### When Consumer Confidence Drops, The S&P 500 Rises

Published February 24th, 2010 in SentimentThe sharp drop in the February Conference Board’s Consumer Confidence numbers surprised many. After climbing for three consecutive months and reaching a 16 month high it dropped 10.5 points (from 56.5 to 46). The “Present Situations Index” fell to 19.4 - the lowest since 1983. The US consumer is hurting real bad. In case the sky high unemployment, food stamp figures and credit contraction statistics didn’t give it away, the consumer confidence numbers reflect a totally despondent consumer.

Naturally, a lot of pessimists are talking about how this is presaging a “double dip” and there is much wailing and gnashing of teeth. I prefer the contrarian approach so I decided to take a look at the actual numbers and see the relationship between consumer confidence and the stock market. My hunch is that if there is a relationship, it is an inverse one.

Since this is a hot topic, Barry quoted Bianco Research’s study of the effects of a drop of 10% or more in Consumer Confidence on the S&P 500 index.

While Bianco’s approach is interesting, I found their data and presentation confusing. For example, I’m not sure what “That Month” and “Next Month” are referring to exactly. As well, I’m assuming that for April 30th 1980 they missed the negative sign in front of the 20.30 drop.

I decided to keep my own study really simple. For each instance of a large monthly drop in Consumer Confidence, (starting from the very next day in the market) I looked at what would happen if you bought the S&P 500 index and held for 1, 3, 6 and 12 months. Here are the results:

On average the S&P 500 performs well for all time frames. The “sweet spot” seems to be between 3-6 months from the drop in sentiment with a return of 5.69-9.08% (+20% annualized). Not surprisingly, the worst instances of 12 month returns occur during the recent bear market. As does the best return when the market bounced back strongly. If we follow the script provided by history, then August 2010 will see the S&P 500 at 1205. If.

For my 3 long term readers, this is old hat. We’ve gone over the contrarian nature of consumer confidence more than a few times already. But it is nice to put some numbers on the board nonetheless.

**Update:**

After a discussion (see below in the comments section) it became clear a consensus was developing on looking at this data from a relative, rather than nominal lense, as Bianco Research had. To be clear, instead of looking at 10 percentage point drops, we should instead look at a percentage drop. So for example while the recent drop from 56.5% to 46% represents a 10.5% point drop, it is an 18.5% decrease.

Being statistically challenged, I turned to Jason Goepfert of SentimenTrader.com and asked for his assistance. Here’s his response:

I show that with any monthly drop larger than 15%, six months later the S&P 500 was positive 14/16 times with an average of +13.3%. The two negative ones were in 1970 and 1973, since then then last 14 signals have been positive. One year, later, it was up 15/16 times with an average of +18.2%.

Barry Ritholtz was curious to know how these returns compared to a random return over the same time period lookup. While Jason didn’t run such a robust study, according to him, looking at the number of times we had a positive return minimizes the chance that it is due to just chance.

The “sweet spot” I found at the half-year point continues with an average return of +13.3% - this is significantly higher than the non-relative study that I did for the same period. So there you have it. A definitive edge from looking at drops in Consumer Confidence, going above and beyond what Bianco Research submitted to their paying clients.

**Enjoyed this? Don't miss the next one, grab the feed or **

useful stats thx

just realized something, instead of looking at point drops it would be more meaningful to look at normalized drops (%)… oh well, that’s what happens when you springboard off someone else’s idea, I suppose.

I couldnt agree with you more Babak…you MUST normalize the data in order for it to have any real context. The guys at QUANTDNA weighed in with their own take except that looked at over 20 markets and they normalized the data by focusing on “prior instances when it declined by at least 15%” While their analysis only looks at the markets performance after 30 days, they nonetheless indicate that S&P 500 is down 80% of the time 10 days later by an average of 3-4%. I wish QUANTDNA would look at stuff beyond 30-days. I wonder if they do that for paying clients. Anyway, here is the link

Thanks Irwin. As you’ll notice when you click on the link, you have to register. For those too lazy, the results for the S&P500 after 11 days is -3.9 (lowest) and after 29 days 2.2 (highest). The lookout is only for 30 days, as Irwin mentioned.

I personally think that point moves are fine as long as you are working with percentages. It takes just as many people to change their mind on a move from 70 to 60 as it does from 40-30. But it might be interesting to see the results broken down as 10% drops when the number is above/below X or at the levels we currently are.

wayne, yeah, you’re probably right. I’m looking into it and if I come up with anything I’ll share the results.

I’ve updated with some interesting relative data (thanks to Jason Goepfert). Take a look (up there ^^)

Now your cooking with gas.

I’m a bit jealous.

After looking into how the index is calculated, I think you do probably need to normalize it as you did. I was previously thinking it was constructed similar to a sentiment index, bounded by x and y.

At some point in the future, Im going to want to get my hands on your database.

I’m trying to get a handle on the implications of your study. Hopefully, I can study it some more this weekend.

wayne, I don’t have the series but instead relied on the dates given by Bianco. But I’m pretty sure you can get it by contacting the Conference Board. The trick is to realize if this is data mining or if it really represents an edge.

i would just like to point out that if you look at the 6 month data on your last data point the snp is down NOT up.