Simple Search Terms Can Predict A Recession

Google Trends is a public web tool that enables users to study hidden relationships based on a selected subset of massive search data collected from Google Search. A breakout search term from Google’s user groups may indicate significant business trends that can be exploited to gain a better understanding of consumer behaviors, for marketing purposes, or  for obtaining a more thorough understanding of a macroeconomic environment.

Methods

In this study, I attempted to find simple search terms from query data that correlated well with a recession pattern in the U.S economy. I entered “foreclosure,” “recession,” “layoffs,” and “bankruptcy” as search terms, using data from 2004 to 2013 and limited to within the U.S. I assumed that when many people had an interest in searching for those four ominous terms, that those people felt pessimistic about the state of the economy. They would consequently curb their spending, leading to further deterioration of the economy. Let’s see if that thesis coincides with the reality that the worst U.S. recession since World War II started in January of 2008 and lasted through March of 2009, as agreed upon by most respected economists.

Results

The graph illustrates Google users’ search interest from 2004 to the present for the four keyword searches mentioned above. The highest search volume of data was denoted as 100 on the Y-axis, with all other data normalized to that scale.


For the green “bankruptcy” curve, I ignored the data pre-2006 because Congress had passed the new bankruptcy law in 2005, which caused an abnormal jump in search interest for “bankruptcy”. As seen in the graph above, when search volume increased (as indicated by an arrow), the change in the trend of that search interest preceded the likely onset of an economic recession. Clearly, the breakout in searches for “foreclosure” was a glaring warning, because the numbers forecasted the onset of a recession in March of 2007 (as indicated by the graph’s blue arrow), nine months ahead of the actual occurrence of the event. The term “recession” signaled the strong likelihood of a downswing in economic conditions in September of 2007 (as indicated by the blue arrow, three months ahead of the recession), while “layoff” (as indicated by a yellow arrow in March of 2008) and “bankruptcy” (as indicated by a green arrow) signaled additionally strong indications of a recession in approximately April of 2008.

 

By analyzing the data trends, I have been better able to understand how the recession occurred. First, people on subprime mortgages were unable to pay their mortgages and lost their homes to foreclosure, leading to the disruption of business. Anxious users began to hysterically search Google for information about a “recession,” while companies reacted to bad business by reducing headcount and laying off large numbers of employees. The unemployed filed for bankruptcy with their houses foreclosed, thus repeating the cycle.

Based on this study, the Federal Reserve Bank should have acted hastily to lower interest rates in March of 2007. Instead, it hesitated until August of 2007 to announce the first rate cut. By then, the damage to the economy had already been sustained, which was struck with a deep recession and total job losses exceeding 9 million between 2008 and 2009.

From this study, it can be seen that a clever algorithm can turn seemingly worthless, raw search data into a tool for obtaining valuable insight or treasured applications. In this case, one of the worst recessions sustained by the world’s highest ranked economy can be interpreted and deciphered by the behavior of Internet search users.