8.2 Really (just) golfing


Screenshot of online article on golfing study. Retrieved February 6, 2020.

Figure 8.4: Screenshot of online article on golfing study. Retrieved February 6, 2020.


In January, NUS researchers released a study to show the following:

  1. After land sales announcements, some developers play golf with other developers more frequently.
  2. Developers who play golf tend to acquire land parcels at lower prices.

While the press release did not include the words “insider trading”, nor did a later version of the manuscript that I found dated 3 Dec 2019 (before the media report, which was released a month later), an earlier version of the manuscript dated 29 April 2019 did so. It seems that the earlier version of the manuscript was circulated to reporters, instead of the later version78.

The phrase “insider trading” offended the Real Estate Developers’ Association of Singapore (REDAS), who came out strongly against what they saw as an attack against them (see here). In it, we see quite a bit of inflammatory language without any form of real statistical justification. Let us examine the critique as it stands in the official statement.


REDAS takes these allegations seriously. We question the validity of the methodology used and reserve the right to examine the findings to address the faulty assumptions in the paper.

Real Estate Developers’ Association of Singapore (REDAS) statement, dated 17 January 2020


First, REDAS points out “faulty assumptions” and questions the “validity of the methodology”. We have then, to ask, what faulty assumptions? The paper details very clearly the methodology that the researchers used, which includes:

  1. How they constructed the dataset through matching golf records, land bidding results, etc.; and
  2. Their statistical model, which is a difference in difference (DID) regression.

Briefly, a difference in difference regression is a commonly used method for causal inference. It basically uses the change in an outcome before and after an event (after - before; the first difference), and compares that change across groups (group A - group B; the second difference). Informally, the idea can be represented as such:

\[ \text{DID estimate} = (\text{After}_A-\text{Before}_A) - (\text{After}_B-\text{Before}_B) \] The full model allows for covariates to account for other factors, and can allow the effect to vary based on how long ago something happened. Now, we return to our question: what faulty assumptions? At the very least, we should go through each part of the methodology to see what assumptions have been made79.

With regard to the data, we could ask questions such as: Who is in the dataset and who is not? Is this representative of the people who bid for land parcels? Are there land sales records that are missing that would alter the finding substantially? These questions all get at selection bias - basically, whether the data captures information in a way that represents the people we want to study. The process of constructing this data is transparent and described clearly by the researchers - if there were obvious selection biases, REDAS could point them out, but they do not.

We could also ask questions with regard to the method (i.e., DID). In addition to the usual assumptions for linear regression models, the DID regression model has a well known assumption that would render its findings questionable if violated - the “parallel trends assumption”. This means that the trends of groups being compared (e.g., A and B) should be parallel in the absence of the event under study (see this for a fuller explanation). But this is an assumption that is explicitly checked within the study. The authors report that “[the] results show no violation of the parallel trend assumption.”

So what “faulty assumptions” are REDAS referring to? Without any kind of elaboration, the accusations here fall flat.


We are appalled by the lead researcher’s unsubstantiated assertion and the conclusions drawn by the authors are misleading.

Real Estate Developers’ Association of Singapore (REDAS) statement, dated 17 January 2020


Second, REDAS says they are “appalled by the lead researcher’s unsubstantiated assertion and the conclusions drawn by the authors are misleading”. As above, we have to ask: how so? How are assertions “unsubstantiated” and conclusions drawn “misleading”? As we already discovered above, the authors are very clear on how they arrived at their conclusions. They quite robustly substantiate their findings with statistical evidence. Their key findings, once again, are that:

  1. After land sales announcements, some developers play golf with other developers more frequently.
  2. Developers who play golf tend to acquire land parcels at lower prices.

These findings are rather straightforward. Perhaps there is a technical definition of “insider trading” that paints a misleading picture, but to a reader like me, whether or not the actual term “insider trading” is used is besides the point. The finding, as the authors state, is that “social interactions enable developers to realize higher profits, while the government loses land sale revenues”. This means there may be a form of bias in the land bidding system that privileges people with important social connections (i.e., social capital). This is really not a contentious finding at all - researchers have found that social capital can lead to finding better jobs or having better health outcomes. It is rather intuitive that maintaining key network connections are advantageous for life in general (this is why alumni associations exist).

My view is that REDAS looked at these findings and felt attacked because they believe the researchers were making a moral judgment on their professionalism. But even if things look bad on the surface, there may be legitimate reasons for observed trends. For example, Minister for Trade and Industry Chan Chun Sing said recently that permanent residents (PRs) took up a disproportionate amount of new jobs created. He reasoned this was because they wouldn’t be taken in as PRs in the first place if they couldn’t get jobs (see this article). The implication of this selection process is that there is no malice and/or discrimination against locals, even though it may “look bad” on the surface.

Therefore, in this “golfing case”, REDAS should really seek to understand the researchers’ findings more thoroughly before making unsubstantiated judgements on them. Even better, they can work with the researchers to figure out why we see this happening, and whether anything can be done to address the presence of unfair advantage.


In the interest of the community and the nation at large, researchers should publish papers which provide objective and balanced perspective and contribute constructive comments to ensure a stable and sustainable property market.

Real Estate Developers’ Association of Singapore (REDAS) statement, dated 17 January 2020


Finally, we examine the statement that “researchers should publish papers which provide objective and balanced perspective and contribute constructive comments to ensure a stable and sustainable property market.” As we have already discussed, there is no reason to reject the researchers’ findings as not “objective or balanced”. I have read the paper and found that by most standards, they present a good case (with regard to the findings I list above) and are transparent about their discussion. The onus is on REDAS to show how these standards are not met, before casting aspersions on their results. What is worrying is the final phrase that suggests perspective and comments should only be given “to ensure a stable and sustainable property market”. Surely research is not at service of the property market, contrary to what REDAS is trying to suggest. If a well-conducted study reveals a phenomenon that upsets the property market, in no way does this automatically mean that it is not balanced and objective. It is unreasonable to want statistics to always operate in your favour, and/or become combative when it doesn’t. The key question to ask is: why do the numbers say this?

  1. Media reports seem to suggest that the paper was edited after REDAS lashed out at the researchers, but I don’t think this is correct - as I mentioned above, I had already found versions of the article without the claim of “insider trading” before REDAS said anything about it.↩︎

  2. Sometimes, we can search for the word “assume” (or similar) within the manuscript to see what assumptions have been explicitly discussed.↩︎