Roger Haston’s Ridiculous Kill Shelter Model

Last January, I wrote about Dr. Roger Haston’s “The Future of Animal Welfare” presentation at an Animal Care Centers of NYC sponsored event. Dr. Haston, who was serving as the Chief of Analytics at PetSmart Charities at the time, was giving the same presentation at events held by shelters across the nation. While I acknowledged Roger Haston made some good points, I was deeply disturbed by his anti-pit bull and pro-killing shelter animals views. Furthermore, I addressed a number of problems with the arguments and so-called facts he presented. Subsequently, Animal Farm Foundation wrote a blog refuting many of Dr. Haston’s points and futile attempts to get Dr. Haston to address these. Additionally, Nathan Winograd dismantled Dr. Haston’s pro-killing arguments.

Shortly thereafter, Roger Haston left PetSmart Charities. Currently, Dr. Haston’s Linkedin profile states he is the President of the Institute for Animals. Unfortunately, I could not find anything about this organization. However, Dr. Haston’s Linkedin profile states the following about his position:

Strategy development, though leadership, research and leadership development services for the animal welfare industry. Focusing on the positive aspects of the relationship between people and animals.

Based on this description, it seems Dr. Haston may provide consulting services to animal shelters. Given the views Dr. Haston expressed in his “The Future of Animal Welfare” presentation, it seems kill shelters could look to him for guidance. In other words, kill shelters might look to his analyses as a way to argue against no kill.

What is the analytical basis of Dr. Haston’s anti-no kill views? Does this analysis make sense? Does this analysis match reality?

Haston’s Anti-No Kill Model

While Roger Haston did not present the model he used as the basis for his recent “The Future of Animal Welfare” events, a presentation from several years before may provide this information. In January 2015, when Dr. Haston was the Executive Director of Colorado’s Animal Assistance Foundation, he gave a presentation titled “Beyond Labels: Understanding the True Impact of Live Release Rates and Intake Policies” in a Society of Animal Welfare Administrators webinar. You can view the presentation here and the accompanying slides here.

Dr. Haston uses an interesting and robust statistical method, stochastic modeling, to conduct his analysis. Most animal sheltering data models, such as the one I created, use “deterministic modeling.” Deterministic modeling yields the same results from the inputs or variables included. On the other hand, stochastic modeling, incorporates the varied results an input or variable could have to predict the results generated from those variables. Therefore, in theory stochastic modeling is a powerful statistical tool.

While the deterministic model I use to target New Jersey animal shelter performance (amounts of animals shelters should adopt out, send to rescues and euthanize) is simpler theoretically, I reduce much of the sources of variability and therefore weaknesses of this type of model. For example, I analyze animal intake on a monthly basis, which accounts for higher intake during warmer months, and incorporate the breeds of dogs and ages of animals shelters impound and the overall adoption demand in a region. Furthermore, since I assess past performance, much of the input data I use has no variability at all. Therefore, my model performs quite well when I compare it to the actual benchmark shelters’ performance I use.

In a nutshell, Dr. Haston uses various shelter data estimates to make future projections. For example, he forecasts if a shelter will exceed capacity, what will happen when it exceeds capacity and what the facility’s future financial performance will look like.

Rigged Assumptions Lead to Anti-No Kill Results

Dr. Haston’s model would yield the same general conclusion regardless if he used stochastic or deterministic modeling. Why? He uses excessive animal intake and insufficient shelter capacity, excludes some of the quickest ways animals leave shelters and ignores how shelters would act if they exceeded capacity.

In Dr. Haston’s model, he assumes the shelter takes 3,000 dogs in during the year and the facility can hold 150 dogs at one time. Additionally, he assumes, based on an undisclosed sample of shelters, that different classes of dogs (from most to least adoptable) make up different portions of shelter intake and have varying average lengths of stay.

Using standard animal shelter capacity calculations, which assume animals come in evenly during the year, the shelter would have to on average move its dogs out of the shelter in 18.3 days or less to avoid consistently going over capacity. However, Dr. Haston’s model, which is based on each major class of dog’s average length of stay, shows these dogs would have an average length of stay of 32.0 days. Thus, a less sophisticated model would also show this shelter quickly exceeding capacity.

If there is anything to take away from this blog, this is it. Why? These key assumptions drive Dr. Haston’s subsequent conclusions that no kill animal control shelters severely restrict intake, are filled with animals few or no people want and financially implode.

Under Dr. Haston’s model, a shelter only adopts out or euthanizes an animal under the assumption all dogs are owner surrenders. Obviously, that is not realistic since stray dogs usually are a larger source of dog intake than owner surrenders. In addition, owners sometimes reclaim dogs they previously surrendered.

Typically, owners reclaim lost dogs within a few days since the animals usually have a license and/or a microchip that allows shelters to quickly identify the owner. Therefore, the model yields an excessively long average length of stay since it excludes owner reclaims.

To incorporate owner reclaims into the analysis, I used Tompkins County SPCA’s most recent statistics. Dr. Haston appeared to use Tompkin County SPCA’s adoption length of stay based off his citation of Brown, et al., 2013. While Dr. Haston did not give the full reference of this source, I believe it is this study that takes place mostly at Tompkins County SPCA from 2008-2011 which I use in my own dog analysis. Since I could only find Tompkins County SPCA’s 2018 data, I used this data to compute a revised average length of stay from Dr. Haston’s model based on an assumed 3 days and 32 days average length of stay for owner reclaims and all other outcomes and the percentage owner reclaims made up of total adoptions, total euthanasia and total owner reclaims at Tompkins County SPCA in 2018. A 3 day average length of stay falls into the middle of the range of owner reclaims’ average length of stay I computed from several New Jersey animal control shelters.

After making this adjustment, the model’s average length of stay decreased from 32.0 days to 23.1 days. As a result, the difference between the average length of stay required to avoid exceeding capacity continuously and the model’s average length of stay dropped significantly.

As you will see below, several no kill animal control shelters have dog average lengths of stay around the required average length of stay to avoid perpetual overcrowding implied in Dr. Haston’s model. First, these shelters generally do a better job adopting out dogs than the facility (primarily Tompkins County SPCA from 2008-2011) Dr. Haston used and most likely adopt out dogs quicker. Second, Dr. Haston’s model does not incorporate dogs shelters transfer to rescues. Frequently, shelters can transfer dogs quicker to rescues, especially when the facilities are rescue friendly and make an effort. For example, the Paterson Animal Shelter, which is far from a progressive shelter, transferred a large percentage of all of its dogs as well as pit bull like dogs in 2015 after just seven days on average. Similarly, 2018 Animal Care Centers of NYC data I obtained showed the organization transferred a large number of dogs to rescues in ten days on average. Additionally, my 2017 analysis of Elizabeth Animal Shelter indicated dogs adopted out and transferred to rescues, which mostly were dogs sent to rescues rather than adopted out, spent only 14 days in the shelter. Thus, Dr. Haston’s failure to use role model no kill animal control shelters and dogs sent to rescues in his model makes the model yield inaccurate or skewed results.

Dr. Haston’s failure to include foster homes in his model grossly understates shelter capacity. While most people appreciate the benefits foster programs can have on both the mental and physical health of animals, many don’t realize how much extra capacity these programs can add to a shelter. For example, Dr. Ellen Jefferson provided a goal in a presentation at the 2019 American Pets Alive Conference for animal control shelters to have 3% of their annual dog intake in foster homes at a given point in time. Dr. Jefferson developed this target based on certain no kill animal control shelters’ successful foster programs. Since average length of stay incorporates animals in foster homes, we can add this to the shelter’s dog holding capacity in Dr. Haston’s example. This adjustment increases the shelter’s capacity from 150 dogs to 240 dogs.

As you can see below, the shelter in Dr. Haston’s example will normally have significant excess capacity even without accounting for animals sent to rescues and no kill animal control shelters with stronger adoption programs. While the inherent volatility of dog intake at an animal control shelter, such as a large hoarding case, could temporarily cause capacity concerns, this data shows Dr. Haston’s perpetual overcrowding and related conclusions are simply incorrect.

Real World Data Contradicts Dr. Haston’s Predictions

Dr. Haston’s model predicts a 95% live release rate animal control shelter will have a ridiculously long average length of stay. While his model implies a 32 day average length of stay based on the make-up of dogs brought to the shelter, the model actually predicts an astonishing 90 day average length of stay after one year.

Why does the model predict such a longer average length of stay? Unfortunately, Dr. Haston doesn’t explain whether he is calculating average length of stay for all the animals that came into the shelter during the period or the actual population of dogs in the shelter at a point in time. Assuming Dr. Haston calculated average length of stay of all dogs taken in during the period, which is how shelters typically calculate this metric, the increase in average length of stay from 32 days to 90 days may be due to the shelter exceeding capacity and not accepting all dogs, including many easy to adopt ones. Therefore, the harder to adopt dogs, which take significantly longer to place, will make up a larger portion of the total dog intake and increase the average length of stay.

Even if shelters consistently exceed capacity, which they shouldn’t as explained above, managed admission policies could mitigate that. For example, a managed admission shelter would be more likely to immediately accept an easier to adopt small dog than a larger dog with behavioral issues if the facility was near or at capacity. Therefore, these shelters would have  a much shorter average length of stay than 90 days if these facilities exceeded capacity consistently and restricted intake.

To analyze the Dr. Haston model’s predictions, I compared his model’s key results to actual data from three no kill animal control shelters. These shelters are as follows:

  1. KC Pet Project serving Kansas City, Missouri
  2. Williamson County Animal Shelter serving most of Williamson County, Texas
  3. Lynchburg Humane Society serving Lynchburg, Virginia during the period of my analysis

Due to the lag in non-profit financial data reporting, I had to use 2017 and 2016 data for KC Pet Project and Lynchburg Humane Society, respectively. I listed the links to the data I used in this analysis at the end of this blog.

The average length of stay computed by Dr. Haston’s model vastly exceeds the three no kill animal control shelters’ average lengths of stay. Specifically, Dr. Haston’s model predicts an animal control shelter with a 95% live release rate will have an average length of stay of 90 days while KC Pet Project, Williamson County Animal Shelter and Lynchburg Humane Society had average lengths of stay of 18 days, 9 days and 19 days. In other words, Dr. Haston’s model predicted average lengths of stay five to ten times longer than these three comparable no kill animal control shelters with the same or higher live release rates. Thus, Dr. Haston’s conclusion that a 95% dog live release rate at an animal control shelter will result in the shelter holding large numbers of animals for extremely long times does not match the reality of well run no kill animal control facilities.

Successful no kill animal control shelters also have significantly lower costs than the amounts Dr. Haston’s model predicts. Dr. Haston’s model appears to only include medical and behavior costs in its “operating costs.” Unfortunately, I don’t have this subset of data for the three no kill animal control shelters. Therefore, I used each organization’s total costs, which would include other costs, such as various fixed and overhead costs, that Dr. Haston’s operating costs do not appear to include. To allocate these costs just to dogs, I used each shelter’s annual intake of dogs and cats as well as an estimate of the per animal cost based on average length of stay from the Maddie’s Fund Financial Management Tool. Even using an apparently broader measure of shelter costs, the estimated total costs per dog at KC Pet Project, Williamson County Animal Shelter and Lynchburg Humane Society were $406, $287 and $635 compared to the $750 per dog figure Dr. Haston’s model predicted.

Dr. Haston’s model also understates shelter revenue at no kill animal control shelters. Specifically, Dr. Haston only measures adoption revenue. In reality, adoption fees usually fall way short of covering animal care costs. No kill animal control shelters recoup some of these costs through funding received from the governments running or contracting with them. However, no kill organizations, especially private ones, receive significant donations since the public wants to support shelters that save lives. As you can see below, the estimated total revenue per dog (allocated the same way as total costs per dog above) was $381, $453 and $701 at KC Pet Project, Williamson County Animal Shelter and Lynchburg Humane Society compared to the $176 of adoption revenue per dog Dr. Haston’s model predicted.

The three no kill animal control shelters’ revenue and cost data disprove Dr. Haston’s implicit assertion that no kill leads to financial ruin. Dr. Haston’s model predicted a net loss of around $574 per dog. During the periods presented, both Williamson County Animal Shelter and Lynchburg Humane Society, which received only modest government funding, turned an estimated profit of $166 per dog and $66 per dog, respectively. While KC Pet Project did have an estimated loss of $25 per dog during the year presented, this was an anomaly. Since KC Pet Project was formed in 2011 and began running a no kill animal control shelter shortly thereafter, its net assets increased from $0 to $1,146,550 due to its revenues exceeding its costs over this time period. Thus, Dr. Haston’s model predicting financial ruin at no kill animal control shelters does not match the experience of these three no kill groups.

These three no kill organizations also disprove Dr. Haston’s assertion that a 95% live release rate animal control shelter turns significant numbers of dogs away. According to widely accepted estimates, the average American animal control shelter takes in 14 dogs and cats per 1,000 people. Based on the ASPCA’s estimated total animal shelter intake in the United States, which includes animal control and rescue oriented facilities, approximately half the animals are dogs and half are cats. Therefore, the average American animal control shelter takes in around 7 dogs per 1,000 people. As you can see below, KC Pet Project, Williamson County Animal Shelter and Lynchburg Humane Society took in 12 dogs per 1,000 people, 8 dogs per 1,000 people and 22 dogs per 1,000 people. In other words, these three shelters received more dogs than the average American animal control shelter. While these three no kill facilities do manage intake at times, its hard to argue they are “turning away” significant numbers of dogs and those dogs are having bad outcomes.

The three no kill animal control shelters also disprove Dr. Haston’s prediction that an animal control shelter with a 95% live release rate will do few adoptions. As you can see below, KC Pet Project, Williamson County Animal Shelter and Lynchburg Humane Society adopted out 6 dogs per 1,000 people, 5 dogs per 1,000 people and 18 dogs per 1,000 people. In other words, these three shelters adopt out around as many or significantly more dogs than the average American animal control shelter takes in let alone adopts out.

Absurd Predictions When Incorporating Rescue Oriented Shelters into the Analysis

Dr. Haston laid out one scenario where a rescue oriented shelter in the community took all dogs in when it had room and the animal control shelter had a 90% live release rate. In a second scenario, Dr. Haston assumed the animal control facility had a 95% live release rate and the rescue oriented shelter in the community did not accept the least adoptable dogs (i.e. the dogs an animal control shelter with an 85% live release rate would kill). In the real world, the rescue oriented shelter’s intake policy almost always is more similar to scenario 2 than scenario 1 since most of these organizations pick and choose which dogs they take in. While some of the qualitative results of the first scenario compared to the second scenario make sense (i.e. the animal control shelter in scenario 2 will have a longer average length of stay and higher operating costs than the animal control facility in scenario 1), the actual model’s results do not match reality.

The table below compares Dr. Haston’s animal control shelter’s predicted results under scenario 2 with successful no kill animal control shelters. All three no kill animal control shelters have selective admission rescue oriented shelters in their areas. Therefore, they are operating in a similar scenario to Dr. Haston’s model. As you can see, Dr. Haston’s model predicts an average length of stay 6-14 times longer than these shelters’ average lengths of stay. Similarly, the no kill animal control shelters pulled in 3-5 times more revenue per dog, incurred 40%-73% lower costs per dog and took in more dogs than Dr. Haston’s model shelter did. Thus, Dr. Haston’s model becomes even more absurd after he incorporates rescue oriented shelters.

So how did Dr. Haston calibrate his model to real world results? He contacted 100 no kill shelters across the country about accepting a large dog with behavioral issues and almost all of the facilities did not agree to take the dog in. Since Dr. Haston did not say which shelters these were, I assume these were selective admission shelters. Given we already know selective admission shelters cherry pick their animals, including those near the three no kill animal control shelters above, this is meaningless.

If that was not bad enough, Dr. Haston used his favorite punching bag, pit bulls, at a regressive shelter to validate his model. In a slide titled “Concentration of Difficult Animals in Open Admission Facilities”, Dr. Haston cited pit bulls making up 45% of dog intake and around 25%-30% of dogs killed at Milwaukee Area Domestic Animal Control Commission (MADACC) as evidence supporting his model’s results. First, Dr. Haston citing pit bulls as “difficult” tells you much about his attitude about these animals. While dogs having a pit bull label do stay longer at shelters, a peer-reviewed scientific study proves removing breed labels significantly reduces pit bulls lengths of stay at shelters. Second, MADACC is a regressive shelter that had 21% of their dogs lose their lives last year and 32% of dogs lose their lives in 2014 (one year after Dr. Haston’s MADACC data goes up to). For example, the Wisconsin Watchdog blog detailed the shelter needlessly killing a “pit bull mix” with a potential adopter waiting in 2014. Does anyone in their right mind think this shelter was doing all it could do five years ago? Thus, the idea that rescue oriented shelters put an unfair burden on animal control shelters and that forces them to kill is absurd.

Dystopian Conclusions

Dr. Haston makes a good point that the live release rate is a key metric, but we must also look at other data as well. I fully agree with this. For this reason, my dog report card blog each year also grades shelters on the number of local animals (which often require more effort to save) these facilities take in and adopt out. Additionally, the no kill and animal welfare movements should also create other metrics of success to ensure shelters follow all parts of the No Kill Equation. That being said, the live release rate will always be extremely important given killing animals is intolerable.

Unfortunately, Dr. Haston repeats the false notion that raising the live release rate from 85% to 95% results in longer lengths of stay, increased costs and refusing animals. While I know some shelters do severely restrict intake in order to raise their live release rates, that is not what well-run no kill animal control shelters do. As the three no kill animal control shelters’ data above showed, large no kill animal control shelters take many dogs in, save around 95% or more of these animals and do so in a financially responsible way. Can a shelter have a shorter average length of stay and lower costs if it settles for an 85% live release rate and quickly kills every challenging dog? Yes, that is likely. However, the three no kill animal control shelters’ data above prove you can still achieve a very short average length of stay and have manageable costs at a 95% or above dog live release rate. Additionally, no kill animal control shelters’ revenue surge when the public realizes these facilities are doing everything possible to save their animals. Thus, Dr. Haston’s thunderous conclusions about doom and gloom for animal control shelters achieving around 95% live release rates are wrong.

Most disturbing, Dr. Haston describes an “optimal” live release rate where killing is not only acceptable, but desirable. In essence, Dr. Haston says we should quickly kill “difficult” animals, such as pit bulls, and take in more easy to adopt dogs. In other words, shelters should operate more like pet stores instead of doing the necessary work to save “difficult” animals. While Dr. Haston doesn’t explicitly state this in his presentation, he did say “we can’t adopt our way out of” the so-called pit bull problem in a presentation he recently gave. Furthermore, Dr. Haston’s 2015 presentation stated saving more lives may mean sacrificing the individual.

Sadly, Dr. Haston’s myopic view need not be true. While shelters will adopt out more easy to adopt dogs all else being equal, all else is not equal. As the no kill movement spreads, the innovative policies will spur positive change in many organizations. As organizations improve, they will responsibly reduce dog intake, increase live outcomes and therefore rescue more at risk animals. By contrast, Dr. Haston’s narrow view only allows shelters to increase adoptions by having easy to adopt animals. That is a recipe for stagnation.

What happens when shelters run out of these easy to adopt animals in the future? Apparently, they may work with “responsible breeders.” According to a recent Animal Farm Foundation Facebook live video (starting at 11:00 minute mark), the 2019 HSUS Animal Expo conference had a session on doing just this. Specifically, shelters would have “responsible breeders” breed desirable dogs for “gold level adopters” since the shelters would be filled with those “difficult” to adopt dogs “nobody wants” like pit bulls. While I can’t say Dr. Haston supports this, it is a logical extension of his kill the “difficult” dogs and adopt out the easy dogs philosophy.

At the end of the day, Dr. Haston’s and many so-called shelter leaders’ anti-no kill views are based on a deeply flawed model. Not only do the model’s conclusions violate basic ethical values, the actual quantitative predictions fall apart when we compare them to well-run no kill animal control shelters. Clearly, no organization should consider this a prediction of what real no kill sheltering looks like. Instead, shelters should consider the model useful if they attempt to implement no kill the wrong way. If that happens, then the model could show what will happen. However, Dr. Haston does not present his model this way and declares no kill/high live release rates a disaster. Sadly, Dr. Haston’s messaging ruins what could be a very good way to illustrate the perils of not implementing no kill the right way. As a result, Dr. Haston’s model will be used by lazy shelter directors to defend the status quo and not improve.

Appendix – No Kill Animal Control Data Used in Comparison to Dr. Haston’s Model

KC Pet Project 2017 Animal Data

KC Pet Project 2016 Dog Average Length of Stay (2017 data not available, but unlikely to differ significantly)

KC Pet Project 2017 Form 990 Tax Return

Williamson County Animal Shelter 2017-2018 Animal Data, Dog Average Length of Stay and Financial Information

Lynchburg Humane Society 2016 Animal Data

Lynchburg Humane Society 2016 Dog Average Length of Stay

Lynchburg Humane Society 2016 Form 990 Tax Return

Maddie’s Fund Financial Management Tool to Estimate Cost to Care for Dogs

One thought on “Roger Haston’s Ridiculous Kill Shelter Model

  1. Killing innocent animals due to homelessness and being the fault of humans failing them is hypocrisy. When we have PRISONS FULL of useless humans who don’t deserve life, are kept alive by us,yet you want the innocent animals to die! WHAT IS WRONG WITH YOU!

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s