Printer Friendly

Philippine electric demand and equivalence scales.

I. Introduction

The literature on residential electricity demand in developing countries is small, based principally on Latin American experience and, with the exception of Westley |24; 25~, employs highly aggregated data.(1) Price elasticities are generally found to be less than U.S. elasticities |26, 39~. One explanation by Westley is that energy substitution possibilities may be more limited in Latin American countries. He further suggests that since the household budget share of electricity is smaller in these countries, the Slutsky equation would imply an own-price elasticity closer to the compensated elasticity and less influenced by income elasticity. As a refinement, Westley also suggests that substitute fuels are more available in rural versus urban settings and that, as a result, price elasticity may be larger. Empirically, however, he found no difference |25, 257, 261; 26, 125~.

Income elasticity findings, on the other hand, are varied. Pindyck's |17~ evidence and inferences are that these elasticities are larger in developing than developed countries and, in both cases, decline with per-capita income. Berndt and Samaniego |3~, find results for Mexico that agree, in the first part, with Pindyck. In fact when electric-system connection rates are considered, they suggest that their findings show an income elasticity greater than one.(2)

Westley, however, finds low income elasticities for Paraguay and Costa Rica. He suggests such results are plausible with the higher estimates in other research possibly attributable to the use of aggregated data and left-out-variable problems |26, 217-19~.

Pindyck's second result is not confirmed. Berndt and Samaniego did not consider a variable elasticity and Westley drew the opposite conclusion. With data for Paraguay, Westley found that income elasticity increased with income. In Costa Rica, the indirect evidence suggested no relationship.(3) Specifically, urbanization, frequently correlated with income, exhibited no influence on either price or income elasticity.

The purpose of this paper is to further investigate electric demand and elasticities in developing countries. The analysis is the first to consider data from the Philippines, household demand with commercial usage excluded and demand disaggregated to the level of rural electric cooperatives.(4) In fact, this research is the first to examine data from an established rural electrification program.

While the model is dynamic, the data are cross-section for only two years and no attempt is made to consider flexible adjustment models of the type proposed by Westley |25; 26~. The relevant electric tariff incorporates a fixed charge or minimum bill entitling a customer to a given or first-block quantity of kilowatt-hours (marginal price = 0) and a marginal price for remaining consumption. The minimum bill is marginal price times the given, first-block quantity of kilowatt-hours. Since "minimum-bill" customers can be numerous, their relevance to price specification and the estimation of price elasticity is an important feature of this research. It leads to a separation of intra-marginal and marginal price effects.

A second and central feature is the use of equivalence scales.(5) Historically a topic in welfare analysis, such scales purport to measure differences in welfare across households by accounting for relevant differences in demographic profiles or environmental factors. The usual approach indexes implied variations in expenditure patterns.

In applied demand research, equivalence scales could be employed to represent or otherwise account for demographic and environmental profiles as determinants of demand. This might involve a priori specification of the model or it may involve the introduction of equivalence scales as variables. Following Pollak and Wales |18~, demand would then exhibit a conditional specification with equivalence scales best referred to as "conditional" equivalence scales. The conditional equivalence scales proposed in this research are useful for model specification and interpretation--but, more directly, enter to provide information on consumption and the accumulation of consumption patterns reflective of the role and importance of underlying variables for which data were not available. Since simultaneity problems can arise, the scales employed are measured with data from a previous period or otherwise specified a priori. To examine the conditional nature of some scales, functions determining these scales are specified and estimated with the earlier data.

In the remainder of the paper, part II presents the model and part III describes data and the sample employed. Part IV summarizes econometric procedures, hypotheses and findings. Finally, part V presents conclusions.

II. Theoretical Framework

Individual household electricity demand requires a stock of electricity-using durable goods. The associated investment problem suggests that lagged responses are likely to be important. One approach in allowing for such responses is, following Taylor |20~, to directly include in the demand model variables measuring stocks of electricity-using durables, related utilization rates and information on stock adjustments. However, the required data are not available. A partial or flow adjustment model of demand is therefore employed which implicitly recognizes investment and appliance-stock adjustments. In this model, desired electricity consumption (|Q*.sub.t~) is a function of appropriate economic and non-economic variables. Actual consumption (|Q.sub.t~) is adjusted to desired consumption in proportion (k : 0 |is less than~ k |is less than~ 1) to any existing disequilibrium (|Q*.sub.t~ - |Q.sub.t-1~).(6) Aggregation and Functional Form

Since the data analyzed are aggregate for residential customers in the service area of each rural cooperative, the individual theory of household demand must be imbedded in an aggregate model.

It is assumed here that the homogeneity postulate carries over and that the adjustment structure rationalized by the individual theory is maintained.(7) While the distribution of household income is measured only by mean income, variables are considered which measure the distribution of households across household size ranges and between the first and second blocks of the electric rate schedule. The specification and estimation of price elasticity depends on the last distribution and is discussed below.

The model, for simplicity and convenience, is specified in log-linear form. In several instances, variables are entered in linear form as they are not amenable to the log transformation or because the theoretical interpretation is improved. Ignoring the variables appearing in linear form, the model is generally given by equations (1) and (2).

ln(|Q*.sub.t~) = |b.sub.0~ + |summation over i~ |b.sub.k~ ln(| + |summation over j~ |b.sub.j~|interact.sub.j~ + |e.sub.t~ (1)

ln(|Q.sub.t~) - ln(|Q.sub.t-1~) = k|ln(|Q*.sub.t~) - ln(|Q.sub.t-1~)~ (2)

or, in reduced form for estimation

ln(|Q.sub.t~) = |c.sub.0~ + |summation over i~ |c.sub.i~ ln(| + |summation over j~ |c.sub.j~|interact.sub.j~ + |c.sub.2l~ ln(|Q.sub.t-1~) + |e.sub.t~ (3)

TABULAR DATA OMITTED where interactions are of the form

|interact.sub.j~ = ln(|| (4)

and where the b's, c's and k are parameters to be estimated; |c.sub.i~ = k|b.sub.i~; i, m and j index independent variables; t indexes time; and all variables are defined in Table I and later listed in Table III as they appear in the model.

Price Specification

Rural electric cooperatives in the Philippines uniformly employ a rate schedule with a minimum bill or fixed charge which allows a household to consume within a first-block of kilowatt-hours at a zero marginal price. A constant, non-zero marginal price (MP) applies to the second block or all additional consumption. Across cooperatives the maximum, first-block consumption (MKWH) under the minimum bill varies significantly.(8) For each cooperative the minimum bill or fixed charge is usually computed as MP times MKWH. Significantly, not all customers in the service areas of rural electric cooperatives in the Philippines consume beyond the first-block or MKWH. In fact, for some cooperatives as many as 80% of the customers may be minimum-bill customers and, for others, as few as 5%.

The presence of minimum-bill customers requires that care be taken in specifying price and price elasticity in a demand equation representing average household electricity usage. Variations in marginal price (MP) produce a pure price response for those customers consuming more than the minimum or allowed kilowatt-hour level.

For those customers consuming within MKWH, the marginal price is zero. At the same time, since the minimum bill itself is usually MP times the minimum-bill consumption, variations in MP are associated with variations in the minimum bill and can be viewed as having an intra-marginal price effect.(9) For this reason, the role and effect of variations in MP on average household electricity demand depends on the proportion of households who are minimum bill customers. As this proportion approaches one, the effect approaches that of a pure intra-marginal price. As the proportion approaches zero, the effect approaches that of a pure marginal price.

At the same time and across cooperatives, MP might be constant while the maximum allowed consumption (MKWH) in the first-block varies. For those customers who are minimum bill customers, this is the same as allowing more or less to be consumed at, effectively, a marginal price of zero. However, since the minimum bill or fixed charge is MP multiplied by MKWH, variation in MKWH means variation in the minimum bill and in the same direction. Therefore while an increase in MKWH would encourage consumption, the increase in the minimum bill would work in reverse but probably not to eliminate the consumption response. To allow for this possibility, MKWH is introduced in the model as an interaction with MBILL (the fraction of customers who are minimum bill customers).

In summary, the approach at this point enters interactions which allow (1) the marginal price elasticity to depend on the fraction of customers that are minimum bill customers and (2) maximum minimum-bill consumption to be a separate variable with an influence dependent on the fraction of customers who are minimum bill customers.(10) In the first case, as the fraction of minimum bill customers increases, price elasticity would be expected to diminish. In the second case, an increase in the allowed minimum-consumption should increase electricity demand (marginal price for increased consumption is zero and should dominate the adverse income effect of a larger fixed charge), and more so in the aggregate, the larger the proportion of customers who are minimum bill customers.

Price Elasticity and Conditional Equivalence Scales

Two other interactions involving marginal price are introduced and as a result appear in the reduced form equation, equation (3). One makes short-run price elasticity (|Mathematical Expression Omitted~) a function of the percentage of household income allocated to housing expenditures (HXP) and the other makes elasticity a function of whether or not the cooperative is in an urban (URB = 1) or rural (URB = 0) location.

|Mathematical Expression Omitted~

Both variables, considered conditional equivalence scales, enter to control for economic, demographic, climatic and resource profiles that are expected to vary in the cross-section sample with location and urban setting.

For example, available data show that, for households in the same income range, the percentage of income devoted to housing expenses (HXP) is significantly different across the Philippines. It is expected that, ceteris paribus, the larger HXP, the more likely it is that electricity use is restricted to necessary applications and associated with a lower price elasticity. It is also the hypothesis that variations in HXP are determined by more than just the price of housing (PHOUS).

The interaction with the urban dummy variable attempts to control for similar types of possibilities. Substitute fuels, particularly wood, are more available in rural locations and this, by itself, may lead to a higher, rural price-elasticity.

Conditional Equivalence Scales and Rural Electrification (RE) Load Research

In order to allow for the possibility that type of housing is related to electricity use or otherwise represents wealth or household preferences for home living, a variable measuring the fraction of housing in the category of single homes (SHOME) as opposed to apartments and duplexes is entered. In addition a variable measuring the quality of roof construction (ROOF) is also included. In rural-electrification cost-benefit studies, preliminary survey work will commonly collect data on such variables as possible indicators of energy use potential. The rational is similarly that they index household wealth or preference for undertaking consumption at home.(11)

While this model does not attempt to account for the type and energy use features of the stock of electricity-using durable goods, the presence of a major item such as a refrigerator may condition demand behavior. Therefore, following Westley |26, 223-37~, a variable measuring the saturation rate of refrigerators (REFRIG) is considered.(12) Saturation rates vary from 3.7% to 43.5% in the sample and average 6.3% for cooperatives in rural locations as opposed to 29.6% for those in urban settings. However, the simple correlation of REFRIG and HXP is .90 and REFRIG may not have a separate role in the model.

Other Equivalence Scales

Underemployment (UNEMP) is included as it is expected to be positively correlated with time spent at home and hence electric usage. Increases in the number of electrified municipalities in the service area of a cooperative (MUN) is expected to show customers more scattered in villages. If markets for second-hand durables are less developed among scattered populations (MUN large), restricted availability and higher prices, ceteris paribus, may reduce electric demand.

III. The Data

Data were collected for 115 Philippine rural electric cooperatives. This sample includes all rural electric cooperatives for which NEA (The National Electrification Administration) collects monthly reports. The data are cross-section for 1986 and 1987.(13)
Table II. Sample Means: Selected Variables

No. Variable Rural Urban

 1. MP (Pesos/kwh) 2.189 1.861
 2. |Q.sub.t~ (kwh/customer) 22.553 35.230
 3. MBILL (fraction) .408 .317
 4. Collection Efficiency (%) 89.625 78.938
 5. System Losses (%) 21.458 24.776
 6. YH (Pesos) 23,660.0 29,130.0
 (U.S. $) 1,154.15 1,420.98
 7. HXP (%) 7.030 10.543
 8. SHOME (%) 95.598 88.045
 9. ROOF (%) 37.393 41.802
10. REFRIG (%) 6.407 29.295
11. MUN (per Cooperative) 11.522 11.286
12. UNEMP (%) 35.041 31.929
13. PFOOD 335.27 343.85
14. PFUEL 501.78 501.83
15. PHOUS 381.30 385.74
16. P (all items) 359.22 367.64
17. Sample Cooperatives (#) 46 42

Note: Household income is measured in pesos. But it is also adjusted with a
regional price index to reflect real differences in income across regions of
the Philippines. Also, the reason that the percentage of families living in
single homes is smaller in urban than rural areas is related to the appearance
and use of apartments, condominiums and duplexes.

Data for cooperative-specific variables (Q, MP, MBILL, MKWH, MUN, URB) were obtained from the National Electrification Administration |15; 16~. Data for all other variables were collected from the National Census and Statistics Office |11; 12; 13~ and the National Statistical Coordination Board |14~. The data for these last variables were available by cities or regions and were associated with the service areas of cooperatives. In most cases, regions were sufficiently small to coincide well with cooperative service areas.

Missing observations in the data for the number of minimum bill customers and several other variables resulted in a reduction in the number of usable observations from 115 to 88.

Research is frequently concerned with the economic and non-economic differences between rural and urban locations |8; 9; 5; 22~. The viability of cooperatives in rural and urban locations is also an issue. Therefore, Table II provides descriptive statistics for the 88 cooperatives studied separated into two samples: one predominantly rural and the other urban.

As expected, electricity price is higher and quantity consumed lower for cooperatives in rural locations. The picture is also that household income, the percentage of household income allocated to housing, the percentage of houses with strong roofs, refrigerator saturation, the price of housing and the price of food are all lower. Underemployment is higher in rural areas and, on a percentage basis, single homes are more prevalent. Finally, differences in collection efficiency and system losses suggest that fewer problems exist with operating cooperatives and handling theft in rural locations in the Philippines.

IV. Estimation and Results

Table III reports estimation results, including short-run price elasticities and estimated adjustment rates for four versions of the model differentiated by the exclusion of variables. Long-run price elasticities are not reported, but are easily obtained by dividing the short-run elasticity by the estimated rate of adjustment. The concepts of adjustment rate and long-run elasticity should be qualified with the presence of equivalence scales. This might be expected particularly with the introduction of the saturation rate for refrigerators. However, the variable was unimportant and results were robust with respect to inclusion/exclusion.

Price Elasticity

Price elasticity is statistically significant. However, the low adjustment rates (.18 to .20) imply that long-run elasticities are much larger than the reported short-run elasticities. Larger adjustment rates might have been expected and further research with longer time-series data would raise the level of confidence in these results.

Figure 1 shows the relationship between price elasticity, the percentage of family income spent on housing (HXP), rural/urban location (URB) and the proportion of customers that are minimum bill customers (MBILL). Considering only the rural/urban dummy variable, demand is marginally more price elastic in rural than urban settings--weakly supporting the contention that energy substitutes such as wood may create greater substitution possibilities. As expected, HXP is inversely related to price elasticity. Since households in rural locations spend a significantly smaller fraction of income on housing, this further and significantly widens the difference between rural and urban price elasticities.

Figure 1 also shows that as the proportion of customers who are minimum bill customers increases, price elasticity falls. This is expected as the estimated price elasticity now becomes more nearly a pure intra-marginal price elasticity. For alternative values of HXP, rows 1 and 2 estimate pure marginal price elasticities and rows 11 and 12 estimate pure intra-marginal elasticities.

Figure 2 provides results similar to those in Figure 1, but for the model with REFRIG excluded. REFRIG was statistically insignificant, was highly correlated with HXP and may not have a role separate from HXP and other variables in the model.

Finally, an equation explaining HXP, the share of income going to housing, was estimated with earlier data and with electricity price entered.(14) All variables were important with the exception of household income. Therefore with the single modification to the demand model that HXP be a function of price, equation (5) (defining short-run price elasticity for electric demand) becomes: |Mathematical Expression Omitted~


where the coefficients and variables are defined by equation (3) and Tables I and III. Re-estimated price elasticities, reported in Figure 3, indicate only marginal differences from those in Figure 2 where HXP is assumed fixed.

Income Elasticity and Other Findings

Consistent with Pindyck's findings, income elasticity declines with income. Table IV reports estimates for a range of income levels and for the different versions of the model reported in Table III.

The remaining results were as expected--with two exceptions. First, while price indices individually and as a group were important in the equation explaining HXP, they were not important when entered separately from HXP in the demand model.

Secondly and as previously mentioned, REFRIG was statistically insignificant suggesting that it might not have a role separate from other variables.

Exceptions aside, the positive and statistically significant coefficient associated with SHOME suggests that households not living in single homes but instead living in duplexes or apartments will use less electricity. The results for the variable measuring quality of roof construction (ROOF) substantiate the presumption, frequently found in rural electrification feasibility studies, that households having homes with stronger construction (good roofs) connect to a cooperative's distribution system with higher demand. Since income is also entered in the model, it is possible that ROOF is simply a proxy for wealth or for non-monetary elements of income.

Finally, it is worth noting that, of the variables measuring the size distribution of households, only F2 and SIZHLD are retained in the results reported. The other variables considered were statistically insignificant and their removal had no effect. Uniformly the pattern of results was that SIZHLD was inversely related to electric demand and F2 (the fraction of households with 3-4 members) positively related.
Table IV. Short-run Income Elasticity Estimates

 Estimate (t-statistic)

Income Level (1) (2) (4)

1. 15,000 .662 (2.274) .674 (2.332) .943 (3.068)
2. 20,000 .423 (2.240) .432 (2.309) .604 (3.036)
3. 23,660 (Rural Sample) .283 (2.098) .291 (2.177) .406 (2.856)
4. 25,000 .237 (1.984) .245 (2.067) .341 (2.705)
5. 29,130 (Urban Sample) .110 (1.220) .116 (1.301) .161 (1.675)
6. 30,000 .085 (0.973) .091 (1.048) .126 (1.343)

Note: Income is measured in pesos.

Specification Issues

The model explains per household or per customer demand, ignoring connection rates which in developing countries can be much less than 100%. There are two immediate implications. One is that demand elasticities may be larger when impacts on connection rates and hence impacts on total demand are considered.

The other is that, following Westley |26, 207-8~, there may be errors-in-variables problems. Specifically while the object of study is the electric demand of households connected to the electric system, data for independent variables frequently measure conditions for all households in an area. Now, it is possible that the RE program in the Philippines did generally connect households that were representative of area populations. However, if it did not, the most likely hypothesis is that households with higher incomes were connected first. Income per household for those connected would then exceed that for all households in the area with the difference approaching zero as the connection rate approached 100%.

In this research, the first issue--the relation between connection rates and total demand--is not pursued. Using Westley's methodology, the errors-in-variables problem was addressed and rejected as statistically not important.(15)

V. Conclusion

This research is the first to study electric demand in a developing country with data disaggregated to level of electric cooperatives. It is the first to employ Philippine data. The previous literature principally involves Latin America and, with the exception of Westley |24; 25~, the data studied are all highly aggregated. The determinants of demand in this literature have correspondingly been restricted to a few variables. This research, offers findings on the importance of more detailed and disaggregated information--including variable introduced as conditional equivalence scales. It also offers results relevant to load research and pricing in third world RE programs.


Aggregate price elasticity is found to be a mixture of intra-marginal and marginal price effects generated by a two-block rate schedule. Depending on interactions which allow elasticity to vary with economic factors, a pure intra-marginal elasticity is somewhere between 1/2 to 1/4 of the pure marginal price elasticity.

Next, the finding that urban price elasticities are only 31% to 36% of rural elasticities is a new result in the literature. Westley |26~ suggested that availability of substitute fuels in rural locations in Latin America might lead to such findings. But his research provided no supporting evidence. While the results here do, the explanation is not so much differences in fuel availability as it is variations in the percentage of household income required for housing. The percentage increases dramatically in urban locations leaving less income for other consumption--including the purchase of electricity. In addition, it is also worth suggesting that electricity use and some corresponding appliance applications may be viewed as more necessary in urban settings. For example, if markets for fresh produce are less accessible or convenient on a daily basis, appliances such as refrigerators may be considered more necessary. In fact the sample correlation between refrigerator saturation (REFRIG) and the percentage of income devoted to housing (HXP) is .90. REFRIG was statistically insignificant and excluding this variable from the model left results essentially unaffected.

To compare price elasticity estimates here with those obtained in Latin American research involves some careful consideration. As indicated in Figures 1 to 3, estimated short-run elasticities in the Philippines vary significantly depending on determinants. For example and assuming no intra-marginal effects (MBILL = 0%), the short-run price elasticity lies in the range from -.182 to -.2 when HXP is as large as 12.6%. When HXP is reduced to 5.5% the range becomes -.55 to -.63. In Latin America, short-run price elasticities range from -.17 to -.44.(16) It is not clear what value for HXP should be assumed for the Latin American data.

The adjustment rate, estimated to be .20 yields long-run price elasticities that are less than one in absolute value for urban cooperatives but larger than one for rural cooperatives. Westley's results for Paraguay, Costa Rica and the Dominican Republic yield larger adjustment rates in the range from .35 to .88. Berndt and Samaniego obtained a rate of .35 for Mexico. Estimated U.S. adjustment rates--with the exception of Lyman |6~--are more in line with the estimate of .2 here. However, any conclusions warrant caution. The number of time-series observations in the pooled, Philippine sample is small, and with the importance of adjustment rates for the calculation of long-run elasticities further research with more extensive time-series data is warranted. Turning to income elasticity findings, Pindyck's inference that income elasticities are larger in developing rather than developed countries and decline with income are, in both cases, supported. One seeming difference is that the short-run income elasticity, estimated for the urban group of cooperatives, is approximately the same as an average elasticity for 26 developed country studies (see Westley |26, 36~). However, this is the urban group of cooperatives.

The result, that income elasticity declines with income in the Philippines, suggests that Latin American experience may be non-typical of developing countries. Of course, Berndt and Samaniego did not consider a variable income elasticity in Mexico. But Westley, in Paraguay, found that elasticity increased with income. In Costa Rica, his indirect evidence was for no relationship.(17) The size distribution of households was found to be important beyond simply a measure of mean household size. Berndt and Samaniego did not include household size. Westley |24~ included a variable defined to be the average number of occupants per dwelling but in later research |25~ with disaggregated data excluded it on the grounds it was statistically insignificant. Household size variables commonly are entered and found to be important in empirical studies of demand and this research establishes their significance for electricity demand in developing countries.

This is also the first research on a developing country to include price indices for food, housing and fuels. Westley |24; 25; 26~ has introduced individual fuel prices and an appliance price index. The price of liquified petroleum gas was found to be important in Costa Rica but not the price of Kerosene. The appliance price index was also important in Costa Rica but not in Paraguay. Here, price indices are important both individually and as a group in explaining HXP, but when entered with HXP in the demand model have limited significance.

Rural Electrification Programs

This is the first research to study electric demand for an established rural electrification program in a developing country. The findings provide information useful to load/demand analysis and pricing.

Load or demand analysis is crucial both in the preliminary screening of RE projects and later economic justification. The demand forecast is the basis for measuring benefits and defining investment and subsequent cost streams. Considering only residential demand, the results here offer support for the strategy of developing separately and then aggregating demand forecasts for those customers living in housing with strong roofs and those that do not. Income data might be better to collect along with forecasts for low and high-income households weighted by the numbers of households in each category. Unfortunately this is easier said than done and the "strong-roof" methodology may be the only resort.

Secondly, while all determinants of demand are important, the finding that urban price elasticities are only 31% to 36% of rural elasticities has substantive implications. With standard methodology (see Munasinghe |8~), much larger benefits would readily be attributable to programs in urban areas or otherwise falling under the heading of "village electrification".

Finally, turning to pricing, the results support the view that the Philippine rate schedule, with a minimum bill entitling customers to consumption at a marginal price of zero anywhere in a first block, encourages waste. This schedule should be replaced with a new schedule that would adopt an access charge reflecting marginal costs of access and a marginal price reflecting the marginal costs of supply. The marginal price should apply to all kilowatt hour consumption and, if too much revenue is collected, access charges could be reduced to token levels. If revenues remain excessive, inverted rate schedules could be considered as a further application of Ramsey Pricing |2~.

1. See Munasinghe |10, 5~ and Amarullah |1, 1~ for observations on the general literature. Amarullah provides some aggregate results for Indonesia. The Latin American literature consists of papers by Berndt and Samaniego |3~ and Westley |23; 24; 25; 26~. Most of Westley's earlier work appears in his New Directions in Econometric Modelling of Energy Demand |26~.

I am unaware of any published research involving electric cooperatives in developing countries. In the U.S., the only published research on cooperatives is Maddigan, Chern, and Rizy |7~. However, the data are not for individual cooperatives but for state level aggregates. The model used is the partial adjustment type but with differences from usual specifications. Instead of explaining per customer electricity use, the focus is total electricity consumption. The number of households, total electric consumption lagged one year, household size and income per capita appear as independent variables.

It is worth noting that Maddigan, Chern, and Rizy do propose to compare urban and rural patterns of electricity use by contrasting their results with those for the same model estimated with state data for all residential customers. They find no difference.

2. Berndt and Samaniego |3~ first define their model as explaining total rather than per-customer electric demand. Income elasticity is then composed of a direct effect of income on total demand and an indirect effect operating through the customer connection rate. The structural model consists of a connection rate equation and a total demand equation with the connection rate entered. The methodology is appealing, but implementation raises questions. First, the total demand equation is replaced by a per customer equation. Second, the connection rate variable is measured by the number of hook-ups divided by population.

It is not clear that the first step is valid and clearly, the connection rate should be measured by the number of hook-ups divided by the number of households--not population. The inverse of hook-ups over population resembles a variable measuring household size--an approximation that becomes better as hook-ups increase. Since household size is typically an important variable influencing electric usage, caution must be exercised in interpreting results with this specification.

3. For a more complete discussion of results for Paraguay see Westley |27, 102-3~. In Costa Rica, interactions were employed to see if price and income elasticities were affected by urbanization. Urbanization is highly correlated with income. See Westley |25, 261~.

4. The nomenclature, rural electric cooperative, derives from the United States. In developing countries these cooperatives can frequently be located in villages or urban areas. For a description of rural electrification and the development of these cooperatives in the Philippines, see Price Waterhouse |19~ and World Bank |28; 29~.

5. See Pollak and Wales |18~. Conditional equivalence scales are most frequently discussed in their application to welfare comparisons. See Tsakloglou |21~.

6. The model can be developed and justified within a utility maximization framework with an infinite planning horizon (results available on request). Otherwise Westley |26, 141-57~ employs the model and considers the possibility of a flexible adjustment rate rationalized through the investment literature.

7. Aggregation can produce problems. See Brown and Deaton |4~.

8. MKWH has a mean of about 12 kwh/month and ranges from a minimum of 0 to a maximum of 35 kwh. Marginal price has a mean of 2.12 pesos/kwh and ranges from a low of 1 peso/kwh to a high of 5.1 pesos/kwh.

9. 106 out of 115 cooperatives exactly determined the fixed charge or minimum bill as marginal price (MP) times the consumption (MKWH) allowed by the minimum bill. Of the remaining 9 cooperatives, 6 approximately do this.

10. If customers were to move between the two rate schedule blocks with some frequency, an appropriate price specification would have to take this into account--especially with annual data employed. However, this does not appear to be the case and minimum-bill customers are minimum-bill customers for the entire year.

11. This strategy was common in RE studies in Bangladesh and Zaire involving the National Rural Electric Cooperative Association and the U.S. Agency for International Development.

12. REFRIG is highly correlated with HXP (% of income allocated to housing costs) and, a priori, might not have an independent role in the model. Other saturation data were available for television sets and radios. However, these data were very highly correlated and by choice were not entered in the model.

13. The 1986 data essentially allow the model to be dynamic. A more extensive time series would have been useful.

14. It should be kept in mind that because HXP is measured with earlier data, it is a predetermined variable from the point of view of estimating the demand model itself--equation (3). It is an endogenous variable only in the earlier period.

The independent variables employed in the HXP equation examined with the earlier data were YH, P, PFOOD, PFUEL, PHOUS, URB, ROOF, SHOME, SIZHLD, MP, REFRIG. Dependent and independent variables (with the exception of URB) were entered in log form. With the exception of YH and SIZHLD, all variables had coefficients that were statistically significant. The corrected |R.sup.2~ statistic was .929. Results are available on request.

15. See Westley |26~, pages 207-8. The connection rate variable equivalent to Westley's share-of-households-connected variable was entered and found to be unimportant. On this basis the errors-in-variables problem is assumed to be unimportant.

16. See Westley |26, 22~. Short-run elasticities are easily derived as the long-run elasticity multiplied by the rate of adjustment.

17. See footnote 6.


1. Amarullah, Ir. Munawar. Electricity Demand in Indonesia: An Econometric Analysis. Publikasi LMK, No. 01-EP-84, Jakarta, 1984.

2. Baumol, William J. and David F. Bradford, "Optimal Departures from Marginal Cost Pricing." American Economic Review, June 1970, 265-83.

3. Berndt, Ernst R. and Ricardo Samaniego, "Residential Electricity Demand in Mexico: A Model Distinguishing Access from Consumption." Land Economics, August 1984, 268-77.

4. Brown, Alan and Angus S. Deaton, "Surveys in Applied Economics: Models of Consumer Behavior." Economic Journal, December 1972, 1145-236.

5. Jones, Donald W., "Urbanization and Energy Use in Economic Development." The Energy Journal, October 1989, 29-44.

6. Lyman, R. Ashley. "Advertising and Sales Promotion in Electricity." Journal of Regulatory Economics, March 1994.

7. Maddigan, Ruth J., Wen S. Chern, and Colleen Gallagher Rizy, "Rural Residential Demand for Electricity." Land Economics, May 1983, 150-61.

8. Munasinghe, Mohan, "The Economics of Rural Electrification Projects." Energy Economics, January 1988, 3-17.

9. -----, "Energy Economics in Developing Countries: Analytical Framework and Problems of Application." The Energy Journal, January 1988, 1-17.

10. -----. Rural Electrification for Development Policy Analysis and Applications. Boulder, Colorado: Westview Press, 1987.

11. National Census and Statistics Office. 1985 Family Income and Expenditure Survey. Vol. I-V. Republic of the Philippines, Manila, 1985.

12. -----. Monthly Bulletin of Statistics. Republic of the Philippines, Manila, monthly, 1988.

13. -----. Statistical Handbook of the Philippines, 1988. Republic of the Philippines, Manila, 1988.

14. National Statistical Coordination Board. Philippines Statistical Yearbook 1988. Republic of the Philippines, Manila, 1988.

15. National Electrification Administration. Monthly Financial and Statistical Reports (for individual electric cooperatives). Quezon City, Philippines, 1988. 16. -----. Status of Program Implementation: as of December 31, 1986. Quezon City, Philippines, 1987.

17. Pindyck, Robert S. The Structure of World Energy Demand. Cambridge: The MIT Press, 1979.

18. Pollak, Robert A. and Terence J. Wales, "Welfare Comparisons and Equivalence Scales." American Economic Review, (Papers and Proceedings), May 1979, 216-21.

19. Price Waterhouse. National Administration and Rural Electric Cooperatives Financial, Organizational and Technical Assessment. Report submitted to USAID Philippines, Contract No. OTR-0000-1-07-4427-00, March 13, 1987, Manila.

20. Taylor, Lester D., "The Demand for Electricity: A Survey." Bell Journal of Economics, Spring 1975, 74-110.

21. Tsakloglou, Panos, "Estimation and Comparison of Two Simple Models of Equivalence Scales for the Cost of Children." The Economic Journal, March 1991, 343-57.

22. U.S. Agency for International Development. A.I.D. Policy Paper: Energy. Washington, D.C., July 1984.

23. Westley, Glenn D. An Aggregate Time Series Study of Sectoral Electricity Demand in the Dominican Republic. ECC/MET P-25, Washington, D.C.: Inter-American Development Bank, 1984.

24. -----, "Electricity Demand in a Developing Country." The Review of Economics and Statistics, August 1984, 459-67.

25. -----. "Nontraditional Partial Adjustment Models and Their Use in Estimating the Residential Demand for Electricity in Costa Rica." Land Economics, August 1989, 254-71.

26. -----. New Directions in Econometric Modelling of Energy Demand: With Applications to Latin America. Washington, D.C.: Inter-American Development Bank, February 1992.

27. -----. The Residential and Commercial Demand for Electricity in Paraguay. Papers on Project Analysis No. 19, Washington, D.C.: Inter-American Development Bank, 1981.

28. World Bank. Philippines: Energy Sector Study. Washington, D.C., Sept. 15, 1988.

29. -----. Philippines Rural Electrification Sector Study: An Integrated Program to Revitalize the Sector. Washington, D.C., 1989.
COPYRIGHT 1994 Southern Economic Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1994, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Lyman, R. Ashley
Publication:Southern Economic Journal
Date:Jan 1, 1994
Previous Article:Estimating the choice behavior of self-employed business proprietors: an application to dairy farmers.
Next Article:The substitutability of equities and consumer durable goods: a portfolio-choice approach.

Related Articles
Stronger support for equivalence principle.
Sea lion makes waves with logical leaps.
New wine in old bottles: a meta-analysis of Ricardian equivalence.
Outfall from Bonn: a host of renewable energy initiatives.
JETRO Releases Results of Monthly Survey on Japanese Business Sentiment in East Asia.
Firms collect $27m to provide electric demand reduction projects.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters