Every week/month/quarter, a number of closely-watched economic indicators are reported that have the potential to move markets. For example, the monthly jobs report is hotly anticipated, while same-store sales numbers for individual retailers can move the stocks of these companies significantly. Gathering, formatting and then reporting this data takes time, however, resulting in lags between when the events are occurring and when they are reported to us. New technology, however, may be rendering these reports obsolete, by allowing us to access closely correlated proxies to these and other data in real-time.
Google Correlate allows users to submit a time series (e.g. monthly new house sales in the US over the last several years) and then spits out a search term that correlates best with that data (e.g. "real estate agencies"). Google Trends can then be used on the relevant search term (e.g. "real estate agencies") to predict what this month's new house sales are looking like, well before the official data is released following the end of the month.
Wharton economist Justin Wolfers takes us through an example. He uploaded weekly unemployment claims to Google Correlate, which offered him a search phrase which best fit this data: "filing for unemployment". The correlation was 0.91 (a correlation of 1 means perfect correlation, while a correlation of 0 means no correlation), suggesting that we can guess how many people are filing for unemployment based on aggregate data of how many people are searching using the term "filing for unemployment"! Here's how the data track over time.
In many cases, the economic reports we (eventually) receive are derived from survey and not actual data. As such, it is reasonable to believe that there are some instances where aggregated search (or other real-time) data may be a more accurate measure of the economic event that is being measured! If this is true, many of the economic reports we consider dear today may be obsolete relatively soon.
For more examples of this idea in action, see this paper from Choi and Varian.