A Panel Data Estimation of Domestic Water Demand with IRT Tariff Structure: The Case of the City of Valencia (Spain)

Round 1

Reviewer 1 Report

This paper intends to value water services demand. A very difficult task. One of the main keys in this kind of researches is to evaluate price elasticity. However, as it happens to many services, water prices have two different parts. And it is not easy to estimate price effects on water demand when one part of the price is fixed and not related to water use or consumption. This is well known by the authors. In fact, it is assumed by researchers in this field to evaluate changes in average prices instead of the marginal price.

Another main key in this researches is the different components of water demand. Water prices are very low and there are many criteria to be taken into consideration along with the price. Many of these issues are collected by the authors in the paper.

The paper tries to integrate all of these elements in a water demand model. But in the process of incorporated all these issues into the water demand model, they simplify too much. My personal point of view is that they have missed two important elements in water demand:

• A technical variable that affects water use. Technology improves reduce water consumption and has not a negative impact on welfare (washing machines and dishwashers).
• Other socioeconomic changes, not only related to the levels of education attained by the population (environmental consciousness, water savings programs, etc.).

All of these elements must be integrated into a water demand model to be completed.

However, this paper has made good progress in knowledge.

Author Response

Thank you for your comments.

Unfortunately, the lack of availability of data at the household level means that this work does not include these two important elements in water demand.  But considering your comments, we have added the following paragraph at the end of Discussion:

“The lack of availability of data at the household level means that this work does not include two important elements in water demand. The first is a technical variable that affects water use as an improvement of technology reduces water consumption and has not a negative impact on welfare (for instance, washing machines and dishwashers). The second one is related with socioeconomic changes not related to the levels of education attained by the population (environmental consciousness, water savings programs, etc.).”

Reviewer 2 Report

Please see the attached file.

Comments for author File: Comments.pdf

Author Response

Thank you to the reviewer 2 for these comments.

Reviewer 2:  On p. 14, l.476 and down, you mention that you test uncorrelation and homoscedasticity for the residuals. You reject these null hypotheses, and then, to alleviate the problems, you use some kind of robust variance estimation procedure instead. You then say that you have a significance loss of the neighbourhood effect (where are these results?), but then, a Wald test of “joint significance” of the neighbourhood parameters rejects the hypothesis of no neighbourhood effect. My comments on this are the following:

Response. This is the table can be incorporated as an appendix.

Reviewer 2:  As a statistician (not an econometrician), I am not much fond of robustifying variances rather than finding a better model. Have you considered improving on the model instead, for example by adding a second order lag Price variable?

Response. Both the inclusion of other variables (such as income) and a second price gap were considered, which was ruled out as it was not significant. From the economic point of view, given that the periodicity of the bill in Valencia is bimonthly, the water consumer is less likely to react to the price after four months (so the loss of significance of the price with 2 lags makes sense).

The robust estimation of the variances is carried out following the method of Arellano (2003) for the estimation with micropanels in the presence of heteroscedasticity and correlation of the residuals.

Reviewer 2:   When you have a lot of data and a small model, as in your case, it is not surprising that you reject any null hypotheses and that you get very small p values. The cause of rejection might be a very small and unimportant deviation from the null model. So I am not so surprised by your results here. I think a comment along these lines would be in place.

Response. We have added the following paragraph at the end of Discussion:

Another limitation would be related to the fact that when you have a lot of data and a small model (as in this work) it is very possible that any null hypothesis will be rejected and that very small p-values are obtained. The cause of rejection can be a very small and unimportant deviation from the null model.

Reviewer 2:    Table A1, AIC: I don’t think it says much to compare AIC across models that transform data in different ways, as you do here. I suppose this is why you also give something like “Corrected AIC” in this table? But what is the corrected AIC?

Another option for model comparisons like this is cross validation (Leave one individual out, estimate the model on the other individuals, then check the prediction error on the left out individual. Do this for all individuals and calculate some kind of overall error measure.) Did you consider that?

 

Response. Thank you for these comments.  We had not considered cross validation but we find it very interesting, and we will take it into account for future works. The use of the corrected AIC is very common in economic models (which is why we chose it).

Woo, C. K. (1992). Drought management, service interruption, and water pricing: Evidence from Hong Kong. Water Resources Research, 28(10), 2591-2595

 

The AIC statistic can also be applied to compare models in which the functional form in which the dependent variable appears is different. The most common case in economics is the comparison of two models, the regression can be done with Y and ln (Y). To do this, the AIC is transformed and the corrected AIC is obtained.

Jiménez, E. U., & GeaRosat, I. (1997). Econometría aplicada: Incluye diskettes con el programa ECOMET y datos. Editorial AC.

 

 

 

Reviewer 2:  

 

1. Abstract, l.9: the demand...	Added “the”
2. p.3, l.88: the panel data estimation method...	Added “the”
3. p.4, l.148: they	Before “They” Now “they” in small letter
4. p.4, l.163-: It seems that much the same sentence appears twice here.	Erased the second one.
5. p.7, l.222: makes	Added “s”
6. p.12, l.412: With a panel data sample...	Added “a”
7. p.12, l.422: ...or a panel data model...	Added “a”
8. p.12, l.425: ...tested if a pooled model...	Added “if”
9. p.15, l.523: was	Before “wss” Now “was”
10. p.15, l.542: located in (add a space)	Added a space, now “located in”
11. p.17, l.575: Erase the second "and".	Erased the second "and".

 

Response. The errors detected by the reviewer 2 have been corrected and furthermore, as a result of a revision of the text, others have been detected that have also been corrected (changes in red in the text).