I am not saying you don't have a good source. I am saying these numbers are not going to be exact.

EX: The 2017-2018 APPA National Pet Owners Survey

https://americanpetproducts.org/Uploads/MemServices/GPE2017_...

Overall dog ownership in the US at 48% +/- 4.4 based on 505 completed surveys. But, that's not nearly enough data to slice things down to 30-40 year olds and worse it's a servery which have dubious accuracy overall.

But that's not the survey they used. A search on Mintel reveals surveys on pet ownership that seem targeted for consumer research and pet industry (eg: https://store.mintel.com/americas-pet-owners-us-august-2018 and https://store.mintel.com/us-americas-pet-owners-market-repor...).

Their methodology as outlined on their web site (https://store.mintel.com/mintel-methodology) and details the quotas they utilize to get what they feel is a statistically relevant cross section of data.

Anytime sampling is used, of course it's not going to be exact. But a 2000 person survey is, given equal statistical rigor, going to be better than a 500 person online survey and might be enough of a population to justify more detail.

(Unfortunately, the Mintel survey results cost $4300, so other than their methodology page it's going to be difficult for me to infer much more. But the methodology page seems to suggest they aim for +/-2 or 3% confidence level.)

First it was 500 dog owners out of ~1000 respondents ~= +/— 4.4%. So, Double that to 2000 people and you can get 3% accuracy on overall per ownership, assuming no systemic bias. But, slice that into 30-39 year olds and you have ~1/7 the data and thus much higher error bars.

Now, if they had sampled ~20,000 people then sure you can slice and dice like that. But slicing data always increases the risk of sampling bias as would anything that targets 30-39 year olds specifically.

You work for an e-rag.

We've banned this account for violating the site guidelines.