OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 54,2(1992) 0305-9049 S3.00 THE DEMAND FOR CIDER IN THE UNITED KINGDOM* David Blake and Sean Boyle I. INTRODUCTION Over the last thirty-five years, there has been a steady expansion in the total consumption of alcoholic drink in the UK, and in particular in the market for long drinks, which covers cider, beer and wine. Table 1 gives consumption in pints per head over the period 1952-1986, and we observe growth in the demands for cider, beer and wine of 150, 25 and 700 percent respectively (see also Figures 1, 2 and 3). So cider, although it still represents only a small proportion of the total market, has been making inroads into the long drinks industry. A glance at the table reveals that this expansion began in the mid-i 96 Os. For a long time cider was thought of as a cheap drink, consumed mainly by men on low incomes. In the 1 950s and 1 960s, many public houses removed draught cider in order to encourage a better class of customer', and we find the number of keg points, i.e. on-licence suppliers of draught cider, falling. The consumption TABLE 1 Consumption of Long Drinks in the (1K. ('Pints per head per annum,) 1952-1986 *We are extremely grateful for the comments of both the Editor and an anonymous referee. 73 Cider Beer Wine 1952 4.5 187.3 2.8 1957 4.1 185.0 4.0 1962 3.6 199.2 5.8 1967 4.8 217.2 8.5 1972 6.0 249.1 12.3 1977 8.2 270.4 15.1 1982 10.7 239.6 17.2 1986 11.4 234.3 22.6
Transcript
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 54,2(1992)0305-9049 S3.00
THE DEMAND FOR CIDER IN THE UNITEDKINGDOM*
David Blake and Sean Boyle
I. INTRODUCTION
Over the last thirty-five years, there has been a steady expansion in the totalconsumption of alcoholic drink in the UK, and in particular in the market forlong drinks, which covers cider, beer and wine. Table 1 gives consumption inpints per head over the period 1952-1986, and we observe growth in thedemands for cider, beer and wine of 150, 25 and 700 percent respectively(see also Figures 1, 2 and 3).
So cider, although it still represents only a small proportion of the totalmarket, has been making inroads into the long drinks industry. A glance atthe table reveals that this expansion began in the mid-i 96 Os. For a long timecider was thought of as a cheap drink, consumed mainly by men on lowincomes. In the 1 950s and 1 960s, many public houses removed draught ciderin order to encourage a better class of customer', and we find the number ofkeg points, i.e. on-licence suppliers of draught cider, falling. The consumption
TABLE 1Consumption of Long Drinks in the (1K.
('Pints per head per annum,)1952-1986
*We are extremely grateful for the comments of both the Editor and an anonymous referee.
. ,...,...........i... ICCC CD CD O CCI CD CD O ('JIt) CD CI) CD CD CO CD CD CD N N NOC OC OC OC OC OC OC O) OC OC OC OC
Fig. 3. Wine consumption in the UK, 1952-1986
of cider was associated with drunkenness, and indeed dry cider has a muchhigher alcohol content than most beers.
The percentage of total cider consumption attributable to draught ciderwas falling during the period up to 1974, when it accounted for only 15 per-cent of total cider consumption. However the major cider producers (H. P.Bulmer, Taunton Cider and Showerings)' identified the potential for growthin marketing cider as a young person's drink - rather as lager has beenmarketed so successfully. It was already fairly common for a young person'sfirst experience of alcohol to be a bottle of cider from the local supermarketor off-licence; now the objective was to encourage younger drinkers to askfor draught cider in public houses. An advertising campaign was designed toestablish cider as the regular drink of trendy young people - no longer thetipple of winos and 14 year olds.
Since the mid-1960s, there has been a turn around in the cider market,with both growth in overall consumption and also a switch from packaged(i.e. bottled and canned) to draught cider; by 1986 draught consumptionaccounted for 45 percent of the total cider market.
However, since the imposition of higher excise duties on cider in the 1984and 1985 Budgets there has been a curtailment in the growth of the cidermarket. The current trend towards non-alcohol/low-in-alcohol drinks has
111'1 ICD CD ON N CDOC OC OC
CCI CDCD CD CDOC OC OC
'By 1985, H.P. Bulmer, Taunton Cider and Showerings accounted for 48, 29 and 18 percentof the market respectively.
75THE DEMAND FOR CIDER IN THE UNITED KINGDOM
76 BULLETIN
meant that firms such as H. P. Bulmer have had to look elsewhere for growthpotential, such as Perrier sales, over which they hold exclusive distributionrights in the U.K. (Bulmer Annual Report 1987).
In this paper, we are concerned with estimating demand for cider equa-tions based on overall U. K. consumption of cider during the period1952-1986. We estimate three equations - total cider consumption, share ofcider in the long drinks market, and packaged cider consumption - in orderto investigate the change in attitude towards cider.
II. SPECIFICATION OF THE DEMAND EQUATIONS
There has been a number of econometric studies of the demand for alcohol,usually at highly aggregated levels (McGuinness (1980, 1982), Walsh (1982),Duffy (1982a, 1982b, 1982c, 1983), Salvananthan (1988), Jones (1989)).Only in the last four papers do we find an attempt to estimate from disaggre-gated data. Duffy, for example, finds that both relative prices and advertisinghave significant effects on the demand for beers, wines and spirits, though theeffects of advertising pale by comparison with those of price effects.
We employ the following specification:
Q=f(P,A, Y,Z),
where Q is quantity of cider consumed, P is a vector of prices, A is a vector ofadvertising expenditures, Y is a vector of income variables, and Z is a vectorof other relevant variables. Included in Z are the number of on- and off-licences serving cider, the percentage of the population. who are unemployed(as a measure of demand shift when people lose their jobs; since unemploy-ment benefit is a transfer payment, the effect that unemployment has on anindividual's purchasing power is not captured in any of the income variablesthat we use); weather variables such as annual mean daily sunshine, annualmean daily rainfall, mean daily air temperature at sea level, annually, in Mayand June, and between June and September; social class variables such as theproportion of the population in social classes A, B and Cl, and the propor-tion of males working in the production sector (as a measure of the workingclass demand for cider); demographic variables such as the percentage of the'population living in the south-west of England, and the number of males andfemales between the ages 15 and 24. Alternative income variables were con-sidered including real G.D.P., real personal disposable income (P.D.L), totalreal consumers' expenditure, and real consumers' expenditure on food, drinkand tobacco.
Our estimation method is ordinary least squares (OLS). Given the simul-taneous determination of prices and quantities, it is, in general, more appro-priate to use instrumental variables (IV). However, in our case, the samplesize is small and the number of regressors quite large, so the number ofdegrees of freedom is low. Since this will induce large finite sample biases and
THE DEMAND FOR CIDER IN THE UNITED KINGDOM 77
large standard errors, IV is not obviously superior to OLS in this case. There-fore we present only OLS estimates.
All equations were estimated in log-linear form, so that the coefficients areestimates of elasticities. We included lagged dependent variables up to thethird order, lagged independent variables up to the second order, and someintercept and slope dummies and a time trend (which will be discussed later).
The model that we estimate is much more richly specified than otherstudies in this field, including that of Jones (1989), the only other investigatorto estimate a cider demand equation. Nevertheless, the inclusion of non-economic' variables such as advertising, climatic, social class and demo-graphic variables is a common feature of micro-economic demand studies inother fields.
The hypotheses that we are interested in investigating are
to what extent are economic variables such as prices and incomesignificant in explaining the demand for cider?is cider a habit-forming beverage?to what extent is advertising important?how does weather affect demand?
(y) does the demand for cider depend on the number of outlets offeringcider?to what extent does cider demand shift when people becomeunemployed?to what extent does social class affect the demand for cider?to what extent has cider become a young person's drink?do large changes in price lead to large changes in the price elasticityof demand?
III. EMPIRICAL RESULTS
The model was estimated using annual data on three versions of the depen-dent variable:
the log of total consumption of cider, in pints per head of populationover the age of 15 (for the period 1952-1986; see Figure 4),the logit of the share of cider in the total consumption of long drinks(for the period 195 2-1986; see Figure 5),the log of the total consumption of packaged cider, in pints per headof population over the age of 15 (for the period 1962-1986; seeFigure 6; data on draught and packaged cider has been separatelyavailable only since 1962).
In Table 2, we present OLS estimates for each of these equations. Theoverall statistical performance of each equation is good. We have R2's of0.995, and 0.997 respectively; goodness-of-fit F-statistics of 333, 548 and602 respectively thus comfortably rejecting the joint null hypothesis that the
N)
1952
1954
-
1956
-
1958
-
1960
-
1962
-
1964
-
1966
-
1968
-
1970
-
1972
-
1974
-
1976
-
1978
-
1980
-
1982
-
1984
-
1986
Logi
tsc
ale
o
1952
N)
1954
1956
-
1958
-0) -
1960
-
1962
-
1964
-e
1966
-
1968
-
1970
-o
1972
-
rD19
74-
u19
76-
1978
-00 O
\19
80-
1982
-
1984
-
1986
Log
scal
e
NN
)N
)o
N)
N)
2.4
2.2-
2.0-
1.8-
1.0-
Fig. 6. Per capita consumption of packaged cider, 1962-1986
coefficients are not significantly different from zero; h-statistics of - 1.25 and0.32 for equations (i) and (iii) (the 5 percent critical value is ± 1.645), and aDW statistic of 2.04 for equation (ii), together with LM statistics for the threeequations of 2.11, 2.42 and 1.44, thus failing to reject the null hypothesis ofno residual autocorrelation; and Chow-statistics of 1.90, 0.35 and 0.14respectively. So there is little evidence of autocorrelation in the residuals, andthe Chow test indicates good predictive performance, particularly for theshare equation.
The estimated values of the parameters generally conform with a prioriexpectations. We find that the own-price effect (LPC) is significant in eachequation, with the correct negative sign, and not very elastic. In the equationfor total per capita consumption, it is - 0.413; in the share in long drinksequation, it is - 0.348; and in the packaged cider equation, it is more elasticat - 0.852, reflecting the fact that it is competing with a more perfect substi-
0.8 -
0.6C\ICD
.
CDO) O)
1
CDCDO)
ICDCDa)
IoNa)
CJr-a)
r-a)
ICDr-O)
ICDr-O)
IO
O)CDa)
1CDO)
COCDa)
THE DEMAND FOR CIDER IN THE UNITED KINGDOM 79
TABLE 2OLS Estimates of the Cider Demand Equations
NOTE: R2 is R2 adjusted for degrees of freedom, SER is the standard error of the regression, his the Durbin h-statistic (the Durbin-Watson (DW) statistic is not appropriate in the presence ofa lagged dependent variable), CHOW is the F-distributed Chow statistic for goodness-of-fit ofthe last four observations, LM is the Lagrange-multiplier F-statistic for residual autocorrelationup to order 4, and FIT is the goodness-of-fit F-statistic for the equation. Critical values (5 per-cent) are presented besides the statistics: the rejection region for the null hypotheses lies abovethe critical values. The figures in parentheses are t-ratios. We adopted the rule of only includingthose explanatory variables for which t> 1, though in our final equations, the t-ratios areusually significant at the 5 percent level, if not better. For definitions of regressors used, see thedata appendix. All computations were undertaken using TSP.
¡'er capitaTotal per capita Share of cider consumptionconsumption of in long drinks of packagedcider market cider
THE DEMAND FOR CIDER IN THE UNITED KINGDOM 81
tute, draught cider.2 However, the own-price effect in the total cider equationbecomes more elastic after 1977, as is discussed later. In the first equation,the cross-price elasticities of beer and wine are significant and positive (indi-cating that beer and wine (LPB and LPW) are substitutes for cider).3 Theprice of wine (LPW) is also significant in the share equation.
Advertising expenditure on cider (LAC) was found to have a positive, sig-nificant but small effect in cases (j) and (ii). In the case of packaged cider, thenegative effect of expenditure on advertising of beer (LAB) seemed to bemost significant, having an elasticity of -0.167. This may not be surprisingwhen we consider that the cider industry's advertising campaign has beendirected towards increasing draught cider consumption. Our results on own-advertising elasticities concur with previous work, such as that of Duffy(1982b).
As indicated previously, we tried various income variables and found thatG.D.P. (LYN) performed best in equations (j) and (iii), and P.D.1. (LPYN) inequation (ii). In the latter case, cider appears to be an inferior product with anincome elasticity of - 0.887; it may be that it is inferior in terms of the budgetallocated to long drinks expenditure but normal with respect to total expendi-ture. We must also remember that the image of the cider drinker waschanging during this period.4
In two of the three equations, the lagged dependent variable is significant,and positive, indicating some role for habit formation in our analysis.5 Wetried lagged versions of various explanatory variables but these did notimprove our preferred specifications.
We introduced linear, quadratic and cubic time trends in our equations andfound significant values for a quadratic trend in the packaged equation (T andTSQ), indicating some growth over time, with the quadratic term damping itin about forty years. In September 1976, excise duty was imposed on cider,which previously had been free of such duty. So we tried an intercept dummyvariable (D77) and a dummy for a change in the slope of the own-price term(DPC77) for the subsequent years, 1977-86, and we found these were signi-ficant factors for the total cider equation. When we include the effect of theslope dummy in the first equation, we find that the demand for cider changesfrom being inelastic before 1977 to being highly elastic after 1977 - a shift inelasticity from -0.413 to - 1.584 (i.e. -0.413 - 1.171). This is associatedwith a sharp rise in the price of cider in 1977 and a corresponding movement
2 Jones (1989) finds (using quarterly data over a similar period to ours), that the demand forcider is more own-price elastic than we do (about 0.6 compared with 0.4 for the nearestcomparable model).
3iones (op.cit.) finds that beer and wine are poorer substitutes for cider than we do; hiscross-price elasticities are very close to zero.
4iones (op.cit.) finds that the income elasticity of the demand for cider is around unity,slightly more income elastic than our nearest comparable model.
This result is also confirmed by Jones (op.cit).
82 BULLETIN
to a more elastic part of the demand curve.6 The intercept dummy provedsignificant for the share equation, though neither performed well in the caseof packaged cider.
Of the other explanatory variables included, we find the following results.The mean temperature during the summer months (LTSUM) was alwayssignificant and positive, as would be expected, while mean daily sunshine(LSUN) proved significant and positive in equation (i). The percentage ofunemployed (LUN) proved to have a significant positive effect in equations (i)and (ii), and the proportion of males employed in production industries(LEMPN) had a highly significant and positive effect in the case of total ciderconsumption. The proportion of the population in social classes A, B and Cl(LPOPAB) did not have a significant effect on cider demand. In the shareequation, the on-licences variable (LONL) was significant and positive, indi-cating the importance of public houses in determining the division of the longdrinks trade between beer and cider. In the packaged cider equation, thenumber of on-licences has a significant but negative effect, which corre-sponds with the fact that packaged cider is essentially an off-licence trade.Nevertheless, the number of off-licences (LOFFL) did not have a significanteffect in the packaged cider equation, although it does have a significanteffect in the total cider equation. To account for differing regional tastes, weincluded in our analysis the proportion of the population living in the southwest of England (LPOPSW); this was found to have a significant positiveeffect in the share equation.
The number of males and the number of females between 15 and 24 arehighly correlated, making it difficult to include both as explanatory variables.However it was consistently found that each variable had a significant butopposite-in-direction effect in our equations, and that this effect was lost if avariable such as the total number of females and males between 15 and 24 areincluded alone. The number of young males in the population has a strongnegative effect on cider consumption, whereas the number of young femaleshas a (slightly less) strong positive effect on cider consumption - youngwomen like to try cider. So we looked for the best fit using any two of thefollowing variables: the percentage of 15 to 24 year olds in the total popula-tion (L 1524), the percentage of 15 to 24 year old males in the total popula-tion (LM1524), and the percentage of 15 to 24 year old females in the totalpopulation (LF1 524).
Our results indicate, for example, that a one percent increase in the pro-portion of young women will increase total per capita cider consumption by3.1 percent (calculated approximately as 6.837 (0.5 X 7.440)), the share ofcider in long drinks by 1.6 percent, and per capita consumption of packagedcider by 4.2 percent. On the other hand, a one percent increase in the propor-
The loglinear specification that we have chosen assumes that the elasticities are constantunder all circumstances. This may be valid for small changes in, say, own-price. But for largechanges in own-price that result from, say, the imposition of excise duty, the constant elasticitymodel may not be valid. This is confirmed by the significance of the dummy DPC77.
THE DEMAND FOR CIDER IN THE UNITED KINGDOM 83
tion of young people (holding the ratio of young males to females constant)will reduce total per capita cider consumption by 0.6 percent (calculated as6.837 - 7.440), will increase the share of cider in long drinks by 0.3 percentand reduce per capita consumption of packaged cider by 0.2 percent. Theseresults show that cider is a popular drink amongst young women, but youngmen soon graduate from cider to other long drinks such as beer.
IV. CONCLUSIONS
We have considered the demand for cider in terms of total cider consump-tion, cider consumption as a share of the long drinks market, and the con-sumption of packaged cider alone, for the years 1952-1986 in the first twocases, and 1962-1986 in the latter case. This was an interesting period forthe cider market. As can be seen in Figure 1, there was an initial decline incider consumption, followed by a remarkable resurgence in the late 1 960sand 1 970s, but petering out in the 1 980s.
To capture these dynamic changes, we have used a standard demandmodel, but rather more substantially specified than existing studies in thisarea. We find correct and significant signs for most of our important eco-nomic variables; more than this, we find that non-economic variables have asignificant role to play in influencing cider demand.
In terms of the hypotheses that we are interested in investigating, we findthat
Standard economic variables are important in explaining the demandfor cider; in particular, we find that when considered just in terms ofthe long drinks market, cider is an inferior good (i.e. it has a negativeincome elasticity), but in terms of total consumption, it is a necessity(i.e. it has a positive income elasticity less than unity); furthermore,before 1977, demand was price inelastic, but after 1977, it becamehighly price elastic, causing cider manufacturers' revenues to becomeconsiderably more price sensitive;7the significant lagged dependent variable indicates that cider is ahabit-forming long drink (in terms of per capita consumption): morethan a third of a percentage point of additional current cider con-sumption is due to previous additional cider consumption;advertising appears to have a rather weak effect on cider demand;as with all cold drinks (both alcoholic and non-alcoholic), goodweather, especially in the summer, is good for cider consumption;
Our model also satisfies the restrictions imposed by demand theory. For example, demandtheory requires that the sum of the uncompensated price elasticities equals the negative of theincome elasticity. In our total per capita equation, the sum of the price elasticities equals- 0.842 (i.e. - 1.584 + 0.326 + 0.416), while the negative of the income elasticity is - 0.947.Since the standard error of the income elasticity is 0.26, this means that these two figures arenot significantly different from each other.
84 BULLETIN
(y) cider demand depends significantly on the number of outlets, so thattotal cider demand is positively related to the number of off-licences,while packaged cider demand is negatively related to the number ofon-licenced premises: cider availability at outlets selling alcoholicdrinks can be thought of as a form of immediate advertising and assuch has a more important impact than direct advertising whoseeffect is inevitably delayed;an increase in the unemployment rate has a positive effect on thedemand for cider: this means that as people become unemployed andsuffer a substantial reduction in purchasing power, they neverthelessincrease their demand for cider at the expense of other forms of con-sumption; note also the opposite effects of increases in personal dis-posable income and unemployment on the share of cider in longdrinks;we find that social classes A, B and Cl do not consume cider to anysignificant extent, but that working class demand for cider (asproxied by production sector male employment) is statistically sig-nificant, especially for total cider consumption: as the role of themanufacturing sector in providing male employment declines, so toowill the consumption of cider.young women have a significant demand for cider: any decline in thisgroup will do substantial harm to the cider market;large changes in the price of cider do lead to changes in own-priceelasticity, thus rejecting the constant elasticity model.
The implications for the cider industry are clear. A declining share ofmanufacturing in total production will continue to depress working-classdemand, while demographic factors, in the form of a declining share of youngpeople in the total population, will reduce demand from young women, theother mainstay of the cider market. Our model also suggests that advertising,which has typically targeted young men, is not likely to reverse the decline.With the current high elasticity of demand for cider, any significant rise in theprice of cider in real terms is likely to substantially reduce demand.
Our analysis has shown the importance of developing a model that is fullyspecified in terms of both economic and non-economic variables. There maybe lessons here for other studies of demand, particularly on the alcoholmarket, where we have shown that climatic, social class and demographicvariables have a significant impact.
Birkbeck College, LondonKing's Fund Institute, London
Date of Receipt of Final Manuscript: September /99/
THE DEMAND FOR CIDER IN THE UNITED KINGDOM 85
REFERENCES
Brewers Society (1987). UK Statistical Handbook.Bulmer, H.P. (1987). Annual Report.Duffy, M. (1982a). A Case Study in Econometric Forecasting for Alcoholic Drinks',
Omega, 10, 597-611.Duffy, M. (1 982b). 'The Effect of Advertising on the Total Consumption of Alcoholic
Drinks in the UK: Some Econometric Estimates', Journal of Advertising, 1,
105-17.Duffy, M. (1982c). An Econometric Study of the Demand for Various Types of
Wines and Spirits in the UK', Occasional paper 8212, Department of ManagementSciences, UMIST, Manchester.
Duffy, M. (1983). 'The Demand for Alcoholic Drinks in the UK, 1963-78', AppliedEconomics, 15, 125-40.
Euromonitor Consumer Market Surveys (1984). The Cider Report.Jones, A. M. (1989). A Systems Approach to the Demand for Alcohol and Tobacco',
Bulletin of Economic Research, 41,85-106.McGuinness, T. (1980). An Econometric Analysis of Total Demand for Alcoholic
Beverages in the UK, 1956-75', Journal of industria! Economics, 29,85-109.McGuinness, T. (1982). 'The Demand for Beer, Spirits and Wine in the UK,
1956-79'. in Grant, M., Plant, M. and Williams, A. (eds), Economics and Alcohol,Croom Helm, London.
Mintel Market intelligence (1986).Salvananthan, E. A. (1988). 'Alcoholic Consumption in the UK, 1955-85: A System-
wide Analysis', Applied Economics, 20, 1071-86.Walsh, B. (1982), 'The Demand for Alcohol in the UK: A Comment', Journal of
industrial Economics, 30, 439-46.
l)ata Appendix
We present a list of the data and symbols that were considered in the empirical analysis.
LPC: log of the retail price of cider relative to the general retail price index (G.R.P.l.).LPB: log of the retail price of beer relative to the G.R.P.i.LPW: log of the retail price of wine relative to the G.R.P.l.LAC: log of per capita real expenditure on cider advertising (1980 prices).LA B: log of per capita real expenditure on beer advertising (1980 prices).LAW: log of per capita real expenditure on wine advertising (1980 prices).LUN: log of the percentage of the workforce who are unemployed.
LEMPN: log of per capita male employment in production industries.LYN: log of per capita real G.D.P. at market prices (1980 prices).
LPYN: log of per capita real personal disposable income (P.D.l.) (1980 prices).LCN: log of per capita real consumers' expenditure (1980 prices).
LCFN: log of per capita real consumption of food, drink and tobacco (1980 prices).LOFFL: log of the number of off-licences in the UK.LONL: log of the number of on-licences in the UK.
LMI 524: log of the percentage of 15-24 year old males in the population over the age of15.
LFI 524: log of the percentage of I 5-24 year old females in the population over the ageof 15.
Li 524: log of the percentage of I 5-24 year-olds in the population over the age of 15.LPOPAB: log of the percentage of the population in social classes A, B and Cl.
86 BULLETIN
LPOPSW: log of the percentage of the population living in the south-west of England.LSUN log of mean daily sunshine.
LTSUM: log of mean daily summer temperature.LTEMP: log of mean daily temperature.LRAIN log of mean daily rainfall.
LTMYJN: log of mean daily temperature in May and June.LAG) - I): previous year's value of the dependent variable.
D77: intercept dummy for 1977-86 (taking a value of unity during these years andzero otherwise).
DPC77: slope dummy for the price of cider in 1977-86 (taking a value of unity duringthese years and zero otherwise).
T,TSQ,TCUB: linear, squared and cubic time trends.
The demand variables and LONL and LOFFL were calculated from the Brewers' SocietyU.K. Statistical Handbooks. The price series LPC. LPB and LPW were formed using data fromthe same source together with price information supplied by the Victoria Wine Co. Ltd., anddata from Mintel Market Intelligence and Euromonitor.
The advertising variables LAC, LAB and LAW were obtained from the following sources:Statistical Review of Press Advertising, Statistical Review of Independent TV Advertising.Statistical Review of Press and TV Advertising, and The Media Expenditure Analysis Ltd(MEAL) Digest.
The macro-variables LYN. LPYN. LCN, LCFN and the retail price index, were obtainedfrom the Economic Trends Annual Supplement, and LUN and LEMPN from the MonthlyDigest of Statistics (M.D.S.). The climatic variables also came from the M.D.S. The demo-graphic variables LM1524, LFI 524, LI524 and LPOPSW were calculated from the AnnualAbstract of Statistics, and LPOPAB from the Family Expenditure Survey.
