(A) The number of births for 1,000 people in each US state. Birth rate is defined as the number of births for 1,000 people. (B) We use Google Correlate to find terms for which the number of searches is higher in U.S. states with higher birth rates. Similarly, we identify terms for which the number of searches is higher in states with lower birth rates. Here, we list the 31 terms which showed the strongest positive correlation (left) and negative correlation (right) with state wide birth rate. To determine the significance of these correlations, we generate 1,000 random samples from a multivariate Gaussian distribution where states which are closer together tend to have a similar value. We submit these samples to Google Correlate and build a distribution of correlation coefficients for each of the 31 top most search terms. We depict the strength of correlation required for the correlation to be significant at the p < 0.05 and p < 0.01 level, given this null hypothesis distribution. (C) To allow us to generalise beyond individual search terms, we conduct an online survey asking participants to identify the main topic in each list of 31 terms. Here, we depict all survey responses which account for more than 5% of submitted responses. Our results suggest that users in states with higher birth rates search for more information about pregnancy, while those in states with lower birth rates search for more information about cats (“baby car seat”, p = 0.051, all remaining ps <0.05).

(A) Infant mortality rates for each state in the US. An infant is defined as any person one year old or younger. Infant mortality rate is defined as the number of infant deaths per 1,000 births. (B) In a similar fashion to our investigation of birth rates (), we use Google Correlate to find terms for which the number of searches is higher in U.S. states with higher infant mortality rates, and with lower infant mortality rates. We list the 31 terms for which differences in search volume across U.S. states shows the strongest positive correlation (left) and negative correlation (right) with state wide infant mortality rate. Again, we generate 1,000 random samples from a multivariate Gaussian distribution where states which are closer together tend to have a similar value. We submit these samples to Google Correlate and build a distribution of correlation coefficients for each of the 31 top most search terms. We depict the strength of correlation required for the correlation to be significant at the p < 0.05 and p < 0.01 level, given this null hypothesis distribution. (C) Again, we ask Amazon Mechanical Turk users to identify the most prominent topic in each of these lists of terms. We depict all survey responses which account for more than 5% of submitted responses, along with the percentage and number of respondents who gave each response. Our results suggest that users in states with higher infant mortality rates search for more information about credit and loans, as well as sexually transmitted diseases (all search terms p < 0.05).