Google Search Terms: A Better Way to Predict Financial Vulnerability Indicators?

Many country-level macro financial indicators such as total bank credit to the private sector are produced with a lag. Sometimes a month, but often longer.

So you can try to predict the current period's credit change by estimating a statistical model that relies on the credit increase and the credit stock of previous periods. One such simple model for Australian monthly private credit is depicted by the red line in the graph below (the actual change is in blue). As you can see, the fit of the red line is "smooth" and not very impressive.

But Google recently launched a new tool that allows the user to track search volume data for certain categories of search queries on a weekly basis. Using this information, Google researchers have been able to predict flu epidemics in the United States 1-2 weeks ahead of the official numbers from the Center of Disease Control and Prevention.

Following this logic, when people search for "Banking and Personal Finance"-related queries they might consider a new loan for example. So can the related search volume indicator be used to help predict credit changes before the official numbers are released at the end of the month (or later)?

The answer seems to be affirmative. When the Google volume indicator for "Banking and Personal Finance" for the first week of each month is added to the model, the predictions improve substantially (the green line in the graph above). The indicator is statistically very significant and the mean absolute prediction error of the model is reduced by almost 15%!

I expect that my simple model can be much improved. So, similar to flu monitoring, could such models be used for early warning systems of financial vulnerability indicators?