Speaking of $100 bills on the pavement, I haven’t looked into this – but look forward to doing so at some stage. Given the preponderance of IT systems which generate real time data for their organisations – firms and agencies – why aren’t we trying to do more of this with our national statistical collections, particularly those that we rely heavily on to manage the economy and others systems in need of ‘real time’ management. If it’s too hard and expensive to peek into every business’s accounting system, why not try to peek into a few – with assurances that the information would be anonymised or otherwise protected from the tax office finding irregularities between firms reporting in real time and to itself. If necessary one could pay a representative sample of firms to participate – they could press a button and send in data every day or week or month – or one could see if one could get buy-in by offering participating firms advance access to the collective information good generated from their data.
In any event it seems odd that so little is being done in this area. But then perhaps it is and I just don’t know about it. I was prompted to write this by this article (pdf) which shows that using Google searches can help in forecasting economic developments.
From the Abstract.
Most economic variables are released with a lag, making it difficult for policy-makers to make an accurate assessment of current conditions. This paper explores whether observing Internet browsing habits can inform practitioners about real-time aggregate consumer behavior in an emerging market. Using data on Google search queries, we introduce a simple index of interest in automobile purchases in Chile and test whether it improves the fit and efficiency of nowcasting models for automobile sales. We also examine to what extent our index helps us identify turning points in sales data. Despite relatively low rates of Internet usage among the population, we find that models incorporating our Google Trends Automotive Index outperform benchmark specifications in both in-sample and outof- sample nowcasts while providing substantial gains in information delivery times.
And from the conclusion.
Our results show that models incorporating Google search resultsoutperform competing benchmark speci?cations in both in and out-of-sample nowcastingexercises. The Google data have a number of characteristics that should make them particularly attractive to decision-makers in emerging markets: (i) They are derived directly from micro user data; (ii) They contain information on a large proportion of Internet users, whichis a far more extensive sample than is commonly employed by surveying agencies; and (iii)They are released at high frequency and at regular intervals. Our ?nding that the accuracyof nowcasting models for automobile sales in an emerging market can be improved usingcontemporaneous search patterns suggests that Google data is a promising source of infor-mation for nowcasting components of aggregate demand in short-run models, an exercisewhich we leave to future research.