Americans own 70 million dogs and 74 million cats, and pet ownership is waning. Or no, wait. It’s 83 million dogs and 96 million cats, and pet ownership is at a record high.

These contradictory messages come from two organizations that separately track the population of pets in U.S. households and are cited equally as authoritative sources.

Dog and cat counts by the American Veterinary Medical Association (AVMA) and the American Pet Products Association (APPA) never have matched exactly in 17 years of their asking similar questions in pet ownership surveys of American households. But the gaps in their most recent tallies yawn wider than ever.

The dog counts diverge by 13 million; the cat counts, 22 million. That's as many cats as the AVMA estimates live in California, Texas, Florida and New York households combined.

The inconsistent numbers yield opposite conclusions about pet ownership in America. The AVMA reported in 2012 that pet ownership had declined, reversing a trend the professional association said had lasted at least 20 years.
 
Six months later, in early 2013, the trade organization APPA said its latest survey showed pet ownership at an all-time high.

The number of pets in American households and prospects for population growth or decline matter to a range of industries, sectors and interest groups, including the veterinary profession and businesses that support veterinary medicine; animal-welfare organizations; local governments with animal control and shelter operations; and makers and merchants of pet foods, drugs, accessories and related merchandise.
 
From the perspective of veterinarians, for example, if the number of pets is increasing, then demand for veterinary services conceivably will rise. Conversely, if the number is shrinking, then demand conceivably will fall. Which way demand is headed is especially pertinent to the profession today owing to concerns about an excess of companion animal veterinarians in the United States.

Having reliable figures on the population of pets is important, too, to animal-welfare advocates, many of whom mark success by individual lives saved.

“We’re well aware of the discrepancy (in statistics), and it’s a bit frustrating,” said Aimee Gilbreath, executive director of Found Animals, a nonprofit organization in Los Angeles. “For a variety of programs, whether they are for dog licensing or spay and neuter, it would be useful to know how many dogs and cats there are ... so you know how much you have to do to move things.”

The APPA and AVMA acknowledge that their results differ but each expresses confidence in its figures.

Anna Ferrante, APPA senior vice president for member relations and business development, said by email: "While we cannot comment on research done by other organizations, any variations are most likely due to differences in methodology, such as sample size and composition, question wording, and survey frequency, to name just a few.” She added, “APPA has duplicated the Pet Owners Survey results in subsequent studies, and therefore is confident in the data."

Similarly, Sharon Granskog, a spokeswoman for the AVMA, said by email: “I know that some of the differences … are due to different samples.” She, too, mentioned differences in question wording.

To explore why the AVMA and APPA survey results deviate markedly, the VIN News Service examined the methodology of each and consulted experts in polling.

For its surveys, the AVMA hires Irwin Broh Research, while the APPA contracts with Ipsos. AVMA surveys are conducted every five years; APPA, every two years. Both have fairly high numbers of respondents.  In the most recent polls, 50,347 responded to the AVMA; 25,109 responded to the APPA. By comparison, public opinion and political polls such as the CBS News Poll typically survey roughly 1,000 people.

Although the AVMA netted twice as many respondents as the APPA, statistical theory says that both sample sizes are large enough to yield small margins of error — more than 95 percent confidence that the results are accurate to within 1 percent — provided the surveys are based on random samples of the entire population.

In previous years, both surveys were conducted by mail, but the most recent editions were conducted online. The change in approach coincides with a marked jump in the size of the discrepancies between the survey results.

The difference in the past was as small as 3 percent and as great as 12 percent. The latest surveys produced a discrepancy of 19 percent in the number of dogs, and 29 percent in the number of cats.

Don Dillman, a survey expert and sociology professor at Washington State University, said both surveys are likely to produce unreliable results in part due to their use of online panels, which he said can give skewed samples of the population.

“My suggestion would be to be really careful with these numbers,” Dillman said.

Here’s why: When conducting a survey by mail, it's possible to start with a list of nearly every physical address in the nation — provided by the U.S. Postal Service, with names removed — and then select households to survey.

Online methods, by contrast, don't reach significant chunks of the population, particularly older and low-income households. At least 15 percent of American adults do not use the Internet, according to a 2013 report by the Pew Research Center. In addition, databases that correlate email addresses with physical locations are far less complete than the Postal Service's address list, Dillman said, so it's not possible to create a truly random sample even of all Internet-connected households.

Further, he said, common problems exist with the types of volunteer online panels used for the APPA and AVMA surveys. These panels are collections of hundreds of thousands of people who have provided basic demographic information — mailing address, age, income, gender, race, household size, marital status, education and employment — and signed up to complete online surveys in exchange for modest rewards. Hours of survey-taking are required to earn a $10 gift card. (Anyone interested may volunteer for the Ipsos panel, the panel used by Irwin Broh, or at any of at least 20 similar websites.)

“They have the aura of being scientific because they have so many respondents,” Dillman said. “… But they’re just not good science.”

The trouble with online panels, he said, is that they tend to be composed of people who have spare time and Internet access and for whom a small reward is a significant incentive — and that's not likely to be a representative slice of the national population.

The two surveying companies tried to account for that. The methodology sections of the AVMA and APPA survey publications state that demographic information about panel members was used to adjust data from survey respondents to produce results that are valid for the U.S. population as a whole. For instance, if certain age groups of people are underrepresented among survey respondents, their answers are given greater weight when calculating survey results.

But Dillman said data-weighting is highly reliable only when there are well-established correlations between demographic traits and the variable being surveyed. For instance, data-weighting in election-day exit polls works fairly well because the demographics of voting have been studied exhaustively. Pet ownership, he said, likely doesn't fall into that category.

The upshot, Dillman said, is that both pet ownership surveys may be significantly biased.

Ipsos, Irwin Broh and MySurvey.com, the panel used by Broh, did not respond to requests from the VIN News Service for more information about their techniques. Without more information, Dillman said it is impossible to assess whether one survey is more reliable than the other. "I don't think a scientist can make anything out of it," he said.

Granskog at the AVMA acknowledged that the switch to online survey techniques may have accounted for some of the shift in its survey results. "It is certainly something we will be looking at as we move forward with the next survey, which will also be via an electronic sample," she said by email.

Specific numbers not ‘God-given’

Andrew Rowan, chief scientific officer of the Humane Society of the United States (HSUS) and a former director of the Center for Animals and Public Policy at Tufts University Cummings School of Veterinary Medicine, has followed animal population numbers for years and is highly familiar with the discrepancies. They caught his attention even before the release of the latest figures with the largest variance.

In 2008, Rowan convened a workshop in Houston involving the AVMA, APPA and others in an attempt to reconcile divergent statistics. In a report he distributed afterward to participants and subsequently shared with the VIN News Service, Rowan wrote:

“…it was generally accepted that the AVMA pet population estimates from their quinquennial surveys are likely to be more reliable than the biennial surveys used to produce the APPMA demographic reports. This is because the APPMA is mainly focused on a detailed analysis of pet owner behavior and purchases rather than estimating national population trends.”

(The report refers to the APPA as the APPMA because at the time, the trade organization went by the name American Pet Products Manufacturers Association.)

A spokeswoman for the APPA said it could not comment on the meeting, but that the group stands behind its survey "100 percent.”

Despite Rowan’s understanding that the AVMA results are likely to be more reliable, the HSUS presents APPA figures on its website. Rowan said that decision was made by his organization’s companion animal section. “I’ve talked to them about it, but they do it,” he said in a telephone interview. “I don’t feel strongly enough about it to push for a change.”

Rowan explained that the APPA report appeals to HSUS staff because it is updated more frequently than the AVMA’s and includes a variety of data. “It comes out every two years in a big, professional-looking book. It’s got lots of interesting information, so they use it,” he said.

(Following the interview, Rowan reported that the HSUS plans to change the website reference.)

The APPA figures are useful for tracking trends in pet ownership, if not exact numbers of pets, Rowan added. In the 2008 report, he wrote: “The reason these trend data are reliable is because the methods used were broadly the same each time the survey was done. Therefore, even though the APPMA data might overestimate dog and cat populations, the method would have the same bias towards overestimation each time.”

He told the VIN News Service: “You don’t want to take each number as a particularly God-given number. You want to look at the trends.” Looking at APPA and AVMA data since their surveys began in the 1980s, he said, “Both show an increase (in pet ownership). But the AVMA data is much lower than the APPA data. I suspect the APPA is quite comfortable having it be high,” he added. “It serves their purposes.”

As for the conflict between the AVMA’s latest finding that pet ownership dipped and the APPA’s finding that pet ownership continued to climb, Rowan suspects the AVMA’s two most recent results are the problem.

“I’m fairly confident their 2011 number is low, and their 2006 number is high, and the real number is somewhere in between,” he said. Why? Because looking back 20 years, he said, the survey’s trend line is “almost exactly parallel” with the number of households in the United States. Rowan does not believe the recent economic slump drove a downturn in pet ownership. The popular perception that many Americans couldn’t afford to keep pets during the bad times, Rowan said, isn’t borne out by databases tracking shelter activity.

Inconsistent results among surveys not unusual

Another organization that uses AVMA and APPA data is the Bureau of Labor Statistics in the U.S. Department of Labor.

In an article published last May, “Spending on pets: ‘Tails’ from the Consumer Expenditure Survey,” the bureau obtained from the AVMA the proportion of American households with pets, and from the APPA, the total number of pets in the United States.

Steve Henderson, chief of the information and analysis branch in the Consumer Expenditure Survey Division and author of the article, didn’t recall noticing discrepancies between population figures from the two sources. “I did not use the information in a way that I needed to figure out one number over the other,” he said in a telephone interview.

Learning that the surveys produce inconsistent answers didn’t surprise him, though. Henderson observed that surveys on household and personal spending conducted separately by his agency and by the U.S. Department of Commerce don’t always yield consistent information, either.

That’s to be expected to some degree, he said, due to differences in the questions and when they're asked. Thinking about pets, Henderson noted that asking “Do you have any pets now?”  and “How many pets have you had during the past 12 months?” could elicit different answers.

In fact, the AVMA and APPA do ask about pet ownership in different ways. The AVMA asks two questions: whether a household owned a pet on Dec. 31 of the previous year, and whether a household owned a pet at any point in the previous calendar year. (In summaries of survey results, it usually gives the number derived from answers to the Dec. 31 question.) APPA asks whether a household currently owns a pet.

Beyond the AVMA and APPA, no other organization attempts to tally the nation’s pets regularly.

The U.S. Census Bureau doesn’t ask about pet ownership during its once-a-decade count of the American populace. The Labor Department tracks pet ownership only peripherally through its Consumer Expenditures Survey. “We don’t ask if people have pets,” Henderson said. “We ask, ‘Did you buy any pet food?’ ‘Any trips to the vet?’ ”

Asking about pets and pet ownership “would be fun,” Henderson said, but the survey, which is conducted in person continuously through the year, already is long. “(It) takes about an hour, so it would take longer,” which potentially would add cost, he said.

Five tallies, five different answers

Counting companion animals appears to be difficult even in a much smaller geographic region than the United States. When the pet welfare advocacy organization Found Animals set out to determine the population of dogs and cats in the city and county of Los Angeles, it identified five sources of figures. None matched.

“The estimated dog and cat populations for the City of Los Angeles and Los Angeles County vary drastically and are nearly impossible to calculate accurately,” states the report published in 2009.

“More specifically, dog and cat population estimates for the City of Los Angeles range from the 200,000s to nearly 900,000 for each species. In Los Angeles County, estimates range from approximately 700,000 to over 2 million for both cats and dogs.”

The sources Found Animals consulted are the APPMA, the AVMA, the City of Los Angeles Animal Services, 1-800-SAVE-A-PET and Zogby International.

Gilbreath, executive director of Found Animals, said she decided after that frustrating exercise that the only way to obtain reliable data for a specific community is to do one’s own survey.

Found Animals has not done so, preferring to spend its money on spay and neuter and adoption programs, Gilbreath said. It makes do with secondary data such as pet product spending or the number of pet supply stores in the community.

“Either you have to be willing to look at secondary sources, or be willing to develop a program to get the data yourself,” she concluded.