The amenity value of the British climate

Abstract

This study explores the amenity value of climate to households in Britain. We employ the hedonic technique and use household panel data to derive the marginal willingness to pay for small changes in climate variables. We analyse both the housing and the labour market. Climate is described in terms of heating, cooling degree and rain days. Evidence suggests that it is the housing market that mainly compensates for climate amenities. According to the results, the relationship between property prices and climate has an inverted u-shape. For the sample, about 50 per cent of the locations are close to the preferred levels of heating and cooling degree days while the results for rain days are more varied. Connecting the results to projections of climate change leads to the conclusion that future developments are likely to move a number of households further away from their preferred climate, thus leading to larger regional disparities.

Keywords

Amenity value climate climate change Great Britain hedonic pricing

Introduction

Climate influences humans in various ways affecting, among others, consumption choices such as clothing, housing, heating and nutritional expenditures, as well as recreational possibilities and activities.¹ Changes in climate will affect all of these. However, little is known about the amenity value of climate since the value of services provided by climate is not readily observable as markets for climate services are inexistent. Providing information on the value of climate amenities to households helps to fill this gap and, therefore, contributes to the overall assessment of climate change impacts.

One methodology that suggests itself for valuing climate amenities is the hedonic technique. Hedonic theory suggests that the costs and benefits associated with non-market goods like climate are capitalised into property prices and wage rates (Roback, 1982; Rosen, 1974). This leads to an expectation that households should pay higher property prices for a house with a preferable set of climate amenities and that they might also have to accept lower wage rates. Information on the implicit value that households place on these amenities can, therefore, be obtained by examining the household’s locational choice.

Although a number of hedonic studies have included climate variables for purposes incidental to their main aims, few have deliberately set out to measure the amenity value of climate. The results of these studies indicate that people are willing to pay substantial sums to enjoy more preferred climates. In contrast to this analysis, none of the existing studies explores household panel data.

Based on data from the British Household Panel Survey (BHPS), this paper analyses the amenity value of the British climate. Two other studies have investigated the amenity value of the climate of Britain (Maddison, 2001; Rehdanz, 2006). This study differs from these earlier ones in several ways. First, both Maddison (2001) and Rehdanz (2006) use aggregated data at the regional level, while here individual property price and wage data at a much more disaggregated level are explored. Preferences with respect to amenities differ in a range of observed and unobserved factors that cannot be accounted for using aggregate data. Second, a variety of individual and household specific characteristics are considered. Third, the analysis employs panel data while previous studies relied on cross-sectional data. The panel dataset enables us to account for unobserved heterogeneity that previous studies were unable to account for. Given that individual preferences are not identical, controlling for unobserved heterogeneity is a major improvement. Finally, the analysis is based on a different representation of climate. In the existing studies, climate is most often represented either in terms of annual average temperature and annual average temperature squared or temperature of the hottest and coldest months. However, with the choice of representation, implicit assumptions about the impact of climate change are made. Representing climate by the temperature of the hottest month and coldest month implies that the impact of changes in climate will be independent of the baseline climate; using only annual average temperatures to represent the climate implicitly suggests that individuals are indifferent between climates which might differ substantially in terms of seasonal variation. This paper by contrast describes climate in terms of heating degree days (HDD), cooling degree days (CDD) and rain days (RD) observed at the local authority level.

The remainder of the paper is structured as follows. The next section provides an overview of the hedonic literature on the amenity value of climate. Methodology and data are described in the following section, after which the results of the econometric analysis are presented. The final section discusses and concludes.

Literature review

Different studies have measured the amenity value of climate (see Table A1 for an overview). For the US, Hoch and Drake (1974) found evidence that climate influences wages. As noted above, Roback (1982) made the first attempt to estimate the effects of climate on both wages and house prices. Hoehn et al. (1987) and Blomquist et al. (1988) draw upon the work of Roback but use more detailed data. Other studies include Nordhaus (1996), Cragg and Kahn (1997, 1999) and Kahn (2009). A more recent study by Albouy et al. (2013) explores US Census data to analyse differences across cities in the US in 2000.

One reason why most of the studies focus on the US is that, as Rehdanz and Maddison (2009) note, as climate varies over large distances, the hedonic technique risks violating its assumption of a unified market for housing and labour. Barriers to mobility prevent the net benefits of different locations becoming eliminated. Therefore, a single hedonic function cannot be imposed on separate markets (see Straszheim, 1974). The hedonic technique is thus unsuitable for inter-country analysis.

Hedonic studies focusing on countries in Europe with varying climates are: Maddison and Bigano (2003) who apply the hedonic technique to determine the amenity value of climate to Italian households; Maddison (2001) and Rehdanz (2006) to British; Srinivasan and Stewart (2004) to English and Welsh; Rehdanz and Maddison (2009) to German; and Cavailhès et al. (2012) to French households. These analyses differ in various aspects including the level of data aggregation, the representation of climate, the specification of the dependent variable and the measurement of local amenities.

Most of the above papers for Europe use data at different levels of aggregation (county level for Germany or level of post town for the UK). None of the existing studies however uses panel data at the micro level. Instead, the existing literature is based on cross-sectional analyses.

In Maddison (2001) climate is measured by annual averages. According to this study, British households regard higher temperatures as an amenity and higher precipitation levels as a disamenity. The other three papers mentioned above (Maddison and Bigano, 2003; Rehdanz, 2006; and Rehdanz and Maddison, 2009) account for seasonal differences in climate by taking January and July averages for temperature and precipitation. They follow Cushing (1987) who analyses different configurations of temperature to determine what fits best in influencing population migration decisions. They find high January precipitation levels to be a significant disamenity. However, in Maddison and Bigano (2003) January temperatures have no significant effect on welfare in Italy, and Rehdanz (2006) as well as Rehdanz and Maddison (2009) find that British and German households prefer warmer Januarys. High July temperatures are a significant disamenity in Italy and in Germany. July precipitation plays no significant role.

A further noticeable difference is that Maddison and Bigano (2003) and Rehdanz (2006) derive a single dependent variable given by household income net of taxes and housing costs. Maddison (2001) and Rehdanz and Maddison (2009) estimate house price and wage rate regressions separately.

Also, the use of GIS data varies between studies. Some capture geographical information by the use of dummy variables (e.g. Maddison and Bigano, 2003) while others include measures of distance (e.g. Rehdanz and Maddison, 2009).

Methodology and data

General approach and empirical strategy

Our empirical approach follows the theoretical work of Roback (1982) in exploring the effects of climate on property prices and wage rates. According to her theory, individuals maximise their utility as a function of some consumption goods, property, as well as amenities subject to a budget constraint. An individual’s indirect utility function can then be derived and the value of an amenity to individual or households be measured.

The critical assumptions of the approach are well known (e.g. Palmquist 1991; Freeman et al. 2014). Hedonic studies are based on the assumption of equilibrium in wages and rents, leaving individuals indifferent across locations. However, the existence of equilibria requires perfect information, zero moving and transaction costs and rapidly adjusting prices. Although these are rather unrealistic assumptions, to the extent that disequilibrium is present, a correlation with the level of particular amenities in the model is unlikely. While a divergence from a full equilibrium is possible, it is unlikely to bias hedonic regressions (Palmquist, 1991).

We explore the effect of climate on property prices and wage rates applying fixed effects models, taking into account the unobservable and non-measurable effects of all the different household or individual units, thereby addressing heterogeneity between households or individuals. In the context of this study, unobservable household or individual characteristics might include differences in attitudes such as awareness of landscape features in the house price regression or levels of motivation in the wage regression.²

To estimate the effect of climate, first, the effects of different building, neighbourhood, geographical and climate characteristics as well as time on property prices are estimated using the following regression equation:

\begin{matrix} \ln H_{ijkt} = & β_{0} + β_{1} B_{i} + β_{2} N_{i} + β_{3} L_{j} + β_{4} C_{j} \\ + β_{5} R_{k} + β_{6} Y_{t} + ν_{i} + ε_{ijkt} \end{matrix}

(1)

The dependent variable, H_ijkt, is the price of property i a homeowner living in district j belonging to region k expects to get at the time of the interview (t for the survey year), measured as property price per room. Explanatory variables are captured in vectors B_i and N_i which contain building and neighbourhood characteristics. L_j covers geographical information. C_j represents the climate variables. R_k captures dummy variables for each region k to control for region-specific effects. Y_t captures the year t of the survey while $ν_{i}$ is the household specific fixed effect and captures unobserved heterogeneity between households. ε represents the error term.

Next, the effects of different employee, employer, geographical and climate characteristics as well as time on wage rates are given by the following regression equation:

\begin{matrix} \ln W_{ijkt} = & β_{0} + β_{1} E_{i} + β_{2} I_{i} + β_{3} L_{j} + β_{4} C_{j} \\ + β_{5} S_{k} + β_{6} Y_{t} + ν_{i} + ε_{ijkt} \end{matrix}

(2)

The dependent variable, W_ijkt, is the net hourly wage of individual i living in district j belonging to region k in a particular year t. Explanatory variables are captured in vectors E_i and I_i which contain employee and employer characteristics. L_j is a vector of geographical characteristics. The remaining vectors and terms are as in equation (1).

The ultimate goal of an analysis of the amenity value of climate is to calculate the marginal willingness to pay (MWTP) for climate. The MWTP for climate is equal to the first derivative of the hedonic price function. Households will demand more of a particular climate until the cost of doing so is equal to the MWTP for it.

Data

The analysis is largely based on the BHPS. The BHPS contains data on more than 5000 households and 10,000 individuals that have been re-interviewed annually since 1991. The unbalanced panel provides annual information on housing and occupation, employment history and earnings of individuals. In addition to a fixed set of core questions, in each year the survey focuses on a special topic. As a consequence, the 1998, 2003 and 2008 surveys include detailed information on neighbourhood characteristics. In order to take advantage of this information, the analysis relies exclusively on these survey years. We further include information on area of living (local authority). In total, the analysis covers 262 different local authority districts. These areas are sufficiently small to assume that individuals living within a district enjoy the same climate.³ Each district is assigned to one of the 18 regions (16 English regions, Scotland and Wales) of Britain.

The MetOffice⁴ provides information on HDD and CDD. CDD and HDD are measured as annual totals of the day-by-day sum of the mean number of degrees by which the air temperature is more than a value of 22°C or less than a value of 15.5°C, respectively. Rain days (RD), also provided by the MetOffice, are measured as the number of days with at least 10mm precipitation. This information is available on a 5km grid level. We used GIS to match the data to the respective district to calculate the average number of degree and rain days by local authority district. Next, we calculated 10-year averages to smoothen the effect of individual years being particularly hot, cold or wet, referring to the 10 years prior to the year of the survey. For the 1998 survey, for example, we use the period 1988–1997.

According to the data, the most HDDs were reported for the year 1996 (2563) and the fewest for 2007 (1998). The fewest CDDs were observed in 1988 (six) and the most in 1995 (57) while 2003 was the driest year (18 RDs) and 2000 the wettest (33 RDs). At the regional level, the data clearly reflects an expected north–south decline in HDDs and a north–south increase in CDDs. Over the period 1988–1997, Scotland had on average the most HDDs (maximum of 2961 in 1993) and the fewest CDDs (minimum of five in 1988)), whereas England had the fewest HDDs (minimum of 1948 in 2007) and the most CDDs (maximum of 61 in 1995). Wales, on average, experienced the highest number of RDs (maximum of 60 in 2000). Figure A1 in the Appendix displays the variation of HDDs, CDDs and RDs over time and across regions.

The hedonic analysis is based on two sub-samples: a house price sample and a wage sample. The description of all variables used can be found in Table A2 in the Appendix. Summary statistics of the two sub-samples are presented in Tables A3 and A4, respectively.

Starting with the house price sample, homeowners were asked for a number of housing characteristics including the building type (HOUSE, SHOUSE, ETHOUSE, THOUSE and FLAT), number of rooms (ROOMS), availability of central heating (CENT HEATING), condensation problems (CONDENSATION), leaky roofs and windows (LEAKY ROOF), damp walls (DAMP) and rot in windows (ROT).

Further, homeowners were asked what price they expect to receive for selling their property at the time of the interview. Information on actual selling prices would have been preferred but are, unfortunately, unavailable. Given that a fixed effects approach is applied it is assumed that possible biases due to individual over-/undervaluation of homeowners of the actual price of their property on the market are covered within the unobserved heterogeneity.

Neighbourhood specific information is covered for by the suitability of raising children (CHILDREN), local standards of schools (SCHOOL), medical facilities (MEDICAL), public transport services (TRANSPORT), shopping (SHOPPING) and leisure facilities (LEISURE). Controls are included for the type of town or city in which the property is located (URBAN AREA, TOWN and VILLAGE). This information is taken from the BHPS and collected at the household level.

Alongside the climate variables (HDD, CDD and RD), a number of further controls are included at the level of the local authority district. These cover information on whether there are rivers (RIVER) and lakes (LAKE), the percentage share of urban areas (URBAN), small urban areas (SMALL URBAN) and woodland areas (WOODLAND), the distance to London (LONDON) as well as to the coast (COAST).⁵ All this data is generated using GIS. Also included is information on the proportion of the resident population aged 16–64 claiming unemployment-related benefits in September of the respective year (CC RATE).⁶ Further, we control for the number of people per square km (PPSQKM).⁷ In order to capture any differences in fiscal structure and the provision of public goods, a separate dummy variable is included for the 18 regions of Britain. Further dummy variables describe the year of the survey.

Turning to the wage sample, an individual’s net hourly wage (WAGE) is derived based on information on the usual net pay per month in the current job as well as the number of hours normally worked per week. An individual’s work situation is characterised by the type of occupation (PROFESSIONAL, SKILLED or UNSKILLED), position (PERMANENT) and terms (FULL TIME) as well as employment (EMPLOYED or SELF EMPLOYED). The workplace is described by its size, i.e. the number of employees at the workplace (EMPL0-EMPL7) and the organisation of workers (TRADE UNION). All this information is taken from the BHPS and was collected at the individual level.

Econometric analysis

Main results

In line with earlier studies, log-linear specifications and fixed effects models to account for unobserved heterogeneity at the household and individual level are applied.⁸ As described above, the study is based on separate regressions for house prices and wage. In total, the samples cover 5910 observations in the house price regression and 13,232 observations in the wage regression. In Table 1 the results for the climate variables are presented. For full regression results see Table A5 in the Appendix.

Table 1.

Regression results.

House Price Regression^[a]		Wage Regression^[b]
Variables	Coefficient	Variables	Coefficient
HDD	0.004***	HDD	−0.0003
CDD	0.0118	CDD	0.0131
RD	0.0168**	RD	0.0073
HDD SQ	−7.72E-07***	HDD SQ	0
CDD SQ	−0.0007***	CDD SQ	−0.0001
RD SQ	−0.0002**	RD SQ	−0.0001
Further controls		Further controls
Building characteristics	YES	Employee characteristics	YES
Neighbourhood characteristics	YES	Employer characteristics	YES
Geographic characteristics	YES	Geographic characteristics	YES
Time trend	YES	Time trend	YES
Observations	5910	Observations	13,232
R-squared (within)	0.7949	R-squared (within)	0.2474
Number of households	3876	Number of individuals	8107
F-tests		F-tests
HDD and HDD SQ	F(2,1986) = 21.49 (Prob>F = 0.000)	HDD and HDD SQ	F(2,5077) = 0.42 (Prob>F = 0.658)
CDD and CDD SQ	F(2,1986) = 23.27 (Prob>F = 0.000)	CDD and CDD SQ	F(2,5077) = 4.50 (Prob>F = 0.011)
RD and RD SQ	F(2,1986) = 2.33 (Prob>F = 0.097)	RD and RD SQ	F(2,5077) = 1.12 (Prob>F = 0.328)

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1. Own calculations based on the BHPS data. Method: fixed-effects regression. [a] Dependent variable: Log of the price of the property (GBP per room) in real terms. [b] Dependent variable: Log of the net hourly wage in pennies, in real terms.

House price regression

Discussing the results of the house price regression first, the dependent variable is the logarithm of the property price in GBP per room.⁹ The property price per room as a single measurement unit facilitates the comparison between properties since property size has an important impact on property prices. The property price per square metre would be preferable, since rooms may vary in size. However, data on property size in square metres are rarely collected in the UK.¹⁰

We experimented with a variety of concepts for describing the climate of a district including annual averages of HDD, CDD and RD; annual averages and annual averages squared; maximum and minimum values; averages and ranges; and means and variances. Evidence suggests that the best description of climate in both the hedonic wage and house price regression in terms of the R-squared within was provided by the use of annual averages and annual averages squared of CDD, HDD and RD. As shown in Table 1, five of the six climate variables are individually significant at least at the 5 per cent level of statistical significance. CDD and HDD are also jointly significant at the 1 per cent level and RD at the 10 per cent level of statistical significance. The sign of the coefficients indicate that more HDDs are an amenity for those with few HDDs.

Reporting on the variables not displayed in Table 1 (but included in Table A5 in the Appendix), a number of building characteristics are significant. Compared to the property type HOUSE, property prices for all other types of property are significantly less expensive, with FLAT being the least expensive. The logarithm of ROOM has a negative effect indicating that the more rooms a property has, the lower the price per room. Properties equipped with CENT HEATING are worth more, as heating in such a property is more efficient in contrast to other heating systems. Properties with DAMP walls indicating insulation problems are worth less. Other property characteristics, in particular CONDENSTATION, LEAKY ROOF and ROT, are statistically not significant. At the neighbourhood level, it matters if the neighbourhood is perceived as suitable for raising children (CHILDREN) or good for LEISURE. In both cases, property prices are significantly higher as they are simply more attractive. None of the other neighbourhood variables (e.g. TRANSPORT, SCHOOL or SHOPPING) are statistically significant.

Turning to the geographic information, a higher rate of unemployment (CCRATE) in the local authority district attenuates property prices. Here, the attractiveness of districts is lower as the economic performance is comparatively lower and offers fewer employment opportunities. Higher prices are obtained for properties in districts that are more populated (PPSQKM) as well as those that are closer to LONDON. Somewhat unexpectedly, the results suggest that property prices per room in a VILLAGE are significantly more expensive compared to URBAN AREA. This can be explained by the fact that in the sample a large number of VILLAGE observations belong to the Inner and Outer London regions; about 70 per cent. For this reason, it is plausible to assume that VILLAGE captures regional amenities we are unable to capture with the other regional and geographical characteristics. The time dummies (Y2003 and Y2008) are highly significant, pointing to the fact that house prices in Great Britain increased steadily over the sample period.

Wage regression

Turning to the wage regression,¹¹ the dependent variable is the logarithm of the net hourly wage in pennies.¹² The estimates for the climate variables suggest that there is no consistent effect of climate on wage rates. The variables are jointly and individually insignificant, only CDD is significant at the 10 per cent level and CDD and CDD SQ are jointly significant at the 5 per cent level of statistical significance. The results for CDD indicate an inverted u-shaped link between wage rates and CDD with a turning point at 65.5 CDD which is at the higher end of CDD. In general, the results with respect to the climate variables are as anticipated given the situation of national wage bargaining. In addition, if a company’s production costs are not affected by the level of the environmental amenity and firms mainly employ labour rather than land, wages are not affected by the level of the amenity (Roback, 1982).

Reporting on the variables not displayed in Table 1 (but included in Table A5 in the Appendix), a number of employee characteristics are significant; SKILLED and UNSKILLED workers earn significantly less per hour than PROFESSIONALs as well as those with full-time employment (FULL TIME). Regarding employers’ characteristics, working at a workplace with more employees (EMPL0-EMPL7) or a trade union (TRADE) increases net hourly wages significantly. A few of the geographical characteristics are significant as well. Workers from a district with a RIVER (LAKE) in the district earn higher (lower) wages per hour. A higher percentage of WOODLAND area in the district increases wages as well. The time dummies (Y2003 and Y2008) are highly significant pointing to a steady increase in wages over time.

Robustness

In order to check for robustness of the results, different specifications of the fixed effects model were estimated. Estimated coefficients of the house price model or the wage model without any climate variables do have the same signs, magnitudes and significance as in the above described results (results not shown). According to the likelihood ratio test, including the climate variables creates a significant improvement in the fit of the two models.¹³

As pointed out above, a reasonable analysis on the amenity value of climate is only possible for regions where climate varies. However, the hedonic technique risks violating its assumption of a unified market for housing and labour if large regions are considered. For this reason, we subdivided our sample into regions of Britain (England, Scotland and Wales). The results suggest that the climate signal is much lower, which is not surprising given the much smaller area (results not shown). Judged on the results of the F-test for joint significance of climate variables, CDDs are significant in England, HDDs in Scotland and RDs in Wales in the house price regression. In the wage regression, HDDs in Scotland are jointly significant, but only at the 10 per cent level of significance. The results of the other variables are similar to those presented in Table A5.

The preferred climate

Based on the results in Table 1, it is possible to characterise the climate that households in Britain prefer. Figure A2 in the Appendix presents the house price as a function of the different climate variables based on the estimation results.¹⁴ It depicts a household’s WTP (WTA) in terms of higher (lower) property prices per room for climate amenities (disamenties). More specifically, property prices per room are c.p. highest in locations with about 2011 HDDs, eight CDDs and 40 RDs per year. Deviations from these values c.p. correspond to lower property prices per room.¹⁵

Figure 1 displays the deviation of a district’s current climate, measured as the 10-year average 1998–2007, from the preferred climate. Here, the preferred climate is the climate that commands the highest property price per room. According to Figure 1, the actual climate in Britain deviates from the preferred for many regions. For Britain as a whole, about 65 per cent of the local authority districts are characterised by HDDs higher than preferred. For CDDs, about 25 per cent of the locations are close to the optimal level. About 80 per cent of the districts have fewer RDs than preferred. At the regional level, northern Britain is generally colder and western parts tend to be wetter than preferred. In the south, HDDs are close to the preferred level but, especially in districts close to London, CDDs are much higher than the preferred level.

Figure 1.

Deviations of a district’s average climate from the preferred climate.

Marginal willingness to pay

Using the results of Table 1 we calculate the MWTP for climate. Evaluated at the sample mean of 10-year average HDDs and CDDs, the MWTP for an extra HDD or an extra CDD is negative (see Table 2). In local authorities with very low (high) values of either of the two the MWTP for an extra HDD or CDD is positive (negative). The MWTP for an extra RD is positive at the sample mean. More extreme values of rainfall though lead to different parameter values. MWTP is positive (negative) in local authority districts with very low (high) rainfall.

Table 2.

MWTP of the average British household and MWTP of households in selected local authorities.

Climate Variable	Value	Region	BHPS Wave	House Price (in £)	MWTP (in £)
HDD
average	2298	GB	all	30,272	−17.35
highest	3142	Perth & Kinross (Scotland)	2nd wave	23,620	−53.42
lowest	1635	Weymouth & Portland (South West)	3rd wave	76,805	57.86
CDD
average	19	GB	all	30,272	−478.1
highest	46	Hillingdon (London)	3rd wave	71,429	−3892.20
lowest	2	Aberdeen City (Scotland)	2nd wave	26,236	230.91
RD
average	30	GB	all	30,272	135.46
highest	73	Argyll and Bute (Scotland)	3rd wave	32,258	−432.93
lowest	11	Stockton-on-Tees (Rest of North)	1st wave	24,759	299.85

Notes: MWTP is calculated for the climate variables at the sample mean. Extreme values for climate variables are the highest or lowest 10-year mean values.

The calculation is, however, based on several assumptions (see also above). Since we analyse data for a 10-year period (1998 to 2007), we implicitly assume that a household’s preferences do not change over this period and that the housing stock is fixed. However, when excluding the few households from our analysis that move over the sample period (about 4 per cent of the total sample), results are robust.

Acknowledging the limitations of the approach, the results allow for the conclusion that British households in general would prefer fewer HDDs and CDDs, implying a preference for milder temperatures and fewer extreme weather conditions but not without limits. In districts with few HDDs and CDDs, more HDDs and CDDs would be preferred.

Previous studies show that Italian households dislike high temperatures in July (Maddison and Bigano, 2003) while households in parts of Britain would be willing to pay for higher temperatures in the same month (Rehdanz, 2006). Maddison and Bigano (2003) as well as Rehdanz (2006) also show that high levels of precipitation in January are a disamenity. According to Srinivasan and Stewart (2004) households’ MWTP in England and Wales is positive for a larger number of hours with sunshine. Maddison (2001) calculates a MWTP of £298 for a one-degree increase in the annual temperature in Britain in 1994. Our results present a more differentiated picture given the level of disaggregation and the differentiation of CDD and HDD, but the MWTP for more CDDs in very cold areas especially is comparable in magnitudes (see Table 2).

Discussion and conclusion

This study has illustrated the extent to which British households’ preferences for climate amenities are capitalised into wage rates and property prices. In contrast to earlier studies implicitly suggesting that individuals are indifferent between climates though seasonal variation might be high, climate is described in terms of HDDs and CDDs. These measure daily deviations from a base mean temperature of 15.5°C and 22°C, respectively. The base temperature is intended to approximate the outside temperature where households need neither heating nor cooling to feel comfortable indoors. RDs are introduced to control for precipitation. The climate variables on the local authority district are combined with panel data of the BHPS and information on the area of living.

For the wage regression results are insignificant. It is the housing market that mainly compensates for climate amenities. In her theoretical model, Roback (1982) shows that it is decisive if an amenity is an input to firms’ activities besides being attractive to households. If firms’ activities are independent from the occurrence of amenities, these only impact on the MWTP in the housing market. On the contrary, if amenities are employed as input factors, they will also drive job markets. Given the nature of the British economy, however, the majority of the economic performance is independent of climate if agriculture is ignored. Therefore, the empirical results are in line with Roback’s theoretical model.

Estimates from the hedonic house price regression point to an inverted u-shaped relationship between property prices and climate. Households pay a significant premium for living in areas offering a particular climate characterised by 2012 HDDs, eight CDDs and 40 RDs. For the sample, about 50% of the locations are close to the preferred level of HDDs and CDDs while the results for RDs are more varied. Here a number of districts exist offering extremes, in both directions. Scientists predict a decrease in HDDs in future (Defra, 2012). In comparison to 1961–1990 levels, by 2080 the number of HDDs will decline by 30 per cent in Scotland and 50 per cent in England. On the contrary, CDDs are expected to increase significantly. With respect to the same baseline, changes in CDDs by 2080 are projected to increase in Southern England (up to 150 per cent) while only moderate increases are projected for northern England and Scotland. With respect to rainfall, projections suggest an increase in heavy rainfalls particularly occurring during winter months (Defra, 2012). Connecting these projections to the results leads to the conclusion that future developments are likely to move a number of households further away from their preferred climate.

Footnotes

Appendix

Table A5.

Regression results of House Price and Wage Regressions.

House Price Regression^[a]			Wage Regression^[b]
VARIABLES	Coefficient	Std. Error	VARIABLES	Coefficient	Std. Error
Building characteristics			Employee and employer characteristics
HOUSE	reference category		PROFESSIONAL	reference category
SHOUSE	−0.1726***	0.0243	SKILLED	−0.0959***	0.0125
ETHOUSE	−0.3054***	0.0349	UNSKILLED	−0.122***	0.0169
THOUSE	−0.3565***	0.0348	PERMANENT	0.0083	0.0218
FLAT	−0.4078***	0.0467	FULL TIME	−0.0732***	0.0136
ROOMS	−0.7719***	0.0356	SELF EMPLOYED	−0.02	0.0593
CENT HEATING	0.0552*	0.0285	EMPL0	reference category
CONDENSATION	0.0035	0.0201	EMPL1	0.0305**	0.0149
LEAKY ROOF	0.0295	0.0292	EMPL2	0.0479***	0.0159
DAMP	−0.0632**	0.0266	EMPL3	0.0949***	0.017
ROT	0.0311	0.0239	EMPL4	0.0718***	0.0176
			EMPL5	0.0935***	0.0175
Neighbourhood characteristics			EMPL6	0.0833***	0.0208
TRANSPORT	−0.0134	0.0118	EMPL7	0.08***	0.0194
CHILDREN	0.0236*	0.0137	TRADE UNION	0.042***	0.0117
SCHOOL	0.0169	0.0127
MEDICAL	0.0059	0.0118
SHOPPING	0.0039	0.0119
LEISURE	0.0218*	0.0112
Climate characteristics			Climate characteristics
HDD	0.004***	0.0006	HDD	−0.0003	0.0004
CDD	0.0118	0.0093	CDD	0.0131*	0.007
RD	0.0168**	0.0078	RD	0.0073	0.0052
HDD SQ	−1.00E-06***	1.19E-07	HDD SQ	6.45E-08	8.89E-08
CDD SQ	−0.0007***	0.0002	CDD SQ	−0.0001	0.0001
RD SQ	−0.0002**	0.0001	RD SQ	−0.0001	0.0001
Geographic characteristics			Geographic characteristics
URBAN AREA	reference category		URBAN AREA	reference category
TOWN	0.0118	0.0464	TOWN	0.0111	0.0252
VILLAGE	0.1221***	0.0455	VILLAGE	0.0183	0.0237
URBAN	−0.0028	0.0026	URBAN	0.001	0.0013
SMALL URBAN	−0.0255	0.031	SMALL URBAN	−0.0165	0.0143
CC RATE	−0.0376***	0.0119	CC RATE	−0.0063	0.0093
PPSQKM	3.62E-05***	8.58E-06	PPSQKM	1.00E-05	7.00E-06
RIVER	0.063	0.094	RIVER	0.0922**	0.0432
LAKE	−0.0679	0.1004	LAKE	−0.1123**	0.0474
WOODLAND	−0.0069	0.0068	WOODLAND	0.0046*	0.0028
LONDON	29.2836***	5.116	LONDON	2.3294	3.7319
COAST	0.0093	0.0078	COAST	0.0032	0.0073
Time trend			Time trend
Y 1998	reference category		Y 1998	reference category
Y 2003	0.4586***	0.0234	Y 2003	0.192***	0.0173
Y 2008	0.7666***	0.0375	Y 2008	0.2996***	0.0247
Further controls			Further controls
REGION	YES		REGION	YES
Constant	5.9144***	0.9134	Constant	6.3081***	0.6091
Observations		5,910	Observations	13,232
R-squared within		0.7949	R-squared within	0.2474
Number of households		3,876	Number of households	8107
F-tests			F-tests
HDD and HDD SQ		F(2,1986) = 21.49 (Prob>F = 0.00)	HDD and HDD SQ	F(2,5077) = 0.42 (Prob>F = 0.6584)
CDD and CDD SQ		F(2,1986) = 23.27 (Prob>F = 0.00)	CDD and CDD SQ	F(2,5077) = 4.50 (Prob>F = 0.0112)
RD and RD SQ		F(2,1986) = 2.33 (Prob>F = 0.0973)	RD and RD SQ	F(2,5077) = 1.12 (Prob>F = 0.3277)

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1. Own calculations based on the BHPS data. Method: fixed-effects regression.[a] Dependent variable: Log of the price of the property (£ per room) in real terms. [b] Dependent variable: Log of the net hourly wage in pennies, in real terms.

Acknowledgements

We would like to thank three anonymous referees for their helpful and constructive comments.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Albouy

Graf

Kellogg

. (2013) Climate amenities, climate change, and American quality of life. NBER Working Paper 18925.

Blomquist

Berger

Hoehn

(1988) New estimates of quality of life in urban areas. American Economic Review 78: 89–107.

Cavailhès

Joly

Cardot

. (2012) The price of climate: French consumer preferences reveal spatial and individual inequalities. In: Hoornweg

Freire

Lee

(eds) Cities and Climate Change: Responding to an Urgent Agenda. Volume 2. Washington, DC: The World Bank, pp. 1–23.

Cragg

Kahn

(1997) New estimates of climate demand: Evidence from location choice. Journal of Urban Economics 42: 261–284.

Cragg

Kahn

(1999) Climate consumption and climate pricing from 1940 to 1990. Regional Science and Urban Econonomics 29: 519–539.

Cushing

(1987) A note on specification of climate variables in models of population migration. Journal of Regional Science 27: 641–649.

Defra (2012) The UK Climate Change Risk Assessment 2012 Evidence Report. Project deliverable number D.4.2.1, Release 8, January. Department for Environment, Food & Rural Affairs. Available at: www.gov.uk/government/organisations/department-for-environment-food-rural-affairs.

Englin

(1996) Estimating the amenity value of rainfall. The Annals of Regional Science 30(3): 273–283.

Freemann

III Herriges

Kling

(2014) The Measurement of Environmental and Resource Values: Theory and Methods, 3rd Edition. Oxford: Routledge.

10.

Gabriel

Mattey

Wascher

(2003) Compensating differentials and evolution in the quality-of-life among US states. Regional Science and Urban Economics 33(5): 619–649.

11.

Gyourko

Tracy

(1991) The structure of local public finance and the quality of life. Journal of Political Economy 99(4): 774–806.

12.

Hoch

Drake

(1974) Wages, climate, and the quality of life. Journal of Environmental Economics and Management 1: 268–295.

13.

Hoehn

Berger

Blomquist

(1987) A hedonic model of interregional wages, rents, and amenity values. Journal of Regional Science 27: 605–620.

14.

Kahn

(2009) Urban growth and climate change. Annual Review of Resource Economics 1: 333–350.

15.

Maddison

(2001) The Amenity Value of the Global Climate. London: Earthscan.

16.

Maddison

Bigano

(2003) The amenity value of the Italian climate. Journal of Environmental Economics and Management 45: 319–332.

17.

Nordhaus

(1996) Climate amenities and global warming. In: Nakicenovic

Nordhaus

Richels

al.

(eds) Climate Change: Integrating Science, Economics, and Policy. Laxenburg: IIASA, pp. 3–45.

18.

Palmquist

(1991) Hedonic methods. In: Braden

Kolstad

(eds) Measuring the Demand for Environmental Quality. Amsterdam: Elsevier, pp. 7–120.

19.

Parker

(1995) Climatic Effects on Individual Social and Economic Behaviour: A Physioeconomic Review of Research Across Disciplines. Westport, CT: Greenwood Press.

20.

Rehdanz

(2006) Hedonic pricing of climate change impacts to households in Great Britain. Climatic Change 74: 413–434.

21.

Rehdanz

Maddison

(2009) The amenity value of climate to households in Germany. Oxford Economic Papers 61: 150–167.

22.

Roback

(1982) Wage, rents and the quality of life. Journal of Political Economy 90: 1257–1278.

23.

Rosen

(1974) Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy 82: 34–55.

24.

Sinha

Cropper

(2013) The value of climate amenities: Evidence from US migration decisions. National Bureau of Economic Research, Working Paper 18756. Available at: www.nber.org/papers/w18756.

25.

Srinivasan

Stewart

(2004) The quality of life in England and Wales. Oxford Bulleting of Economics and Statistics 66(1): 1–22.

26.

Straszheim

(1974) Racial discrimination in the urban housing market and its effect on black housing consumption. In: von Furstenburg

Harrison

Horowitz

(eds) Patterns of Racial Discrimination. Vol. 1. Lexington, MA: D.C. Heath, pp. 139–164.

27.

Wooldridge

(2009) Introductory Econometrics: A Modern Approach. 4th edition. Mason, OH: South Western.