Why is Google Trends data inconsistent?

Google Trends also allows the user to compare the relative search volume of searches between two or more terms.

Originally, Google neglected updating Google Trends on a regular basis. In March 2007, internet bloggers noticed that Google had not added new data since November 2006, and Trends was updated within a week. Google did not update Trends from March until July 30, and only after it was blogged about, again.[2] Google now claims to be "updating the information provided by Google Trends daily; Hot Trends is updated hourly."

On August 6, 2008, Google launched a free service called Insights for Search. Insights for Search is an extension of Google Trends and although the tool is meant for marketers, it can be utilized by any user. The tool allows for the tracking of various words and phrases that are typed into Google's search-box. The tracking device provided a more-indepth analysis of results. It also has the ability to categorize and organize the data, with special attention given to the breakdown of information by geographical areas.[3] In 2012, the Insights for Search has been merged into Google Trends with a new interface.

In 2009, Yossi Matias et al. published research on the predictability of search trends.[4] In a series of articles in The New York Times, Seth Stephens-Davidowitz used Google Trends to measure a variety of behaviors. For example, in June 2012, he argued that search volume for the word "nigger(s)" could be used to measure racism in different parts of the United States. Correlating this measure with Obama's vote share, he calculated that Obama lost about 4 percentage points due to racial animus in the 2008 presidential election.[5] He also used Google data, along with other sources, to estimate the size of the gay population. This article noted that the most popular search beginning "is my husband" is "is my husband gay?"[6] In addition, he found that American parents were more likely to search "is my son gifted?" than "is my daughter gifted?" But they were more likely to search "is my daughter overweight?" than "is my son overweight?"[7] He also examined cultural differences in attitudes around pregnancy.[8]

Google Trends has also been used to forecast economic indicators[9] and financial markets,[10] and analysis of Google Trends data has detected regional flu outbreaks before conventional monitoring systems.[11] Google Trends is increasingly used in ecological and conservation studies, with the number of research articles growing over 50% per year.[12] Google Trends data has been used to examine trends in public interest and awareness on biodiversity and conservation issues,[13][14][15][16][17] species bias in conservation project,[18] and identify cultural aspects of environmental issues.[19] The data obtained from Google Trends has also been used to track changes in the timing biological processes as well as the geographic patterns of biological invasion.[20] Google Trends has been shown to outperform survey‐based indicators.[21]

Evidence is provided by Jeremy Ginsberg et al. that Google Trends data can be used to track influenza-like illness in a population.[22] Because the relative frequency of certain queries is highly correlated with the percentage of physician visits in which a patient presents with influenza-like symptoms, an estimate of weekly influenza activity can be reported. A more sophisticated model for inferring influenza rates from Google Trends, capable of overcoming the mistakes of its predecessors has been proposed by Lampos et al.[23]

The use of Google Trends to study a wide range of medical topics is becoming more widespread. Studies have been performed examining such diverse topics as use of tobacco substitutes,[24] cancer,[25] suicide occurrence,[26] asthma,[27] oral health-seeking behaviors,[28] and parasitic diseases.[29] Further research should extend the utility of Google Trends in healthcare.

Furthermore, it was shown by Tobias Preis et al. that there is a correlation between Google Trends data of company names and transaction volumes of the corresponding stocks on a weekly time scale.[32][33]

In April 2012, Tobias Preis, Helen Susannah Moat, H. Eugene Stanley and Steven R. Bishop used Google Trends data to demonstrate that Internet users from countries with a higher per capita gross domestic product (GDP) are more likely to search for information about the future than information about the past. The findings, published in the journal Scientific Reports, suggest there may be a link between online behaviour and real-world economic indicators.[34][35][36] The authors of the study examined Google search queries made by Internet users in 45 different countries in 2010 and calculated the ratio of the volume of searches for the coming year (2011) to the volume of searches for the previous year (2009), which they call the future orientation index. They compared the future orientation index to the per capita GDP of each country and found a strong tendency for countries in which Google users enquire more about the future to exhibit a higher GDP. The results hint that there may potentially be a relationship between the economic success of a country and the information-seeking behaviour of its citizens online.

In April 2013, Tobias Preis and his colleagues Helen Susannah Moat and H. Eugene Stanley introduced a method to identify online precursors for stock market moves, using trading strategies based on search volume data provided by Google Trends.[37] Their analysis of Google search volume for 98 terms of varying financial relevance, published in Scientific Reports,[38] suggests that increases in search volume for financially relevant search terms tend to precede large losses in financial markets.[39][40][41][42][43][44][45][46]

The analysis of Tobias Preis was later found to be misleading and the results are most likely to be overfitted.[47] The group of Damien Challet tested the same methodology with unrelated to financial markets search words, such as terms for diseases, car brands or computer games. They have found that all these classes provide equally good "predictability" of the financial markets as the original set. For example, the search terms like "bone cancer", "Shelby GT 500" (car brand), "Moon Patrol" (computer game) provide even better performance as those selected in original work.[38]

In 2019, Tom Cochran, from public relations firm 720 Strategies, conducted a study comparing Google Trends to political polling.[48] The study was in response to Pete Buttigieg's surge in a poll of Iowa's likely Democratic caucusgoers conducted between November 8 to 13 by the Des Moines Register. Using Google Trends, he looked into the relationship between polling numbers and Google searches. His findings concluded that, while polling consists of far smaller sample sizes, the primary difference with Google Trends is that it only demonstrates intent to seek information. Google search volume was higher for candidates having higher polling numbers, but the correlation did not mean increased candidate favorability.[49]

Research also shows that Google Trends can be used to forecast stock returns and volatility over a short horizon.[50] Other research has shown that Google Trends has strong predictive power for macroeconomic series. For example, a paper published in 2020 shows that a large panel of Google Trends predictors can forecast employment growth in the United States at both the national and state level with a relatively high degree of accuracy even a year in advance.[51]

In 2020, it was shown that Google Trends data can exhibit a high variability when queried at different points in time, indicating that it may not be reliable except for very frequent search terms due to sampling,[52] and relying on this data for prediction is risky. This research made it to major headlines in Germany.[53]