Dog Longevity 

Created by Dr. Kelly M. Cassidy, 2007





Home Study Citations Breed Data Longevity Trends Survey Comparisons Weight and Lifespan Perception v Reality Surveys - MHO Conclusions Contact

Perception and Reality

Updated 07/04/2007


In the process of creating the tables of longevity data, I noticed that the many sources of breed longevity estimates were overly-optimistic.  They often reported an expected longevity that greater than longevities measured in questionnaire surveys.  I decided to quantitatively compare estimates of breed longevity to the available survey data.  

[Note: Unfamiliar with statistics?  As a rule, a relationship is not considered "statistically significant" unless p, the probability,  is less than 0.05, which means there is less than a 5% chance the relationship is due to chance alone.  r2, the regression coefficient, is a measure of how much of the variance in the dependent variable (the discrepancy in longevity) can be explained by variation in the independent variable.]


Sources of longevity estimates

I found three sources of information that gave expected life spans for a large number of breeds.  The sources were recent (2005 to 2007)  and appeared to be independent of one another, i.e., none seemed to be echoing the expected life spans from either of the other two sources.  One source was an encyclopedic book of dog breed information (Coile 2005). The other two were general dog web sites:

Coile, D. C. 2005. Barron's Encyclopedia of Dog Breeds.  2nd edition. Barron's Educational Series, Hauppauge, NY. Retrieved April 25, 2007. Dog Longevity by Breed.

DNA Diagnostics Center (DDC), Animal DNA Services. Retrieved April 30, 2007.

Comparing estimates and data

For measured longevity, I used the average longevities from all the questionnaire surveys for which the total sample size was at least 20 dogs.  Averages of multiple surveys were weighted by the sample size in each survey (i. e., the blue-colored column on the Breed Data page).

For estimated longevity, I used the average of the longevity range given by the source.  For example, if a source gave Breed A's lifespan as "8-10 years," I used 9 years as the source's estimate of average longevity for that breed.

The number of breeds compared varied by source, depending on whether the source included an estimate of longevity for the breeds for which I had data.  For Coile, I was able to compare 115 breeds; for Dog-Answers 111 breeds, and for DDC 93 breeds.

Estimate Discrepancy - I refer to the difference between the longevity estimate and the breed longevity measured in surveys as the estimate discrepancy.  I call it a "discrepancy" instead of an "error" because error has a statistical meaning and because it is possible that surveys have a systematic bias.  If the estimated longevity minus the longevity measured in questionnaire surveys was positive, then the estimate discrepancy is an overestimate. 


Average magnitude and direction of estimate discrepancy

Source Number of breeds compared Average questionnaire longevity of compared breeds (yrs)  Average estimated longevity of compared breeds (yrs) Average estimate discrepancy (yrs)
Coile 2005 115 11.1 1.9 12.0 1.6 0.9 1.6
Dog-Answers 2007 111 11.0 1.9 12.5 1.2 1.6 1.5
DDC 2007 93 11.0 2.1 12.2 1.9 1.1 1.6

All sources overestimated breed longevity. Coile was closest to being correct, overestimating by an average of 0.9 yrs, but DDC was nearly as accurate. Dog-Answers overestimated by an average 1.6 yr.   

Relationship between estimate discrepancy and data sample size

I hypothesized that there might be a relationship between sample sizes in surveys and the estimate discrepancy.  Perhaps surveys with small sample sizes tended to be skewed in one direction.

Source r2 for regression of estimate discrepancy and survey sample size F statistic p (probability)
Coile 2005 0.012 1.38 0.25
Dog-Answers 2007 0.009 0.94 0.34
DDC 2007 0.038 3.57 0.06

Result: The estimate discrepancies were not significantly related to survey sample size for any of the estimate sources, although DDC was nearly significant.  Even in that case, however, only 4% of the estimate discrepancy could be attributed to sample size.  For the other two sources, only about 1% of the discrepancy could be attributed to sample size.

Relationship between estimate discrepancy and breed weight

I hypothesized that the estimate error might be related to breed size.  Perhaps sources consistently overestimated or underestimated longevity of large or small dogs.

Source r2 for regression of estimate discrepancy and breed weight F statistic p (probability)
Coile 2005 0.0004 0.04 0.83
Dog-Answers 2007 0.031 3.54 0.06
DDC 2007 0.002 0.14 0.71

Result: There was no significant relationship between breed weight and the estimate discrepancy.  Estimates were as accurate (or inaccurate) for small breeds as large breeds.

Relationship between estimate discrepancy and measured breed longevity

I hypothesized that the magnitude of the estimate discrepancy might be related to measured breed longevity.

Source r2 for regression of estimate discrepancy and measured longevity F statistic p (probability)
Coile 2005 0.33 55.5 < 0.0001 (highly significant)
Dog-Answers 2007 0.56 139.3 < 0.0001 (highly significant)
DDC 2007 0.23 27.5 < 0.0001 (highly significant)

Result: There was a highly significant relationship between the estimate discrepancy and measured breed longevity.  For all three sources, a substantial percentage of the variability in the estimate discrepancy could be explained solely by measured longevity.  For Dog-Answers, the relationship was particularly pronounced, with more than 50% of the variability explained by the measured longevity.

Because this relationship was so pronounced, I plotted estimate discrepancy (estimate minus measured longevity) against measured longevity for the three estimate sources.  In the plots below, the points above the zero line are the overestimates. Those below the zero line are underestimates.

For the shortest-lived dogs, those with a surveyed longevity of 5-7 yrs, Coile and DDC overestimated breed longevities by an average of about 2.5-3.5 yrs, but with very large amounts of scatter in the data.  All sources had one or two breeds in which the overestimate was more than 5 yrs for breeds with measured longevities of  6-8 yrs--an estimated longevity nearly twice the measured longevity.

 For breeds with more "normal" longevities of 10-13 yrs, estimates were usually within a year or two of measured longevities, but more often overestimated than underestimated. 

For long-lived breeds (those with surveyed longevities of >13 yrs), estimates were usually within a year of the measured longevities, but longevity was more often underestimated than overestimated.

In brief, there is a general belief that short-lived dogs live longer than they really do and that exceptionally long-lived breeds do not live as long as they really do.


Summary of Results

  1. On average, estimates of breed longevities were 0.9 to 1.6 yrs greater than longevities measured in surveys
  2. There was a highly significant, inverse relationship between measured longevities and the size of the discrepancy between estimated and measured longevity.  The shorter the breed's measured longevity, the greater the discrepancy.    
  3. The discrepancy was not significantly related to survey sample size or breed weight.  

The Relationship between Estimated and Measured Longevity

There are several possible explanations for the inverse relationship between survey longevities and the magnitude of the discrepancy between estimated and surveyed longevities:

  1. Most breeds have a surveyed longevity of 10 to 13 years (see the Weight and Longevity page). Most dog owners' personal experience with dogs will be with dogs that die within that range, so the tendency may be to extrapolate from the usual experience to less commonly encountered breeds.  This potential explanation seems unlikely, since the club web sites for short-lived breeds, presumably authored by people with much experience with the breed, often report expected longevities that are as overly optimistic as the encyclopedic dog references.
  2.  The shorter-lived a breed, the greater the reluctance among breed fanciers to accept how short-lived the breed really is. Hence, the shorter-lived the breed, the more "padding" is added to the longevity estimate. Breed fanciers may like to believe they have a healthy breed and might be inclined to feel the longer-lived individuals are more representative of the breed than dogs that die young. 
  3. The estimate was correct several dog generations ago, but some breed longevities have declined and there is a lag time for perception to catch up with reality.  There is some evidence that breed longevity is declining (See Longevity Trends page), but it is weak because of the lack of comparable data for earlier times.  This possible explanation, if true,  would suggest that longevities of many breeds are declining "under the radar" of those most familiar with the breed.
  4. There is a dog owner tendency to feel ashamed or guilty if their dog dies young, so owners avoid mention of former dogs that have died young.  Conversely, owners of long-lived dogs may attribute their dog's long life to their excellent care.  (These proud owners of long-lived dogs are probably partly correct, but good care cannot completely compensate for a dog's bad draw of the genetic deck.  Many owners of short-lived dogs give their dogs excellent care.)  The combination of vocal owners of long-lived dogs and quiet owners of short-lived dogs could give the impression that long-lived dogs are the norm.  The anonymity of a survey helps reveal the true number of shorter-lived dogs in a breed.  Anecdotal support for this possible explanation can often be seen on breed email lists.  When a dog dies at an early age of cancer, for example, the owner will often publicly agonize over whether they might have used too many lawn chemicals, over-vaccinated, fed the wrong food, etc.  They often seem convinced their dog's early death was an anomaly that must be explained by anything but genetics.  I have also noticed (again, anecdotally since I have not done any analysis) that, among members of breed email lists that name former dogs in their signature files, owners of dogs that died young rarely list those dogs' ages at death while members that have had long-lived dogs often include age at death.   


Why is perception important?

Why is the difference between perception and reality important? After all, the dog will live however long it lives, regardless of expectations.  One important reason is the dog owner.  Many dog owners avoid short-lived breeds to reduce  the number of times they must face the inevitable trauma of losing a dog.  Dog owners planning on a performance (hunting, agility, police work, service work, etc.) dog can expect to invest considerable resources (emotional, financial, and time) into training the dog. They may be loathe to invest such resources in a dog with a high chance of dying before there has been much return on the training.  Breed longevity is not important to all prospective dog owners, but, if it is, the puppy buyer often wants realistic information.

Longevity information is also critical to dog breeders for whom longevity is important.   If longevity is declining, those breeders need to know soon enough to react before loss of genetic diversity makes a reversal of the trend impossible without crossing to another breed.  Breeders also need to know whether expensive strategies to improve breed health are actually paying off.  For example, testing of dogs for various conditions or particular alleles is increasingly expected of responsible breeders prior to breeding.  The number of tests expected was only recently one or two, but is rapidly expanding in many breeds.  For standard poodles, for example, a test for hip dysplasia was once the only expected test, but breeders now may do phenotypic tests for sebaceous adenitis, Addison's disease, and atrial septal defect and genetic tests for von Willebrand's disease, neonatal encephalopathy, and juvenile renal disease.  A breeder that does all of these tests may spend $1,000, but still won't be covering  some of the worst afflictions of the breed for which there are no tests, such as bloat/torsion and early onset cancer.   There is no evidence yet that all this testing is improving health or longevity of standard poodles.

Where does a prospective dog owner go for real data?

If a dog owner or breeder can't find any longevity data for a breed, what source should they trust?  Unfortunately, no estimate is reliable.   Breed club websites are no more reliable than encyclopedic sources.  Unless the breed club or encyclopedic estimate is based on actual data, it should be viewed with skepticism.  Also, a puppy buyer should keep in mind that all breeds have individuals that live unusually long lives, so trumpeted examples of long-lived members of the breed do not mean the average member of the breed is long-lived.   

If a source says a breed lives 13 years, you cannot know whether it is actually a long-lived breed whose longevity is being underestimated by 1 year or a very short-lived breed whose longevity is being overestimated by 4 or 5 years.  In the absence of data for a particular breed, the best fall-back guess is the median longevity of all pure-bred dogs, which is about 10 to 11 years.      


Hit Counter