To evaluate the effect of Green Mark scheme on household utility consumption, we use a staggered difference-in-differences research design. We restrict the baseline sample to public housing buildings constructed between 1981 and 2010 that did not implement other major efficiency improvement programs to ensure similarity in building characteristics between the control and treatment groups. The treated buildings are those that attained Green Mark between 2012, the start of Green Mark for existing residential buildings, and 2017, to ensure at least three years of utility consumption observed post-certification. The control group constitutes of the never-certified residential buildings.
Water consumption: By estimating Eq. (1) in Methods, we find that the Green Mark certification increases household water consumption by 3.3% (Table 1 Panel A Column (1)). Evaluated at the mean pre-certification consumption of 18.4 m3 per month, it translates to an increase of 53.3 m3 in water use per block, given an average block size of 88 households. Meanwhile, using the same empirical method, we estimate a 11.2% or 10.5 m3 monthly reduction in the common area water consumption post-certification in the certified buildings, relative to their never-certified counterpart (Table 1 Panel B Column (1)).
Table 1
Effect of Green Mark certification on water use of existing residential buildings
| (1) | (2) | (3) |
| Monthly | Annual |
| Weighted | Standard |
Panel A: Household | | | |
Post Certification | 0.033*** | 0.027*** | 0.031** |
| (0.004) | (0.003) | (0.004) |
N | 53,407,060 | 4,442,199 | 4,442,199 |
Panel B: Common area | | | |
Post Certification | -0.112*** | -0.108*** | -0.133** |
| (0.021) | (0.050) | (0.042) |
N | 305,380 | 26,351 | 26,351 |
Account FE | Yes | Yes | Yes |
Year FE | No | Yes | Yes |
Year-month FE | Yes | No | No |
Notes: This table presents the estimates on the effect of green building certification on household (Panel A) and common area (Panel B) water consumption of existing residential buildings, relative to their never-certified counterparts, use Eq. (1) Methods. The dependent variable is log of account level monthly water consumption for column (1) and log of account level annual mean monthly water consumption for columns (2) and (3). Columns (1) and (3) show the effects estimated using standard staggered difference-in-differences method, while column (2) uses the weighted estimation method proposed in Callaway and Sant’Anna (2021). All models control for account fixed effects and time fixed effects. The standard errors shown in parentheses are clustered by block. * p < 0.1, ** p < 0.05, *** p < 0.01. |
To mitigate potential biases arising from variations in treatment timing and heterogeneous treatment effects, we employ the weighted staggered difference-in-differences approach as outlined in recent literature26 and find the estimated effects to remain robust (Table 1 column (2)). The corresponding standard difference-in-differences method with annual mean monthly water consumption as dependent variable are presented in Table 1 column (3). We further conduct robustness checks by (1) altering sample restrictions, such as including extreme observations of utility consumption, relaxing restrictions on building age, and using a fully balanced sample; (2) altering control group by restricting the sample to buildings within the same vicinity of the certified buildings; and (3) altering model specifications by adding weather controls. The effects of Green Mark certification on household and common area water consumption remain statistically significant with minimal changes in effect sizes (Table S1). Moreover, the increase in household water consumption is consistent across households with various flat types, building age, and baseline water demand, as shown in Figure S1 using estimates from Eq. (2) described in Methods.
Electricity consumption: Using monthly block-flat type level data, we find that the Green Mark certification increases household electricity consumption by 1.5%, relative to the households residing in never-certified buildings (Table S2 Panel B Column (1)). The estimate is robust across estimation methods (Table S2 Panel B Column (2)), sample restrictions, control groups, and specifications (Table S3), as well as across sub-groups of households (Figure S2).
Limited by data availability, we are unable to estimate the effects of Green Mark certification on account level household electricity consumption nor the electricity use in common areas. The magnitude of the effects for both, however, is expected to be larger. As shown in Table S2 Panel A, the effects of Green Mark certification on household water consumption using block-flat type level data are smaller than the corresponding estimates using account level data (Table 1).
Evolutionary effect: Fig. 2 demonstrates the differences in water consumption between the treatment and control groups each year, relative to the year of certification (T = 0), as estimated using Eq. (3) in Methods. We find the consumption differences in each of the three years before certification to be statistically insignificant, supporting the parallel trend assumption essential for our difference-in-differences research design.
In Fig. 2(a), we observe elevated household water consumption in the treatment group throughout the post-certification years and the difference increases overtime. Similar pattern is observed in household electricity consumption in Figure S3.
In Fig. 2(b), we observe lower common area water consumption in the treatment group for most of the post-certification years. This difference tends to decline over time, which may be attributed to the reduced effectiveness of the water saving measures implemented.
Mechanism: Overall, the magnitude of the unintended increase in household water consumption is five times of the intended reduction in common areas. We explore whether household-level efficiency change, substitution between common area and household utility, rebound effects of falling utility costs, and wealth effects are driving the unintended increase in utility consumption.
Household-level efficiency change
To obtain Green Mark certification, a minimum of six points must be obtained on water efficiency, including two points on water monitoring, two points on water efficiency improvements plan and two points on common area washing. Additional points can be obtained for common area water efficiency fittings (three points), washing of water tanks (two points), irrigation systems (three points), as well as higher ratings for water monitoring (one additional point) and common area washing (three additional points). All 18 points, however, only applies to common area water fittings and water use. Similarly, for energy efficiency, although more points are allocated and minimal pre-certification performance of building energy use/intensity is required, there is no mandate on efficiency improvements at household-level. As such, the unintended increase in post-certification household utility consumption is likely to be a pure behavioural response to the certification.
Substitution effects
One potential behavioural explanation to the increase in household consumption is the substitution between common area and household utility use post-certification. This explanation, however, does not align with the observed effect magnitudes. If the increase in household water consumption were only to offset the water use in common areas, the magnitude of the increase should be no more than that of the reduction, which is not the case as shown in Table 1 Column (1). The explanation also implies widespread water use in common areas for household purposes before certification, which is implausible given that water utilization in common area is preserved for authorized purposes only. Any unauthorized use is illegal according to the Housing and Development (Common Property and Open Spaces) Rules under the Housing and Development Act. We further examine the existence of this potential behavioural mechanism by comparing the post-certification increase in household water consumption between low-floor units (floors 1–3) with comparatively easier access to common area water fixtures, to that of units situated on higher floors. We find no statistically significant difference (p-value = 0.69) in the effects between the two groups (Figure S1(c)). Based on both qualitative and quantitative evidence, we conclude that the increase in household water consumption post-green building certification is not attributable to a substitution between water sources.
Rebound effects
Upon receiving the Green Mark award, the decal/plaque can be displayed at a prominent location in the certified building. Building owners may also publicize the attainment digitally on their websites and other social media platforms. This information may foster a misconception of improved household water and energy efficiency, potentially triggering a rebound effect due to anticipated lower utility costs. However, such a rebound effect would typically be transient. As households notice an increase in utility bill following the initial behavioural rebound, consumption would generally recalibrate, returning to the pre-certification level quickly. In contrast with this conjecture, our event study results (Fig. 2(a) and Figure S3) show a sustained increase in household utility consumption post-certification. We therefore postulate that the increase in household utility consumption is not due to a perceived reduction of utility costs.
Wealth effects
Green building certification leads to higher rental rates and property values13–17,27,28, resulting in green premium, and the anticipated increases in wealth can stimulate household consumption18,19. A survey conducted in 2016 for the Building and Construction Authority, the organization in charge of the Green Mark certification in Singapore, showed that over 70% of the 610 surveyed public-housing flat owners agreed that green buildings offer better resale value.29 Using resale transaction data for all public-housing flats from 2006 to 2020, we estimate that relative to the non-certified buildings, the resale value of flats in certification building is 1.1% higher post-certification, as compared to flats sold pre-certification in the same building, as shown in Table S4. This green premium lasts at least five years post-certification and increases over time to a magnitude of 6.3% (Figure S4), which is consistent with the magnitude in the literature and aligns with our observation of sustained higher utility consumption post-certification.
To test whether the wealth effect explains the increase in household utility consumption, we first compare the effect of Green Mark certification on homeowners and renters. As the benefits of the perceived increase in property value only accrues to the homeowners, we should not observe any change in water consumption among renters. While pinpointing rental apartments with precision remains challenging, we classify accounts remaining active for a duration not exceeding four years post move-in as those of renters, given the mandatory five-year minimal occupancy period when purchasing an HDB flat. By estimating Eq. (2) in Methods, we show in Fig. 3(a) that only the homeowners—those who occupied the same apartment for five-years or more—increase their water consumption once their building receives the Green Mark certification. Consistent with our hypothesis, the change in renters’ water consumption is minimal (-0.1%) and statistically insignificant (p-value = 0.55). This finding reinforces that the unintended increase in water consumption is unlikely a rebound effect, as we do not anticipate the behavioural rebound to be different for renters who bear the cost of utility bill the same way as the homeowners.
Furthermore, a perceived increase in wealth may lead to a reduction in the price elasticity of consumption goods.30,31 In August 2017, Singapore increased its water price for the first time in 20 years, which led to a significant reduction in water consumption among the households residing in public-housing flats, relative to the private apartments.32 Using a similar empirical method as the previous study, we compare the effect of water price increase on water consumption for households residing in certified versus never-certified HDB buildings, relative to the private apartments. By estimating Eq. (4), we find that the reduction in water consumption in response to the price increase is smaller (1.6%, p-value < 0.01) for certified green buildings, which experience a perceived wealth appreciation, relative to the never-certified buildings (Fig. 3(b)). This reduction in price elasticity of water aligns with our hypothesis that the increase in perceived wealth could be driving the increase in utility consumption.
Lastly, if the increase in utility consumption is driven by an increase in housing price, the effect should disappear when the wealth effect is missing. To test this, we consider two other building categories: (1) existing non-residential buildings and (2) new residential buildings. For existing non-residential buildings, as majority of the occupants and utility users are not property owners, they do not benefit from the increase in property value and are more likely to face higher rent post-certification. By estimating Eq. (1) using a fully balanced sample for all non-agriculture and non-industrial utility accounts, we find a 3.6% reduction of water consumption (Table S5) that lasts up to five years post-certification (Figure S5). For new residential buildings, although homeowners may expect better resale values of their properties, they had to pay a premium to the developers. Using a hedonic valuation approach (Eq. (5) in Methods) and a sample of all residential buildings constructed after 2010, we find no difference in household water nor electricity consumption for certified and never-certified buildings (Table S6). In both building categories, where the certification-led wealth effects are missing, we do not observe an increase in household utility consumption.
In summary, the Green Mark certification, when applied to existing residential buildings, precipitates a persistent rise in household utility consumption, absent any engineering modifications at the household level. We attribute this to a long-term perception of increased housing price. Alarmingly, this uptick in utility usage surpasses the intended consumption reduction derived from the green features situated in common areas. This phenomenon could notably undermine our combat against climate change, especially as a growing number of existing buildings are slated for retrofitting and certification in a bid to mitigate the environmental impacts of building operations.