that way. We refer to a "data point,” we think about an “economic snapshot,” we get an earnings “report.” The better way to understand the ebb and flow of the data is to recognize these are released as part of an ongoing stream.
And yet we succumb to what I call the “recency effect,” or the tendency to overemphasize the most recent information and underemphasize the longer-term series.
Consider the payroll report. It is a very noisy bit of near real-time analysis, filled with normal inaccuracies and a huge margin of error. It is subject to large revisions. It is even more challenging in a post-credit crisis recovery. Any one report can be an outlier, and if you ignore the context of the series, you run the risk of overemphasizing statistical noise.
Recall the May jobs report, a negative surprise that led to handwringing, predictions of an imminent recession and many forecasts that the Federal Reserve would keep rates lower for longer. All of which was promptly forgotten when the next month was a positive surprise. (Don’t say you weren’t warned).
I have noticed the same sort of approach to the presidential elections. The daily news cycle requires a narrative, and each and every poll delivers one, regardless of its accuracy. Back when Republican Donald Trump was trailing by a significant margin in the polls, pundits were calling his campaign a “dumpster fire,” he was heading for a landslide loss, the Democrats were going to retake the Senate AND the House. This story line was confirmed by a terrible August for him.
Then the race began tightening in September amid a series of missteps by Democrat Hillary Clinton. This terrified establishment Republicans and the #NeverTrump wing of the party, and reignited speculation on the fringes that Democrats were about to call for Vice President Joe Biden to step in as Clinton's replacement.
The data series paints a very different picture: In a nation with an electoral college, the national polls matter less than the aggregation of state polls and how they add up to the 270 electoral college votes needed to win the presidential race. The Democratic candidate starts out with an advantage because of geography and demographics, while the Republican candidate has tapped into anger and unhappiness.
We still haven’t had any debates between the candidates yet. In a politically divided nation the election can easily go one way or another. Yet the media, and partisans on both sides of the aisle, alternatively gloat and panic over data noise. Is the emphasis on the latest outlier poll any different than the market’s monthly spasm in reaction to the jobs report? It is the same cognitive errors that drive both.
For a baseball hitter in a slump, the longer-term data series is more meaningful that any single at-bat. The same is true in politics, economic data, finance and markets. Investors too should pay more attention to the movie, and not any one snapshot.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Barry Ritholtz at email@example.com
To contact the editor responsible for this story:
James Greiff at firstname.lastname@example.org