Do Lost Pet Ads Predict Earthquakes?

This is a neat little thing you can do for a science fair project. A lot of people say that animals run away before earthquakes and some say you can predict earthquakes that way. One way to count lost pets is to count the number of lost pet ads in a newspaper. That's what we did to see if we could predict earthquakes. You can try it yourself, but first let's see how Andy and I did it.

First, we collected ten dates between 1983 and 1993 where an earthquake occurred with magnitude greater than 4.5 and within 50 km of San Jose. We picked earthquakes near San Jose because we were going to use the San Jose Mercury News to get the lost pet ads. We picked magnitude 4.5 and 50 km so that we got 10 earthquakes in our data set.

Then we chose ten random dates to compare to the earthquake dates. The random dates are our control study. We picked these dates by cutting out pieces of cardboard and writing dates, months, and years on them, putting them in separate jars (one for the dates, one for the months and one for the years) and "picking dates out of a hat", so to speak. Using cardboard is a good idea, because first we tried paper but it was too hard to mix up in the jars.

 

A picture of the cardboard squares that we used - the m's next to the numbers show that they are months because the numbers 1 through 12 are also dates. The numbers 1 to 31 are the dates and the numbers 1983 to 1993 are the years.

We wanted to make sure that the random dates were not just before an earthquake because then we might find a lot of pets running away due to the earthquake. We also wanted to make sure that the random dates weren't just after an earthquake because the earthquake and its aftershocks might make pets run away. So, we picked windows around each earthquake that weren't allowed for random dates.

We then went to the Cecil Green Library at Stanford University where they have the San Jose Mercury News on microfilm. There we counted the number of lost pets on all 20 dates: the ten days just before the earthquake and the ten random days. For each day we counted the number of lost cats, dogs, and other pets separately. We also added together all three kinds.

To study whether there were more lost pet ads before earthquakes we made a chart of the data. We then put the chart in different orders such as by the total number of lost cats, lost dogs, other lost pets, and the total number of lost pets. From these charts we learned that you can't predict earthquakes by counting the number of lost pet ads. We saw that this was true because when we ordered the charts by the number of lost pet ads the earthquake and random days were mixed up. In fact, the day with the highest number of total lost pet ads for this experiment was a random day! Look at the charts to see this.

So the next time you hear someone talking about predicting earthquakes with their pets, tell them about this! If you want to try it yourself here are some helpful hints.

If you have any comments please send them to Andy Michael.