EWC stands for Expropriation Without Compensation. This is the practice that allows a government to take ownership of an individual’s private property without having to pay any compensation. Currently, a constitutional amendment allowing for EWC is making its way through South Africa’s Parliament. The policy change is being justified as a policy of land reform that would correct for the land equity issues that stem from Apartheid.
According to the November 2017 Land Audit Report published by the Department of Rural Development and Land Reform, white individuals own a 72% of the hectarage of individually owned farms and agricultural holdings (Figure 1).
So is EWC a good thing? Though it may be well intentioned, EWC carries the potential to bring disaster to South Africa. Historically, EWC has lead to economic hardship. Zimbabwe passed land reform policies allowing for EWC in 2000 and has since faced many negative economic repercussions. These negative economic impacts are largely attached to investor confidence. EWC may lead investors, both domestic and foreign, to doubt the stability of their property rights and thus they may choose to leave. Investor flight, is only one of the negative economic impacts. The economy is also intertwined with global politics. Alliances may shift due to this critical change in policy. Due to the frightening precedent of Zimbabwe, there is a real concern that EWC could bring to economic and political ruin to South Africa. However, despite the risks attached to the proposed policy, there still is support for it. But what causes this support? What characteristics are attributed to those who favor EWC? By using Google Trends data, we can attempt to construct stories of the factors which may be contributing to support for EWC.
What are Google Trends?
Before delving into the particular mechanisms of the proposed research design, it is important to understand what Google Trends is and what data it offers. Google Trends is a free and publicly available Google service which provides data on the search volume of different queries across time and space. Users choose particular terms, words or phrases and are allowed to specify time periods and regions of interest and Google Trends will produce a graph of the relative search term frequency within the specified parameters. Google Trends sometimes also offers data on further regional levels. For example, if one were to research how often people searched Elon Musk in the United States over the past three months, Trends also provides information on which states had the highest search volume in comparison to others. This search volume is measured in relative terms with 100 signifying the greatest term search frequency. This means that the peak of the graph will always be 100 which signifies the point at which the particular search term was most searched.
This is best illustrated through an example. With the search term “Matthew McConaughey” (specified as the American Actor), the following trend line (Figure 2) was produced for the United States in 2018.
There are several spikes in the line which signify periods during which the search frequency for the term was higher than average. The spike in early March was given the value of 90, the spike in mid September peaked at 80, and the last spike in late December was 100. The 100 value indicates that the term popularity was highest in that period. The other values are all relative to this peak popularity. For example, the value during early June was 42, which means the search term popularity was 42% of what it was during its highest peak. The spike in early March with a value of 90 indicates that the term popularity during that period was 90% of peak popularity.
Not only does Google Trends tell us when a particular term was searched the most, but it even offers some information on where it was most popular. These values are also calculated on a scale from 0 to 100, where 100 is given to the location where the term is the most popular. This value is calculated as the proportion of total searches at the particular location that are of the search term of interest. It is important to understand that the term popularity is relative to the total number of searches at a given location.
For example, suppose there are two regions, Location A and Location B. Suppose Location A produces four times as many searches as Location B does. Let’s say that 10% of Location A’s searches are of the term of interest compared to the 20% of Location B’s. Even though Location A yields the higher quantity of term searches of the two, Location B has the higher proportion of term searches. Google Trends would assign the value of 100 to Location B and 50 to Location B.
Going back to the Matthew McConaughey example, when analyzing the term popularity, Texas has a value of 100 indicating that the term is most popular here. Alaska is ranked 30th and has a value of 62. The District of Colombia is ranked last with a value of 46.
Google Trends also offers options to specify the type of term. The results of the Matthew McConaughey trend data specified Matthew McConaughey as the actor. Similarly, instead of narrowing down the term, it can be generalized by using search topic. The search topic is somewhat of a blackbox because the terms attributed to the topic are not published.
Now that the mechanics behind Google Trends have been explained, the research design can be understood.
What can Google Trends Tell Us about EWC?
For the study, we will be analyzing the search volume of the search topic “Expropriation” in South Africa from 01-01-2017 to 12–31-2018. Over this two year period, Google Trends provides the search popularity data at the provincial level. This study will exploit this regional variation in search popularity in order to correlate it with and regress it on various regional characteristics.
By correlating and regressing the relative measures of search frequency on regional characteristics, we can create a model that predicts how much a region searches EWC related terms based on its regional attributes. However, what does this mean? How should we interpret one region searching EWC related terms proportionately more than another?
At the very least, increased search popularity can proxy for salience of the issue at hand. Additionally, there may be a direction to the relationship. It is possible that searching the topic more expresses more concern and opposition to the policy, or perhaps the opposite is true. It is not even necessary for there to be a relationship between how frequently these terms are searched and the region’s general attitude towards the policy.
However, the possible hypotheses of direction could be explored. One method would be to compare the results from correlations between regional variables and the search popularity values with the results of correlations between the same regional variables and partisan polling data from local elections. If a pattern emerges, perhaps more could be said about the interpretations of search term popularity in the context of EWC.
Limitations to Using Google Trends
There are several issues with using Google Trends data, some of which have been mentioned upon, that should reiterated. Among the problems of using Google Trends data touched upon earlier, is the matter of interpretation. It is often times unclear what Trends data can tell us. Often times, Google Trends is used to support findings produced using other datasets. On its own, Trends data does not come with much confidence behind it. This is also due to its relative nature. Rather than providing absolute measures of search volume, it offers values relative to the most popular moment. Another issue is with the blackbox nature of the search topic queries. Google does not publish the list of terms it associates with the topic of interest. Lastly, another aspect that needs to be taken into account is the very small number of data points offered. This can be mitigated some by constructing panel data manually, however, the sample size would still be very limited in comparison to survey data with thousands of observations.
Coming Soon: Next Steps with EWC
The study is currently in the data processing stage. Most of the annual regional and Trends data has been collected and preliminary analysis is already underway. However, data is still being collected for additional variables of interest. In addition to collecting data for new variables, more data and observations are being sought for more years to allow trends over longer periods of time to emerge.