PSD reviewing paper that alleges civil servants disproportionately bought homes near unannounced MRT stations

SINGAPORE – The Public Service Division (PSD) is reviewing a research paper published by a US-based non-profit organisation that claimed Singapore civil servants disproportionately bought homes near MRT stations that had not yet been publicly announced.

The PSD said on Sept 14 that it is looking at the data and methodology within the academic working paper and will refer any issue to the Corrupt Practices Investigation Bureau “if there is a material basis to do so”.

“We take any integrity concerns extremely seriously,” said a spokesperson for the department, which sits under the Prime Minister’s Office, in response to queries from the media.

The National Bureau of Economic Research (NBER), which is a private outfit headquartered in Cambridge, Massachusetts, published the working paper titled “Do Social Norms Substitute for Enforcement? Evidence from Public Officials’ Home Purchases in Singapore” in September.

The named authors of the paper are Tomasz Piskorski from Columbia University, Amit Seru and Chun Zhao from Stanford University, and Jian Zhang from the University of Hong Kong.

The authors said they wanted to study whether pro-integrity social norms could substitute for formal enforcement in deterring misconduct by public officials.

The paper examines housing transactions around planned expansions of the MRT system, using what it describes as “universe-level housing transactions”.

The authors said they began their study by assembling private-property transactions from legal filings in Singapore and linked buyers to the Singapore Government Directory Interactive (SGDI) – an online database of public officers – to identify civil servants and trace their transaction histories. 

They then focused on purchases near planned – but not yet publicly announced – MRT stations around announcement dates.

“We find clear evidence of informed trading by civil servants,” said the authors.

They cited findings that purchase intensity near planned MRT stations was higher for civil servants compared with a control group of non-civil servants with similar demographic attributes and housing consumption behaviours.

The effect was concentrated one to two years before public announcements and among mid-level officials and agencies connected to rail planning, according to the paper.

The researchers said the pattern generated “meaningful private gains” and also appeared among relatives, which they said was consistent with information leakage.

The paper notes that as a working paper published by the NBER, it is “circulated for discussion and comment purposes”. It has not been peer-reviewed or subject to review by the NBER board of directors that accompanies official NBER publications.

Peer review is a scholarly process where fellow experts or academics independently assess a paper for its validity and quality, among other factors.

The first line of the working paper stated that the authors “benefited immensely from discussions” with a list of individuals, including economist Li Shengwu, the estranged nephew of Senior Minister Lee Hsien Loong. 

Others thanked include several current and former academics at the National University of Singapore such as Sumit Agarwal, Jessica Pan, Ivan Png and Bernard Yeung. 

Png and Pan, both economists, told The Straits Times that they have asked to have their names removed. Png, a former Nominated MP between 2005 and 2006, said he was not aware of the manuscript and did not comment on it.

Pan, who is also NUS’ Vice Provost (Graduate Education), said she was named by mistake and that she only just found out about the paper.

The data and method

The researchers reached their conclusion by tracking the house trading activity of civil servants and other individuals through three data streams, one of which was described as a “proprietary source”.

The proprietary source was described as containing demographic details and residence addresses of “a large sample” of Singaporean residents as of 2013, without going into further detail. ST has contacted the paper’s authors for details on this source.

The other two data streams include six cohorts of civil servants as listed on the SGDI from the years 2007, 2009, 2011, 2013, 2015 and 2022, and the Council for Estate Agencies’ public register of licensed real estate agents in Singapore.

The SGDI data was further augmented using OpenAI’s GPT 4.0, where job titles were mapped into three rank tiers, as the researchers wanted to test if the informed buying behaviour was more concentrated among high-, middle- or low-level officials.

The large language model classified 19 per cent of civil servants of the sample as being of high rank, 39 per cent as mid-rank, and 42 per cent as low rank.

All three data streams were integrated with property transaction data from the legal or caveat filings of private property transactions with the Registry of Land Titles between January 1995 and December 2019, the paper stated. This data included names of buyers and sellers.

The researchers, meanwhile, identified civil servants’ connections by looking up the people listed as co-occupants or co-registrants at the residential address registered to the civil servant.

The paper stated that these individuals were referred to as “relatives” for expositional convenience, as the researchers tested the spillover effects of informed purchases around station announcements.

The researchers used mapping tools to geocode MRT stations and identify private homes that sit within the planned stations’ 1km, 1.5km or 2km radius.

The control group consisted of individuals identified to be broadly similar in age, gender, marital status, ethnicity and estimated monthly income. Income was approximated using a proprietary consumer financial transaction data set from a 2014 paper co-authored by Agarwal, it stated.

The findings

The researchers used their regression estimates to calculate an excess of roughly 113 property purchase incidences by civil servants near planned MRT stations in the two years before announcements, compared with the matched control group. They estimated the gross purchase value associated with these additional incidences at about $138 million.

For people they identified as relatives or co-registrants of civil servants, the researchers similarly estimated an excess of about 108 purchase incidences in the three years before announcements, with an estimated gross acquisition value of about $132 million.

Together, the statistical estimates imply an excess of roughly 221 purchase incidences, with an estimated gross acquisition value of about $270 million.

The estimates do not represent 221 identified buyers or 221 specific transactions that the researchers determined involved misconduct. Rather, they are derived from differences in purchase intensity between civil servants and their matched controls.

As another robustness check, the researchers repeated their analysis using HDB resale flats, where eligibility and resale restrictions make them a less attractive vehicle for trading on advance information. They found no evidence of increased pre-announcement purchases by civil servants relative to matched controls across the different distance cut-offs.

For civil servants, the difference was concentrated in the one- and two-year periods before an announcement.

The researchers’ baseline analysis focused on properties within 1km of planned stations, but they also found similar increases in civil servants’ purchase intensity when they expanded the radius to 1.5km and 2km.

They also found that the properties subsequently generated positive returns relative to comparable properties, with the gains particularly pronounced over shorter holding periods.

The buyers were also more likely to sell earlier, particularly in the first to eighth years after purchase, according to projected estimates – a pattern that the researchers said was consistent with investment-driven behaviour and timely exits.

The purchases were also concentrated among mid-level civil servants rather than the most senior officials or junior staff. Additionally, the researchers found stronger effects among officials working in agencies involved in rail-line planning.

In their analysis, the researchers said the pattern was expected as mid-level officials face the highest net temptation, given that they have sufficient resources to act while bearing lower expected scrutiny and reputational costs than senior officials.

The authors also compared the civil servants’ purchases with those of people who might be expected to have an edge in reading the property market, such as real estate agents and corporate board members.

They did not find the same pattern in either group.

The authors said the absence of similar behaviour strengthened their interpretation that the civil servants’ purchases reflected access to non-public information, rather than “superior processing of public information”.

The study also found that following regulatory and enforcement tightening – changes to the Securities and Futures Act in 2012, among others – there was a “stark” difference in purchasing behaviour.

The authors said the evidence from 2011 onwards showed no comparable increase in informed purchases for both civil servants and their relatives.

This indicates that stricter enforcement substantially reduced both the direct channel of informed trading by officials and the indirect channel through relatives, they added