Endowment Effects of Shared Ownership: Evidence from Hong Kong

ABSTRACT The endowment effect is a behavioral bias; postulating people demand more to part with an object they own than they would offer to acquire the same object owned by others. Recent experimental research, however, suggests that this effect applies to consumption goods rather than exchange goods. Given that housing is a mix of consumption and exchange goods, it presents a unique opportunity to examine whether the effect exists in the real world through a quasi-experiment. Based on a policy change that relaxed the resale constraints of shared ownership housing in Hong Kong, we found that more constrained homeowners were more likely to sell their homes compared to those with fewer constraints when granted the right to exchange, all other factors affecting liquidity being held constant. This marks the first field study revealing that the endowment effect diminishes when a good enhances its exchange nature, such as being given an exchange right.


Introduction
Research in decision sciences makes a strong assumption that subjective and objective ownership are closely related or even identical.This assumption forms the basis for studying consumer behaviours through the principles of preference exogeneity (Korobkin 1998, 611), source independence (Issacharoff 1998(Issacharoff , 1735) ) and status irrelevance (Korobkin 2003(Korobkin , 1228)).However, in certain fields, such as housing, the concept of ownership is multi-faceted (Bright and Hopkins 2011) and varies depending on individual motives (S 2018).In behavioural economics, it is further established that ownership induces a psychological bias called the "endowment effect".A seminal experiment conducted by Thaler (1980) demonstrated that people owning coffee mugs set their minimum selling prices too high, while those without mugs set their maximum offers too low.This phenomenon, known as exchange asymmetry, manifests as an individual's willingness-to-accept (WTA) to sell a good as a seller being significantly higher than their willingness-to-pay (WTP) for an identical good as a buyer (Horowitz and McConnell 2002).As a results, goods subject to endowment effects are less likely to be traded.
These findings challenge the assumption of ownership equivalence and highlight the influence of psychological biases on economic decision-making.While the endowment effect has been extensively examined in controlled laboratory settings (e.g.Kahneman, Knetsch, andThaler 1990, 1991), its implications in real-world contexts remain a subject of investigation.In the related literature, many experimental studies have been conducted, and the theoretical underpinning and interpretation of the effect continue to be the subject of contentious debate among economists, lawyers, and psychologists (Depoorter and Hoeppner 2019).One explanation, based on prospect theory (Kahneman and Tversky 1979;Kahneman, Knetsch, and Thaler 1990), is loss aversion: giving up good from the endowment is treated as a loss, while adding the same good to the endowment is treated as a gain.Since the loss function is assumed to be steeper than the gain function, the negative value of giving up an object is greater than the positive value of acquiring it.
Alternative explanations include trade aversion (Engelmann and Hollard 2010), cognitive focus (Ariely, Huber, and Wertenbroch 2005), and ownership attachment (Carmon, Wertenbroch, and Zeelenberg 2003).Engelmann and Hollard (2010) conjecture that a main driver of the endowment effect is "uncertainty regarding the trading procedure itself," which originated from perceived transaction costs or the risk of offending the experimenter they trade with.Expectations of trade possibilities could explain having subjects repeatedly trade can eliminate the endowment effect.Ariely et al. (2005) propose that the endowment effect arises from different aspects of buyers and sellers emphasized in a transaction.Sellers focus on the item they might give up, whereas buyers focus on the potential expenditure.An experiment using National Collegiate Athletic Association basketball tickets illustrates that selling prices are more sensitive to changes in the salience of specific event attributes than buying prices.Carmon et al. (2003) argue that close consideration of options can induce attachment to them, a "sense of pre-factual ownership."In their account, attachment is generated by psychological proximity, increasing how long or intensely one thinks about an option.People always, in most cases, dislike breaking attachments to an object.
Experimental studies investigating the endowment effect often face limitations due to the nature of goods assigned to participants (e.g.mugs, pens, and chocolates), which are primarily intended for consumption rather than trade.In reality, we rarely exchange our consumption goods -have we ever exchanged our mugs in our lives?The only exception in distinguishing consumption and exchange goods is an experiment by Svirsky (2014), which found the endowment effect only in chocolate coins (which are consumption goods) but not in money or physical tokens (which are exchange goods).Another challenge faced by experimental studies is the potential influence of subject misconceptions inherent in experimental procedures (Plott and Zeiler 2005).These highlight the importance of considering the specific characteristics of the goods being exchanged in studies when examining the endowment effect.
Can the endowment effect survive outside the laboratory and be observed in goods we actually consume and trade?Housing, which is known to have a combination of use value and exchange value, provides an intriguing case to examine the endowment effect in the real world.Previous studies have indicated that goods with the following attributes are important contributors to the endowment effect: high-involvement goods (Saqib, Frohlich, and Bruning 2010), high-quality goods (Azar 2011), and attractive goods (Brenner et al. 2007).Housing fulfils these attributes.Homeowners may be emotionally attached to their homes due to belongingness, lifestyle, and identity (Elliott and Wadley 2013), while their ease of moving is also affected by market conditions in different housing cycles (Jones and Mostafa 2021).However, there have been very few studies on the endowment effect in the housing market (Bao and Gong 2016;Nash and Rosenthal 2014), let alone any quasi-experimental studies.Whether the housing market might exhibit the endowment effect requires a better understanding of its dual nature of consumption and exchange, even though there have been a lot of previous literature on housing commodification (Chiu 2021;Fenton et al. 2013;Wu 2015).
Ideally, a control experiment would involve contrasting the purchase decision of a homeowner with the subsequent sale decision made by the same owner, thereby directly testing the endowment effect in housing purchase.But the reality is that we do not have individual-level data to observe the purchase and sale decisions of the same individual.This absence of individual-level data makes it difficult to ascertain endowment effects or driven by other household-level factors.Nevertheless, in the case of housing transaction, a significant time gap usually exists between the initial purchase and any potential resale, often spanning several years or even a decade.This extended time difference renders direct comparisons between willingness-to-accept (WTA) and willingness-to-pay (WTP) values challenging.Furthermore, as the initial purchase of shared ownership units in this study typically occurs through presale arrangements with the government, the actual product (i.e. the shared homeownership unit) may not yet be observable at the time of evaluation.This complicates the assessment of WTA for homeowners when the tangible product itself is not yet available for examination.These practical considerations make it difficult to design a control experiment that satisfies the conditions necessary for directly examining the endowment effect in the context of housing.
In this study, we test the existence of the endowment effect by exploiting a shared ownership arrangement in Hong Kong, known as Home Ownership Scheme (HOS).The purpose of such shared ownership is to assist middle-income households in ascending the housing ladder (Cheung and Wong 2020).Conceptually, the total value of a shared ownership home belongs to two parties: d% is the government's share, and 1-d% is the owner-occupier's share.The owner-occupier may use the home as if he or she were the sole owner but is restricted from reselling the home unit without first redeeming the government's share.This government's share commonly ranges from 30 to 50 per cent of the market price for comparable units in the private sector.In other words, the owneroccupier can be thought of as being endowed with the entirety of the home's value for consumption but only 1-d% of the home's value for exchange.This "mismatch" between consumption and exchange in HOS is the key feature that allows us to design a quasiexperiment to examine how the endowment effect distorts the trading liquidity of housing.
Our quasi-experiment takes advantage of a policy shock that took place in 2013 when the government suddenly gave owner-occupiers a right to trade without redeeming the government's share.This policy shock offers us an opportunity to isolate the potential endowment effect from other factors.Specifically, this effectively recovered HOS's exchange nature without changing its consumption nature.According to Svirsky (2014), the policy should have made HOS more liquid if exchange goods were freed from endowment effects.In particular, there is a non-trivial testable implication concerning the size of the government's share (d%): HOS units with a higher d% should recover more of their exchange nature and are hypothesized to be traded more frequently than those with a lower d%.Based on a sample of 2,000 HOS units in 170 housing estates as the treatment group and private housing nearby as the control group, our difference-indifferences estimation confirms that after the exogenous policy shock, HOS units with a higher d% in which more exchange value of a good being recovered, will be more actively traded.
The ensuing Section 2 develops testable hypotheses for the quasi-experiment.Section 3 describes the data source.Sections 4 and 5 discuss the quasi-experiment results on transaction liquidity and robustness tests.Sections 6 and 7 discusses and concludes.

Development of Hypotheses
In Hong Kong, the government has implemented the shared ownership policy known as the Home Ownership Scheme (HOS) since the 1970s.The HOS programme aims to respond to the aspirations of people for a better way of life in their own housing units.Since its establishment, the HOS has had over 395,000 units, and it accommodates about one-third of Hong Kong owner-occupiers.Under the HOS, the government sells newly built flats to eligible public housing tenants and low-income residents at a discounted price (1 -d i %) × P it , where d i % is a fixed discount rate for a housing unit i and P it is the market price of a housing unit i at the purchase time t.The d i % can be considered as the government share of the housing unit.This share is usually worth 30-50 per cent of the market price for comparable units in the private sector.
The successful applicant (i.e., the HOS occupier) is usually not allowed to resell her/his flat within the first few years of occupancy.Upon the expiration of this restriction period, the occupier can resell the HOS unit either to anybody at market price after repaying the government's share, i.e. the initial purchase discount amounts plus the capital gain associated with the government share amount; or to the public housing tenants at a negotiated price without repaying the government share value.In either case, the owner-occupiers will receive net sale proceeds of 1 À d i % ð Þ � P iT , where P iT is the market price of housing unit i at the resale time T.
Given the prolonged upwards trend of property prices in Hong Kong, many HOS owner-occupiers should be able to make considerable profits by reselling their units.However, statistics indicate that the HOS resale market has been extremely inactive in Hong Kong.Every year only about 1% of total HOS stock has been resold, well below the 9% observed in the private housing market.In contrast, new HOS units in the first-hand market (sold by the government) are always heavily oversubscribed.Why have so many HOS units in the secondary market remained "frozen" and unsold to potential buyers?Is it because HOS owner-occupiers suffer from the endowment effect, which unreasonably raises their willingness-to-accept the trade of their homes?
With the White Form Buyer (WFB) policy implemented in 2013, some HOS owneroccupiers were enabled to sell their flats to eligible buyers without repaying the government's share.They can also negotiate the price with eligible buyers on the market. 1Having been selected by ballots (a quota of 5,000 in each batch), qualified buyers can then negotiate with HOS owner-occupiers for their favourable HOS units within 12 months.A sale would occur when the willingness-to-pay of the eligible buyers is higher than the willingness-to-accept of the HOS owner-occupiers.Unlike negotiations in private housing markets, there is little information on the market price of government shares.This WFB policy acts as a quasi-experiment to examine how various amounts of initial discount (i.e. government share, d i % of housing unit i) moderate the endowment effect of HOS owneroccupiers to trade their housing units.
Similar to the conditions in most experiments on endowment effects, if the owneroccupiers of shared ownership possess a significant endowment effect, the willingness-to -accept of the owner-occupiers will be much greater than the willingness-to-pay of the buyers, and hence trade is less likely to occur.In this study, however, it is argued that the endowment effect follows Svirsky (2014) and is contingent on the nature of exchange goods.That means that apart from the housing endowment for consumption; we have to take into account the share of home value for exchange, i.e. the government share (d i %).
First, consider housing as a pure exchange good without any endowment effect.For each housing unit, V(S j ) is the willingness-to-accept of the HOS seller or owner j and V(B k ) is the willingness-to-pay of buyer k.A housing unit will be traded if a buyer values the home unit more than its owner, i.e.V(B k ) > V(S j ).If there are many sellers and buyers in the market, and their V(B k ) and V(S j ) are normally distributed, as shown in Figure 1. 2 The probability of trade (T) is indicated by the overlapping area. 3 Next, consider housing as a pure consumption good, the owner of which is subject to endowment effects.Owners' willingness-to-accept becomes V S j À � þ E j , where E j > 0 is the compensation required by owner j to offset the endowment effect.A housing unit will be traded if a buyer values the home unit much more than its owner does, i.e.
With the distribution of sellers' willingness-to-accept shifted to the right, the overlapping region is reduced, hence there is a lower probability of trade T E < T (Figure 1).This can be regarded as the situation for shared ownership homes intended for consumption only before the implementation of White Form Buyer Policy.
Finally, consider a hybrid situation where the nature of housing is partly for consumption and partly for exchange.This is the consequence of the White Form Buyer Policy, which gives owner-occupiers a right to exchange without repaying the government share (d i %).Imposing an exchange nature on a consumption good leads to a reduction of the endowment effect.The bigger the government share, the greater the exchange nature imposed, and the larger the reduction of the endowment effect.We can express the endowment effect of a hybrid good as A housing unit will be traded if this condition is satisfied: The overlapping region is now larger than the one for a pure consumption good but smaller than the one for a pure exchange good.Hence, the probability of trade in this hybrid situation falls between T E and T. This implies our hypothesis that: Hypothesis: Ceteris paribus, the higher the government share (d%), the higher the probability of trade under the WFB policy.
In order to test the hypothesis, we apply the spatial-temporal difference-in-differences (DID) regression model to test how the implementation of the WFB policy affects the transaction liquidity between HOS flats and private flats nearby.
The quasi-experiment considers the WFB policy as an intervention, and the HOS housing units become the treatment group.Private housing units within 400 m of the HOS estates are taken as the control group (counterfactuals).The effect of government shares on the change of HOS transaction liquidity compared to the counterfactuals after the WFB policy is estimated by the DID models (Figure 2).Details of the White Form Buyer policy are in the Appendix.

The Data
To assess the effects of the WFB policy on the monthly transaction liquidity (L%) of HOS estates, the transaction data of about 2,000 shared ownership housing units in 170 housing estates with trading activities from July 2011 to March 2015 were considered, while the comparable private housing units within 400 metres of each HOS estate were used as the counterfactuals.The monthly transactions of private housing units recorded nearby provide a relevant control group (i.e.counterfactual) for benchmarking the impact of the initial discount rate on HOS trading activities under the WFB policy.
The counterfactual is constructed along the Neyman (1923) potential outcome framework of causal inference.Causal effects, in our case, refers to the net increase or decrease observed in the liquidity of the shared ownership attributed to malleable variables in the policy intervention (i.e.Endowment effect).Without counterfactual inference, the observed increase in shared ownership could be attributable to some confounding variables happening behind the scenes, such as market conditions.Using the comparable private housing units within 400 metres of each HOS estate as the counterfactuals, the transaction liquidity of shared ownership can be "benchmarked" against private ownership.If any, the net difference observed in liquidity between shared and private ownership in a 400-metre neighbourhood before and after the policy intervention can be more reasonably described as a "cause" to the endowment effect.
Indeed, it is worth noting that there are some other plausible explanations for the increased liquidity in the market for shared ownership.These include other reasons for moving homes, for example, the downsizing of the life cycle, job change, marriage and other push factors that might encourage movement overwhelming consumption, as well as the transaction costs inhibiting home moves.The difference-in-differences used in this study can capture most, if not all, of the reasons for the increase in transaction liquidity after the WFB policy.The rationale is that these demographic attributes could be similar within 400-metre neighbourhoods owing to the Tiebout sorting.Having said that, not all counterfactuals are perfect.One may still argue that the shared owners and private owners are different.Nevertheless, to a certain extent, the counterfactual analysis can be made more precise estimand by specifying the contrary-to-fact scenario in view.We can be more confident in admitting that the potential increase in liquidity in shared ownership due to the WFB policy is the endowment effect.
The study period has been carefully chosen for good reasons.As 5,000 ballots of white form, secondary market buyers were issued from June 2013 to March 2014, and the selected households were allowed to search for their flats over the next 12 months; all granted white form buying rights should have expired by March 2015.The study period was thus set to commence in July 2011 to balance the time frame before and after the launch of the WFB policy.
Table 1 shows the means of the variables in the dataset.The means for variables in the HOS and private housing estates before and after the launch of the WFB policy are presented in Columns (1) to (4).After the implementation of the policy, the number of monthly HOS transactions averaged 6.19, which slightly exceeded the transaction volume (i.e.6.07) before the 5,000 waivers were introduced.Nevertheless, such an increase should not be regarded as trivial compared to the private housing market figures.Due to various tightening measures, 4 trading activities in the property market remained slack over the study period.The number of transactions for private units adjacent to HOS estates, on average, dropped from 15 to 11 cases per month.This reinforces the argument that the introduction of the WFB policy boosted the relative number of transactions in the HOS resale market.An initial assessment of the WFB policy programme 5 also indicates that the average number of annual transactions in the HOS market increased to 2,810, more than the annual average of 2,100 in the previous decade.The WFB policy programme boosted most of the transactions during the period.
More importantly, the government pledged to have another batch of the WFB policy programme.Those HOS households who had planned to buy their HOS units without getting the ballot were expected to have a strong incentive to wait for the impending ballots rather than buying private housing units nearby.That further helps ensure the validity of the comparison between shared ownership homeowners and private homeowners in the quasi-experiment.Furthermore, the main attributes of housing transactions before and after the WFB policy programme, including estate age (AGE), floor level (FLR), and flat size (SIZE), are likewise found to be similar for the HOS and private estates.Thus, buyers' housing preferences should not have changed much after the implementation of the WFB policy.

Quasi Experiment on Transaction Liquidity
To test the hypothesis, we conducted a quasi-experiment with the WFB policy.Precisely, the identification strategy is to make use of the WFB policy 6 in 2013 as a randomized exogenous shock (i.e.random ballots) to test the reactions of shared ownership homeowners being granted a right to trade without redeeming the government's share.According to our Hypothesis, it can be predicted that those HOS households holding a more significant proportion of the home's value for exchange will be more likely to accept the compensation and dispose of their properties.A higher HOS transaction liquidity can then be envisaged in the following criterion.First, HOS units with a smaller share held by the seller, i.e. a larger share held by the government, which is measured by the initial purchase discount rate times the current market price, d i % � P T for the HOS unit i, are expected to be more frequently traded because under the WFB policy, homeowners own less endowed goods but more exchange goods.The relevant regression models for the analysis of the hypothesis are presented as follows.Equation ( 1a) is a baseline model of the differences-in-difference (DID) model to test whether the HOS market becomes more active after the WFB policy, ceteris paribus.Then β 1 þ γ 1 À X 1 > 0 where β 1 þ γ 1 and λ 1 represent the changes in liquidities of the HOS and private housing markets after the WFB policy, respectively.where Lmt is the monthly transaction liquidity in per cent at estate m (i.e.number of transactions of housing units relative to the total housing stock in the estate, at time t.HOSm is a dummy variable that equals one if the observation refers to a HOS estate, or 0 if it refers to the private housing estates nearby.WFBPt is a WFB policy time dummy that equals 1 when the transaction occurs after 2013M6 (post-treatment) or 0 if otherwise.Estate fixed effects and temporal fixed effects are controlled.
Equation (1b) tests the government share effect on transaction liquidity 7 .The measure of the government share, is the initial purchase discount rate of the HOS estate.The coefficients is expected to be larger than because HOS estates with higher discount rates (i.e. a larger proportion of exchange goods) are more likely to be resold after the WFB policy is implemented.

Results for Quasi Experiment
The results in Table 2 align with our prediction of the hypothesis (Equations 1a and 1b).In Columns (1a) and (1b) of the table, the policy time dummy WFBP (i.e.WFBP = 1 when the transaction occurred after 2013M6 or 0 otherwise) carries a significantly negative sign that coincides with the tightening impact of the property cooling measures.Model (1a), the baseline model, confirms that the HOS estates became more active than private housing estates nearby after the WFB policy is implemented, ceteris paribus.
This shows, after implementation of the WFB policy, a 34.2%net increase in the transaction liquidity of HOS in comparison with that of private housing in the vicinity of the HOS estates.
In Model (1b), of particular note is the DID effect of government share on transaction liquidity, which is captured by the significant and positive results of the coefficients of the interaction term β 3 ¼ 0:010, and β 3 > > β 2 ¼ À 0:023.This confirms the hypothesis that a higher proportion of exchange goods nature in shared ownership housing units increases the transaction liquidity.Furthermore, in comparison with the private housing nearby, the transaction liquidity of the HOS for an additional unit of initial discount (d%) changed from −2.3% before the WFB policy to + 1.0% after the WFB policy.This implies that HOS estates with larger discounts, which contain a higher proportion of exchange goods, are more actively traded after the WFB policy is implemented.
We have carried out typical regression diagnostic tests, including the Variance Inflation Factors test on multicollinearity and Breusch-Pagan-Godfrey (BPG) test on heteroskedasticity.The uncentered VIF of most non-dummy variables is greater than or equal to five.The BPG tests reject the null hypothesis of homoskedasticity, and the models are reestimated by Huber-White-Hinkley (HC1) heteroskedasticity consistent standard errors and covariance as Models (1c) and (1d).The results remain almost intact.

Robustness Tests
To validate the findings, three robustness checks are carried out, encompassing a rigorous test that eliminates any potential alternative explanation, temporally and spatially.
The first robustness test is to confine the sample period to 5-month before and after the WFBP shock, to minimize any effects from other events occurring outside the period window.The subsampling sharply reduces the number of observations to one-fourth of the main regression, yet the model explanatory power doubles as R-squared increases to 31.6%.The results of the sample period test in Table 3 further support the hypothesis that the transaction liquidity of HOS has immediately increased by 0.15% to 0.18% in the two quarters at and immediately after the first trial ballot of WFBP compared to the liquidity at the quarter before.
The parallel trend test is also provided to assess the assumption of parallel trends in our quasi-experimental designs.The parallel trend assumption is that the treatment and control groups would have followed parallel trends over time in the absence of the intervention.In other words, any differences in outcomes observed between the two groups after the treatment is introduced can be attributed to the treatment itself rather than pre-existing differences in trends.In the parallel trend test, the results show a parallel trend between the transaction liquidities of HOS and private housing estates nearby before the policy shock, as reflected by the similar coefficients of 0.20 and 0.21 for HOS � Y 2012 and HOS � Y 2013, respectively.Excluding buildings with housing stock less than or equal to 100 units, the total number of observations is 12,650.***, **, * means the coefficient is at the 1%, 5%, 10% significance levels, respectively.Standard errors are presented in parentheses.WFBP is a policy dummy that equals 1 when the transaction occurred after 2013M6 or 0 if otherwise.HOS equals 1 when the transaction liquidity refers to a HOS estate and 0 for its private counterparts.The variable d% refers to the initial purchase discount, which is the government share in this study.The fixed effect is a set of HOS estate dummies to capture, if any, estate-specific impact on transaction liquidity.
The second robustness test is to confine the sample to its spatial location.Table 4 shows the sample district robustness tests by dividing the HOS estates into the three major districts in Hong Kong, viz.Hong Kong Island (HKI), Kowloon (KLN) and the New Territories (NT), where housing is in general more expensive in HKI than KLN and NT, as HKI is closer to the city centre.The sample sizes are reduced to 1663, 2438 and 6608, yet the results still support the hypothesis that HOS was priced 0.29%, 0.14% and 0.15% higher after the WFBP in HKI, KLN and NT, respectively.
The third set of robustness checks involves further exploration of the implications of WFBP impacts on various sizes of HOS units, aiming to rule out potential alternative explanations.Concerns may arise that WFBP itself could function as a demand factor, offering another rationale for the increase in liquidity.This policy directly engages buyers on a lengthy waiting list for subsidized homeownership, acting as a determinant of demand.Should the WFB policy indeed assume this role, it would contrast with endowment effects, which relate primarily to sellers' behaviour and resemble a supply factor, eliciting a unique market response.Specifically, the endowment effect implies that existing homeowners, especially those with subsidized home units, would not be induced to sell.This delineation between the influence of the WFB policy on demand and the endowment effect on supply positions the latter as a supply-side factor, furnishing testable implications to further explore the endowment effects on various transaction sizes within our sample.
To exclude the WFB as an alternative demand factor and to further test the existence of the endowment effect in the study, we subdivide the transaction liquidity into two groups: i.e. large flat units (55 sq metre or above) and small flat units (below 55 sq metre).The rationale behind this grouping is that endowment effects should not be dependent on unit size, while the policy might particularly stimulate demand for smaller units due to affordability considerations, as evidenced in the Legislative Council Panel on Housing Review of the Interim Scheme to Extend the Home Ownership Scheme Secondary Market to White Form Buyers (Legislative Council of Hong Kong, 2017-2017-2018).Table 5 shows the transactions under the WFB mainly concentrated below the 55 sq metre range.
Under this premise, it is conceivable that policy intervention could heighten the demand for smaller-sized HOS units, reflecting concerns about affordability and credit constraints (Wong, Deng, and Cheung 2022), also referred to as the starter home premium in the literature.Our tests differentiate between the liquidity of larger and smaller HOS units.By contrasting the liquidity of adjacent private housing, our analysis affirms the presence of endowment effects in the context of larger-sized HOS units but does not corroborate similar findings for smaller-sized units.These results bolster the endowment effect hypothesis.WFBP is a policy dummy that equals 1 when the transaction occurred after 2013M6 or 0 if otherwise.HOS equals 1 when the transaction liquidity refers to a HOS estate and 0 for its private counterparts.The fixed effect is a set of HOS estate dummies to capture, if any, estate-specific impact on transaction liquidity.
Table 6 showcases the outcomes of our robustness assessment.Initially, we commenced with an analysis of the socio-economic status of households residing in the corresponding building groups.In this examination, we introduce both the median household income and post-secondary educational attainment of the building group where the HOS estate is situated as a control variable, replacing the fixed estate effect, as detailed in Column 1 of Table 6.The data, drawn from Centamap's 2011 population census on Building Groups (Census and Statistics Department Hong Kong 2011), reveals a positive yet statistically insignificant income influence on transaction liquidity.Whilst post-secondary educational attainment demonstrates a statistically significant positive impact on transaction liquidity, the endowment effect has shown the same (WFBP × HOS), with its statistical significance remaining intact.Nevertheless, it is not feasible to include income and educational attainment variables within the same equation owing to the inherent endogeneity in these two factors.In Column 3 of Table 6, the logarithm of income (ln(INC)) exhibits a negative shift when these variables are introduced.This counterintuitive alteration in the sign of ln(INC) further underscores the presence of such endogeneity.These tests also support the effectiveness of our estate fixed effects in capturing socio-economic status influences on transaction liquidity.The 400-metre radius control in specifying the control in the differences-in-differences estimate appears to be sufficiently precise to eliminate any significant difference in neighbourhood effects.Notably, the endowment effects remain consistent across these results.
In Columns 4 and 5 of Table 6, we extend our robustness examination to delve into an alternative hypothesis.We scrutinize whether the observed change in liquidity can be ascribed to policy objectives rather than the endowment effect.Specifically, only the interaction effect of WFBP × HOS for larger-sized housing proves significant, but not for small-sized flats.This finding further confirms the endowment effect resulting from the WFBP, reinforcing our understanding of how different aspects of the policy interact with housing sizes to influence the market.

Discussion
This study uniquely applies a policy shift in Hong Kong's subsidized housing market as a quasiexperiment, facilitating an empirical investigation of the endowment theory.Eschewing conventional laboratory-based research methods, which may not fully capture real-world HOUSING, THEORY AND SOCIETY behaviour, this investigation adopts a revealed preference for quasi-experimental design.This innovative approach offers a more accurate representation of real-world behaviours in housing markets, contributing to a more robust understanding of the endowment effect.Previous research has indicated the endowment effect, which suggests homeowners value their properties more than potential buyers, as a crucial factor in reducing residential mobility among subsidized housing occupants.As reinforced by the endowment effect, the psychological attachment often deters these homeowners from trading or relocating their properties.In a novel exploration, this study examines the division of exchange value from consumption value within the framework of shared ownership in subsidized housing.In such a setting, homeowners retain a partial claim to their properties while still benefitting from governmental subsidies.This duality appears to dilute the endowment effect, as owners are not entirely relinquishing their consumption rights when they sell their homes.
Interestingly, this research finds that when exchange value is uncoupled from consumption value, shared ownership homeowners demonstrate a higher propensity to trade their properties.This discovery implies that the endowment effect may indeed be mitigated when homeowners can dissociate their emotional attachment to the home (the consumption value) from the potential financial gain of a sale (the exchange value).These insights hold significant implications for social housing policy.The endowment effect has been linked to the illiquidity of the social housing market, complicating the trading process and posing challenges for households seeking relocation to more suitable locations (Clark and Lisowski 2017;Yan and  Notes: The dependent variable is the monthly transaction liquidity (L%) of 170 HOS and private estates within a 400 m radius of the estates from 2011M7 to 2015 M3.L% Large , represents, large flat units (55 sqm or above) and L% small is the flat units (below 55 sqm).**, * means the coefficient is at the 1%, 5% significance levels, respectively.Standard errors are presented in parentheses.WFBP is a policy dummy that equals 1 when the transaction occurred after 2013M6 or 0 if otherwise.HOS equals 1 when the transaction liquidity refers to a HOS estate and 0 for its private counterparts.INC refers to the median household income of the building group where the estate is located.EDU presents the proportion of residents who received post-secondary education within the building group where the estate is located.The fixed effect is a set of HOS estate dummies to capture, if any, estate-specific impact on transaction liquidity.The estimate is in Huber-White-Hinkley (HC1) heteroskedasticity consistent standard errors and covariances.
Bao 2018).Implementing a policy that supports the separation of exchange value from consumption value, akin to shared ownership schemes, could potentially enhance transactional fluidity, thereby improving the liquidity of social housing units.Such an approach could prove especially beneficial in regions where affordable housing demand outstrips supply.By facilitating more efficient housing redistribution and enabling more households to move to appropriate locations, this strategy could effectively address spatial mismatch issues, improving the overall efficiency of the social housing market (Cheung et al. 2021).Thus, this study offers a potentially effective strategy for policymakers to enhance the functioning of social housing markets by addressing the inherent challenges posed by the endowment effect.
One of the limitations of this study is that, we were unable to access comprehensive demographic information for each individual buyer or seller.Furthermore, we lack the ability to track the post-sale movements of owners, limiting our analysis to an estate-level focus.Despite our efforts to control for estate characteristics, these data limitations impede our capacity to ideally observe decisions made by the same individual concerning the same property.Consequently, the absence of individual-level data complicates the task of determining whether our findings are purely indicative of endowment effects or are possibly influenced by other latent variables.

Conclusions
This study examines the concept of endowment effect in decision making and challenges the assumption that subjective and objective ownership are the same.The analysis investigates the implications of the endowment effect beyond controlled laboratory settings, particularly in the housing market.Using a shared ownership arrangement in Hong Kong, the study capitalizes on a policy change in 2013, allowing homeowners to trade without redeeming the government's share, to examine the endowment effect in a real-world context.The study proposes that this policy change should increase the trading liquidity of HOS units, as they gain more of their exchange nature.The results confirmed that after the policy change, units with a higher government share were traded more frequently, affirming the impact of the endowment effect in housing market transactions.
Specifically, two key findings were observed in this quasi-experiment.First, the introduction of the WFBP led to a significant increase in transaction liquidity in the HOS market.This means that after the policy change, there was a 34.2%net increase in transaction activity in the HOS market compared to private housing estates in the same vicinity.This suggests that homeowners were more likely to dispose of their properties under this policy, which allowed them to trade without redeeming the government's share.Second, there was also evidence that HOS estates with a larger government share saw increased transaction liquidity.Before the WFBP, each additional unit of initial discount (a proxy for government share) resulted in a −2.3% change in transaction liquidity.After the WFBP, this changed to a + 1.0% effect.This implies that properties with a larger government share were more actively traded after the policy was implemented.Several robustness checks were conducted for verify the finding, including limiting the sample period to the five months before and after the WFBP policy shock, a parallel trend test, and district-based analyses.All these tests supported the initial findings.The results suggest that changes in policy can significantly affect transaction liquidity in the housing market, particularly in markets with shared ownership schemes.
What does the quasi-experiment result on endowment effects tell us?When Kahneman, Knetsch and Thaler (1990) conducted their behavioural experiment on endowment effects, the purpose was to test whether ownership matters.Contrary to theoretical expectations from neo-classical economics, they found that consumers' willingness-to-accept greatly exceeds their willingness-to-pay, hence the undertrading of consumption goods.They add that "to assess the practical significance of the endowment effect, it is important to consider first some necessary conditions for the effect to be observed" (Kahneman, Knetsch, andThaler 1990, 1344).In other words, there is no presumption that endowment effects must exist for all types of goods, including housing.The contribution of our study is to show that endowment effects diminish when the nature of good changes from consumption to exchange.The quasi-experiment result shows that when a good contains a higher proportion of exchange value, the good will not always be "under-traded".
One practical implication of our quasi-experiment on housing is that the lack of trades in shared ownership results from the owner not being assigned the exchange right.Property rights theory tells us that commodities always consist of a bundle of separable and transferable rights, and ownership of a commodity is a function of the extent and form of control over these rights. 8The key to remedy the under-traded issue is how the exchange rights can be more clearly delineated.This perspective follows Coasian thought (Coase 1937(Coase , 1960) ) and resonates with the ideas of Nygaard, Gibb and Berry (2007) in transferring the de facto property rights to a new legal entity for social housing.Nygaard, Gibb and Berry (2007) examined how ownership transfer of social housing in the United Kingdom.They argued that those ownership transfers only involve the de jure transfer of property to a new legal entity, but significant parts of de facto property rights remain "untransferred" or have been redistributed within the public sector and hence do not necessarily constitute a redraw of the public domain boundary.
Even though our quasi-experiment results are confined to the case of Hong Kong, endowment effects of shared ownership are expected to have a broader significance in other international housing policy contexts, especially when the government-backed shared equity is not uncommon in many commonwealth countries.In the UK, the government seeks to find niche ways to expand homeownership to potential first-time buyers.Council home tenants are given the right to buy their rental units where the purchase discounts rise with the length of tenancy.This was effectively a transfer of assets from the public sector to the eligible household.Indeed, little empirical attention has been given to the mobility constraints linked to homeownership (Jones & Mostafa, 2021).It is crucial to explore alternative avenues for potential solutions, review the key characteristics of owneroccupied housing markets and understand the residential mobility process with institutional and behavioural economics (Marsh and Gibb 2011).
This study has argued that endowment effects vary with the exchange value of a good.A separation of such exchange value from the consumption value enables shared ownership owners to substantially increase the stake and value of their ownership.The proposed property rights arrangement will also increase the flexibility of shared ownership.Its spirit is in line with the notion of housing vouchers, which has been hotly debated in recent housing literature -whether a cash equivalent shared ownership housing enables public housing residents to enjoy greater mobility and thereby leads to a substantial improvement in their well-being (Eriksen & Ross, 2015).Regrettably, most housing policy debates so far, whether in Hong Kong or many other economies, have mainly concentrated on providing adequate, affordable housing through government policies, with little regard to the role of institutional and behavioural economics rooted in shared ownership.This study is meant to fill such a void by an illustration.

Notes
1.The official name of the premium waiver policy is called "The interim scheme of extending the HOS secondary market to White Form buyers".2.Besides probability of trade, the endowment effect hypothesis also implies that the price of the HOS unit will be higher after the policy.The empirical test on the price effect requires a very different approach to model, and therefore it will be reported separately in another working paper.3.As we study the endowment effect empirically, the willingness to pay of all potential buyers is assumed to form a normal distribution, and the intensity of the endowment effect is assumed to be reflected by the change in the trading liquidity (probability of trade) in the specific housing market.4. The tightening measures apply to the resales of HOS units with premium settled. 5.An initial assessment of the WFB policy (i.e.Interim Scheme) can be available at http://www.legco.gov.hk/yr14-15/english/panels/hg/papers/hg20150105cb1-384-4-e.pdf 6.The WFB policy is officially called as "The Interim Scheme to Extend the Home Ownership Scheme Secondary Market to White Form Buyers".7. Two terms are omitted in Equation (1b), they are d m and d m � WFBP t .It is because private housing does not have discount rates.These two variables will be exact collinear with HOS m .8. Nygaard et al. (2007) suggest these attributes include an asset base (wealth), a source of revenue (rent), a source of patronage or constituency building, a community asset and a resource (vehicle) for homelessness and urban renewal policy.9.According to the memorandum for the subsidized housing committee on the initial ideas on price setting and for calculating the premium payable under the new Home Ownership Scheme (HKHA 2012), for the traditional HOS, the discount rate was determined based on the affordability of the households within the HOS income limit, as guided by the following two principles: "(a) eligible households could afford the flats with a mortgage-to-income ratio of not more than 40%; and (b) at least 50% of the flats for sale should be 'affordable' as defined in (a) after applying a discount rate to the market value of flats".

Disclosure Statement
No potential conflict of interest was reported by the author(s).

Funding
The Appendix -Background of the HOS scheme and the White Form Buyer (WFB) policy In Hong Kong, since the 1970s the government has implemented the shared equity affordable homeownership policy, known as the Home Ownership Scheme (HOS).The HOS programme aims to respond to the aspirations of people for a better way of life living in their own housing units.Since its establishment, the HOS has had over 395,000 units and accommodates about one-third of owner-occupiers in Hong Kong.Under the HOS, the government sells newly built flats to eligible public housing tenants and low-income residents at discounted prices, usually at 30-50 per cent (initial discount rate, d%) below-market rates for comparable units in the private sector subject to resale restrictions.The discount rates are determined by the general affordability of eligible households at the time of sale, rather than the affordability of the individual home buyers. 9Before 1997, HOS owners could resell their flats only to the government at their original purchase prices within the first five years of ownership.From the sixth to tenth year, they could resell them to public housing tenants (green form applicants).Only after 10 years could a resale occur on the open market yet still be subject to the payment of a premium proportional to the original purchase price discount.
To show how the HOS resale market works, consider a typical case where the government has sold a flat to a successful HOS applicant at a discounted price.The applicant (i.e., now owner) usually is not allowed to resell his flat within the first few years of occupancy.Upon the expiration of this restriction period, the owner can resell his shared ownership flat, either (1) to anybody at a market price after settling the government share (land premium), which is equivalent to the initial discount rate times the market prices; or (2) to public housing tenants at a negotiated price without repaying the government share (land premium).In either case, the owner will receive net sale proceeds of approximately (1 -d%) × Market prices.Given the prolonged uptrend of property prices in Hong Kong, many HOS flat owners should be able to make considerable profits if they resell their flats.However, many relevant statistics indicate that the HOS resale market is extremely inactive in Hong Kong.Only about 1% of total HOS stock has been sold to public housing tenants every year, which is well below the 9% recorded for private housing.Why have so many shared ownership home units in the secondary market remained "frozen" and unsold to potential buyers on the market?The high price in the private housing markets can partially explain the lower number of HOS transactions, but it should be equally low in both primary and secondary markets of HOS.The frozen state of HOS is concentrated in the secondary market only.In contrast, all sales of new HOS in the primary market are heavily oversubscribed.
In 1997, to increase HOS flat turnover and provide more incentives for public rental housing tenants to own a home, the government relaxed certain HOS resale conditions.After the first three (instead of five) years of ownership, HOS owners could resell their flats to public housing tenants.From the sixth (instead of eleventh) year of residence, they could also choose to remove all the resale restrictions and sell their flats to any buyers by paying the government a premium proportional to the original purchase price discount.In any circumstance, those owners who have disposed of their HOS flats will no longer be eligible for any other form of public housing.
In January 2013, the government implemented a "White Form Buyers Policy" in an attempt to encourage HOS owners to sell their flats to eligible white-form buyers without paying back the share to the government.The qualified white-form buyers, who would not be existing public housing tenants (green form applicants), having been selected by ballots (a quota of 5 000), could then find and negotiate their favourable HOS flats with the homeowners within 12 months.Whether a sale would occur depends on the negotiation between the eligible buyers and HOS owners.The first round of the scheme, with a quota of 5000, was issued from June 2013 to March 2014 and completed in April 2015.The second round of the Interim Scheme with a quota of 2500 was launched in August 2015, and it was also completed in May 2017.Over 4000 WF applicants achieved home ownership through the two rounds of the Interim Scheme.
After a review and taking into account the sustained demand from White Form buyers, the government endorsed the regularization of White Form Buyer (WFB) policy in November 2017.White Form Buyer (WFB) policy 2018 was launched in March 2018 with a quota of 2500 and in May 2019 with a quota of 3000.The latest round in 2020 was launched in September 2020 with a quota of 4500.When the government endorsed the extension of the HOS Secondary Market to White Form buyers in 2012, it had taken into account the needs of family applicants and one-person applicants for SSFS flats.At that time, considering that family applicants had a more pressing need to purchase a flat as compared to one-person applicants, the HA set the allocation ratio for family and oneperson applicants at 9:1.This was in line with the arrangements for other subsidized housing policy where higher priority is allocated to family applicants.

Figure 1 .
Figure1.Probability of trade as the overlapping region of the distributions of willingness-to-accept and willingness-to-buy with and without endowment effect.

Figure 2 .
Figure 2. The major changes brought by the white form Buyer (WFB) policy.

Table 2 .
Results of the quasi-experiment on transaction liquidity (DID estimates).The dependent variable is the monthly transaction liquidity (L%) of 170 HOS and private estates within a 400 m radius of the estates from 2011M7 to 2015 M3.The total number of observations is 15 300 (i.e.170 × 45 × 2).

Table 3 .
Results of the robustness tests of sample period and parallel trend on transaction liquidity (DID estimates).
Notes: The dependent variable is the monthly transaction liquidity (L%) of 170 HOS and private estates within a 400 m radius of the estates.The sample period test limits the sample period from 2013M1 to 2013M12.Excluding buildings with housing stock less than or equal to 100 units, the total number of observations is 3,352.In the parallel trend test, the sample period is from 2011M7 to 2015 M3, the total number of observations is 12,650.***, **, * means the coefficient is at the 1%, 5%, 10% significance levels, respectively.Standard errors are presented in parentheses.HOS equals 1 when the transaction liquidity refers to a HOS estate and 0 for its private counterparts.The fixed effect is a set of HOS estate dummies to capture, if any, estate-specific impact on transaction liquidity.

Table 4 .
Results of the robustness tests of sample district on transaction liquidity (DID estimates).The dependent variable is the monthly transaction liquidity (L%) of 170 HOS and private estates within a 400 m radius of the estates from 2011M7 to 2015 M3.The three sample district tests confine the sampled HOSs from the three major districts of Hong Kong, viz Hong Kong Island (HKI), Kowloon (KLN) and the New Territories (NT).Excluding buildings with housing stock less than or equal to 100 units, the total numbers of observations in HKI, KLN and NT are 1,663, 2,438 and 6,608 respectively.***, **, * means the coefficient is at the 1%, 5%, 10% significance levels, respectively.Standard errors are presented in parentheses.

Table 5 .
The flats purchased by GF and WF buyers, by size under the WFBP.

Table 6 .
Results of the robustness tests on transaction liquidity.