Lack of Confidence in Consumer Confidence

Posted by Milos Sugovic

U.S. consumer sentiment rose in early September, according to the Reuters/University of Michigan Survey, with index values increasing from 65.7 in August to 70.2 in September. Hallelujah! Consumers are finally optimistic, they’ll start spending, and the economy will recover in a jiffy. Oh wait, there’s one caveat: the change in index values is statistically insignificant and therefore useless. 

Here’s why: it’s called “margin of error.” 

If you haven’t heard of it, here’s a good place to start. The Reuters/University of Michigan Survey is based on a relatively small sample size of roughly 500 respondents. So if the sentiment of these 500 respondents is projected onto the U.S. population, that gives us a margin of error of +/- 3.3 index points. 

That means in August, we were 95% confident that the actual or “true” population index value was between 62.4 and 69. In September, the confidence interval was between 66.9 and 73.5. Note the overlap.  

So, I pose to you the following question: If the index value in August could be as high as 69, and as low as 66.9 in September, are we confident that it actually increased? No! As long as the confidence intervals, or margins of error - depending on your perspective - overlap, the change is statistically insignificant. And since statistical significance is a precondition for economic significance, any hype and optimism around the reported index values is unfounded. 

So before you let “news” like this affect your business plans, read the fine print. Statisticians know to look past the estimated values - confidence intervals are the end all be all. After all, do the guys at University of Michigan really know that consumer confidence is EXACTLY at 70.2, down to the first decimal place? Their best guess is a range, that’s it. So, whenever you see a figure reported, think margin of error, and you should be confident in your conclusions.