Google Trends and Fun Facts About The Turkish Economy-2
Here is the continuation of Ali G.’s previous guest post, with my comments right below the post:
In the last post, I briefly talked about the value of data, how hard it is to obtain reliable data and how technology provides new sources of data. This week I will focus on a particular source, Google and its Google Trends and Google Insights for Search services. Then, I will show some interesting results from Turkey. This might be a little old for some people but I think it is still underappreciated.
Google Trends is the older brother of the search volume providers. It basically shows the search intensity of a particular item through time (starting from 2004). You can apply a region filter and top news results are also displayed. Then in 2008, Google launched Insights for Search which is a fancier service which essentially does the same thing. So what makes this data so special? Well for one, although it is still sampled, there is a lot of data! Moreover, people provide this data with their free will, without any external factors coming into play. So it may be especially useful for sentiment studies.
First, let us look at who makes use of these services. One example is the Bank of England (1). According to the Guardian, the bank admits, in an article for a Quarterly Bulletin, that “As further developments are made in this area, and the backrun of the data increases, these data are likely to become an increasingly useful source of information about economic behavior.” There are several examples as to how the Bank uses these services and the strongest one is for the housing market. Apparently the search results for “real estate” gave better results than housing surveys regarding average house prices over eight years. Searches also gave hints on the state of the job market and the search volume, for example search terms “jobseeker’s allowance” and “unemployment” rose through the recession. Again Google results were more accurate than surveys, but official data (lagged one month) was more accurate than both.
Let’s move on to Turkey. Although it is possible to construct alternative indices and models with proper statistical methods(refer to the paper in the last post), I won’t go into that here and I’ll just give a few fun results. I looked at credit growth, consumer sentiment, real estate and a few financial data. I’ll just present the two most striking examples, starting with real estate results. First let’s look at the growth in Turkey under the “Real Estate” category of Google Insights for Search as a whole, starting from 2004 in the figure below
The growth since 2004 is amazing but it seems to have stabilized since the last two years. To take the analysis one step further and view the demand in houses for sale and houses for rent, I searched for those terms separately. First, there is cyclicality for apartments for rent, which is expected. What is interesting though is, prior to 2008, search volume for apartments for rent and apartments for sale moved hand in hand, with search for rentals generally staying on top. However, that picture changed after the second half of 2009 and the gap that opened between the two during that period is still present. These may point at the fact that the “buying frenzy” in the real estate market that has been going on for a few years is still on, but it has somewhat slowed in the recent period.
Another very interesting result is presented below. The graph shows the search volume for “Dolar” (Turkish for the USD/TRL foreign exchange pair) versus the implied 1 month at the money volatility for the USD/TRL pair (which could roughly be translated as an indicator for Turkish Lira risk as priced by the market). The correlation between the two series is amazing. I don’t know if the naïve Google data has any predictive power, but at any rate it seems to be a very easy to obtain, timely, free and reliable risk sentiment indicator for the Lira. Just remarkable.
Data often provides the leverage in any type of empirical work, be it a journal paper or a just simple analysis. However, especially for some questions, data obtained from standard sources are either lagging, expensive or subject to biases such as in survey data. Although they have their drawbacks like not having a long history, Google’s search volume-based insight services provide a very strong and cheap alternative. This is more valuable especially for emerging markets, where quality data is even harder to come by. I will try to do some more serious stuff with it if I have the time and I’ll share it as I go along, but also try them out for yourself, I guarantee that you’re going to be amazed at the results.
Ali Gökhan is the acting economist at a Turkish conglomerate. The views expressed here are his personal views only and do not represent the views of his company.
If you haven’t see it already, have a look at my latest column, on the Turkish real estate market- Ali G.’s post (at least the first part) is a nice complement to my piece. However, I am not sure if his Google data support my arguments- I couldn’t be sure from his graphs.
As a more general point, using Google to get info. about the economy has been popularized in the last few years. For example, Central Bank economists just published a paper, where they “nowcast” the non-agricultural rate by using Google search queries.
3 Responses to “Google Trends and Fun Facts About The Turkish Economy-2”
Hi Ali – thanks for the guest posts… I don't understand the "dolar" graph (the last one you discuss) at all. Maybe it's because it's in tiny font… But which series is the "dolar" search query and which is the implied 1 month volatility? How do you see the correlation at all? All I see is a bunch of lines that graph USD against different currencies…
Hi; I relayed your question to Ali. My take is that the wrong graph was posted, but I'll post here as soon as I get an answer from him…
As I guessed…. The wrong graph- he sent me the right one…