Jobless Claims Prediction Hits the Mark

Earlier this week, following a Google Insights for Search snafu, Markets with Search predicted that jobless claims would come in flat for the week (in line with consensus). The numbers released today validated the claim. Chalk one up to another win for using search to "predict the present."

Looking further into the issue which resulted in the errant prediction earlier, it is to my understanding that Insights for Search data may take up to 2 days to finalize following the end of a periodic week. The initial prediction was made on data that has not been fully normalized. While this is not ideal, it still enables you to be ahead of data that is delayed by 5 days.

Correction: Jobless Claims Likely to Be In-Line with Consensus this week

There appears to have been a glitch in the data analyzed for yesterday's post. Insights for Search was listing an 18% lift in popularity for key searches in yesterday's evaluation; today's data shows a different story (highlighted below).

The reason for the discrepancy is currently under investigation and I apologize for the confusion.

The current data reflects no major changes in the key searches, which indicates a likely flat week for initial claims, in line with consensus expectations.

Initial Jobless Claims Likely to Miss Consensus this Week

This upcoming Thursday, September 3rd, the initial jobless claims report for the week ending 8/28 will be released. Initial claims have been trending downward over the past weeks, which is a positive sign for the economy, as job retention has been looking fairly stable.

The consensus pick for this week's data is a continuing trend downward, with an expectation around a 1.5% drop. Looking at the search data, however, indicates that there was a sharp increase in searches for key unemployment-driven queries.

On July 16th, we posted the correlation between key searches for unemployment and jobless claims. While it is unlikely that this week's jobless claims release will increase by 18% (as searches have week over week), it is highly likely that the concensus pick will miss the mark in the wrong direction.

[PLEASE NOTE THIS POST HAS BEEN UPDATED TO REFLECT NEW, CONTRARY DATA]

Bank of Israel Using Search Data to Get a Pulse on Economic Activity

The Bank of Israel recently posted a brief analysis highlighting how they have looked at Google Insights for Search to predict economic activity. I highly encourage reading through the white paper linked from the press release; it truly shows the power and depth you can get into when analyzing search data.

A very interesting result from this analysis shows some good news for the Israeli economy:

In the last few months the query indices in Israel point to a reduction in the probability of a slowdown; that conclusion is based on the moderation of the decline in private consumption. The indices do, however, indicate a steep increase in unemployment in the second quarter of the year, to more than 8 percent.

It is reassuring to see these kinds of analyses pop up outside of Markets with Search. Thanks for the tip on this press release, Hal.

I will try to repost these articles as I come across them to point to ideas outside of MWS. Have a great weekend and happy trading.

Getting Ahead of Initial Jobless Claims

Have you ever wanted to get ahead of data before the official release to the street? Have a play based on the numbers?

The time has finally come. After doing some backtesting, some content has been geared up and ready for release. Today’s trend: using Google to predict where seasonally adjusted job claims will turn… with 88% confidence.

Consumers turn to the search engine juggernaut to indicate interest in topics that have been on their minds, making it the world’s largest focus group. One of the most interested topics of 2009: What do I do if I just lost my job?

Answer: search for filing unemployment. With an R² of 0.8815, searches for “file unemployment” and “filing unemployment” fall right in line with actual seasonally adjusted Initial Jobless claims for a given week.

It is important to note that the indexing algorithm for Insights for Search readjusts with every new peak in popularity, requiring you to index SA Initial Claims accordingly using the methodology (SA Initial Claims_n / Max SA Initial Claims * 100). Given the values from July 16, 2009 the following model will allow you to fairly accurately predict the movement in jobless claims:

Predicted SA Initial Claims = 5006.2 * (Search Index for Unemployment Filing) + 225,046