Over at his newish blog, Jonathan Law politely points out that WTP4 opponents, conservationists and I are sort of full of it. He writes:

Unfortunately, opponents of this plan have chosen a convenient sticking point – the statistical methods employed – as the basis of their objections. These objections shouldn’t be enough to stop construction of the plant. This analysis isn’t perfect, but anyone dealing with statistics knows that literally no analysis will be. Austin’s City Council shouldn’t get bogged down by just another case of a group of people trying to play politics with what shouldn’t even be an issue.

Nothing more to see here. Full speed ahead on the new water treatment plant.

This argument is a bit off so I want to engage in some friendly back and forth.

For starters, I can’t speak for every person that is skeptical of WTP4, but in my case, I am not categorically opposed to it being constructed. I am interested in uncovering what is the optimal water management policy for my community, assuming we include environmental impacts (a personal preference) in determining what is optimal.

The decision about WTP4 is actually about what our overall policy to shape demand will be: pricing, voluntary conservation programs, regulation, treatment, recapture, etc. It’s not a debate about whether Austin is actually experiencing growth and needs infrastructure. I am reality-based community member. Thus, I suspect many people like myself would support the eventual building of additional capacity, but only after we have reached the most efficiency realistically possible on conservation. As a matter of fact, I believe Mayor Leffingwell’s position evolved from oppostion to WTP4 to support because his personal threshold for efficiency was satisfied. A lot of us skeptics are trying to find out what that efficiency frontier is. Therefore, it makes sense to look at the way the demand projections were constructed.

Jonathan makes a few other points which I paraphrase and then refute below:

1. “It doesn’t matter that the per capita rate is going down. Overall demand is still going up.” Actually, as I pointed out in this post, Austin Water’s own chart points out that this is not the case in from 2007-09. I reproduce the chart again below. Those years were not included in the Plummer model, as they were subsequent to their work. Regardless, nobody serious is contesting that Austin is growing and might need more infrastructure. We are arguing that if we add the treatment capacity, the urgency of preserving existing freshwater through conservation is destroyed because we can just treat ourselves out of the problem. Scarcity of treatment capacity better aligns with appropriate costing of the environmental externalities of our growth. So the question is about *when *we add the capacity – after we have abated the externalities as much as possible or before.

2. “Population explains 75% of the usage, so other modeling isn’t necessary.” Here Jonathan uses my own analysis against me (!), since I pointed out how surprised I was at the Plummer high r-squared. But I also pointed out that the consultant was aware of the changing nature of water consumption and decided to split the sample – acknowledging a one variable linear model’s problems with inflection points. Moreover, the reason I cautioned about inflection points was precisely because the recent data shows some significant changes, and it is possible that a single variable model will miss that because it is not including pricing, conservation incentives, regulation, etc. The danger of a single variable model is that it will derive the bulk of its fit from a period that is not immediately relevant (in this case the 70s,80s, and 90s) and omit a key variable that is the true driver of fit (say price per gallon) because it correlated with the regressed independent variable (population). Besides adding more independent variables, the consultant could have used a non-linear relationship or used time-series instead of conventional regression, just to name some obvious choices.

3. “No statistical basis for a 2.5 MGD per 10k residents scenario.” Agreed. There is no statistical basis for that as a strong predictor. It was only advanced as a way of dealing with the possible inflection point we are going through that the Plummer model can’t capture and therefore setting an upper bound on the discussion. I did not intend to suggest it was a full fledged and accurate alternative.

4. “What if population grows even faster than assumed by the model.” I make this same suggestion to proponents of the plant. I think the reason this hasn’t been pushed is that the City Demographer numbers might not be empirically verifiable and run into their own issues.

5. “No proof of additional conservation potential.” Well, this one always confounds me. Since we can price water, I am pretty sure that if we priced it at $10 per gallon there would be more conservation. Obviously that would be unjust, but my point is that there are effective, coercive mechanisms to reduce use. I am not advocating that, but it is important to understand the elasticity. Further, as we have repeatedly heard during this process, San Antonio has almost one third lower per capita residential use and I don’t consider it some water pricing and regulation dystopia.

That’s it for the back and forth. I hope it has generated some insight to our readers. And I very much look to Jonathan’s future blogging.

Julio,

Your comments are well received, and I don’t plan to post further about the WTP issue, but I would like to respond quickly here.

Historically, it did appear that the lion’s share of the fall in per capita consumption did occur in the few years immediately following the city’s implementation of the conservation program in 1983. So i’m not sure I would agree with the statement that the model derives the bulk of its fit from a period that’s not relevant, since it’s showed little change since the late 1980s. Furthermore, it’s not really clear that demand is going down from 2007-2009. There is the obvious low point there in July 2007, but it’s gone back up and of course numbers from 2009 wouldn’t really even be in yet.

As for possibly omitting true drivers of fit that would correlate with population, this makes sense but I can’t think of a variable that this would likely apply to. The fact that the model is 75% specified is still pretty huge, I think.

Admittedly, my focus was to attempt to show that the WTP should be built, which I do still believe. I mostly disregarded issues of efficiency and other ways to create conservation. These elements are very important and you’ve offered a thorough discussion and analysis of them. I appreciate your comments and look forward to your future posts as well.

– Jonathan Law

Another wastewater treatment plant = selling more water. Full steam ahead, right?

I wonder where conservation fits in to this…