“Worthless” was how Michigan's annual report on the registration of births, marriages, deaths and divorces described the antiquated law for enumerating births that was in place from 1867-1905.  Not until 1906 could the returns of birth in Michigan be considered accurate. Only then were birth certificates required to be filled out proximal to the time of birth. Before then, the process for recording births in the state was, well . . . not exactly what a good genealogist would hope for.
So, how were birth records for Michigan collected prior to 1906? Once a year the supervisors or assessors canvassed their districts to conduct the annual assessment, lasting from mid-April to the close of May. At this time they also asked about any births or deaths (until 1897 when a death certificate was required to be filled out before a body's removal or burial) that had occurred in the previous calendar year. So, a worst case scenario for recording a birth might look like this: “As an extreme case, a birth which occurs in January may not be noted by the supervisor until May of the following year, or sixteen months after it occurred. Is it any wonder that many births are overlooked and forgotten, even when supervisors honestly try to do their duty? And when they perform it in a merely perfunctory manner, without personal house-to-house inquiry after children, the amount of imperfection must be even greater.” 
“The enumeration of births by the supervisors of Michigan. . . is little more than a farce, so far as completeness is concerned. . . Not all of [the births] could possibly be obtained by an enumeration conducted after the close of the year, as some families have moved out of the districts where the births occurred and others have forgotten the exact dates of birth, etc. At present, however, the enumeration made by many supervisors is merely perfunctory. All of the domiciles in the township or village are not visited, and sometimes none of them, the list of births being jotted down in the supervisor's office from recollection. Not only are many births entirely omitted from the records but even those registered, especially after they have been copied in the county clerks' offices, are often grossly incorrect in important personal and statistical details. In some years the returns have been falsified deliberately in order to obtain increased compensation for returns” [they were paid by the number transcribed]. 
Not only were the births not recorded at the time of birth, but even after they had been collected it took a while for them to make their way to the State, during which time records could have been lost. The Michigan annual report again minced no words on this topic: “for births. . . it would seem that a systematic system of procrastination was authorized by law.”  “And when the law is dilatory, officials acting under it may be trusted to add another measure of delay, on the principle that as the returns are tardy anyway, a little more delay will not hurt them.”  Once the births were turned over to the county clerk from the assessor they were permitted three months time to copy and transcribe the records before sending them on to the State or until nine months after the close of the previous year. “Some of the births have been over nineteen months on the way. The copies of the copies of the supervisor's hearsay reports are frequently inaccurate when received, and further time is necessary in endeavoring to remove some of the most obvious blunders.”  These, then are the state of the records that we genealogists depend upon to source our trees.
“This 'census method' is universally recognized as a failure for the purpose of collecting vital statistics,” declared the Michigan annual report on births, etc. and yet it took until 1906 for it to be changed. [5,6] A bill was brought before the state legislature in 1899, but received some push back. Some felt the new law would be an invasion of privacy, particularly in the case of illegitimate births or stillborns.  The fact was that many illegitimate births and stillborns were already on the books and in the opinion of the Secretary of State's office “it would not seem to be a matter of greater indelicacy to record a stillborn child or an illegitimate birth immediately upon the event than to wait a year or so before making the record.”  As far as recording the birth of a stillborn was concerned, the new registration law for deaths required it so why would recording the birth of that child be any more controversial.
I don't know exactly how birth records (or death records) were collected in other states prior to the requirement of certificates, but I wouldn't be surprised if it was similar to the system used in Michigan prior to 1906 (for births) and 1897 (for deaths). Some people might think they have a primary source if it is stamped with the official seal of a state, but this foray into birth records demonstrates the importance of understanding where a record came from. Knowing the circuitous path the information traveled before it landed on a roll of microfilm, if I have to choose between a family bible or church baptism record and an early state record, I know which one I'll put more faith in.
- Secretary of State of Michigan. Thirty-third Annual Report of the Secretary of State on the Registration of Births And Deaths Marriages And Divorces in Michigan For The Year 1899. (Lansing, Michigan: Wynkoop Hallenbeck Crawford Co., 1902), clvix.
- Secretary of State of Michigan. Thirty-Second Annual Report of the Secretary of State on the Registration of Births And Deaths Marriages And Divorces in Michigan For The Year 1898. (Lansing, Michigan: Robert Smith Printing Co, 1900), viii.
- Secretary of State of Michigan. Thirty-Second Annual Report . . . 1898., clix.
- Secretary of State of Michigan. Thirty-Second Annual Report . . . 1898., vii.
- Secretary of State of Michigan. Thirty-Second Annual Report . . . 1898., v.
- Physicians Say Good Thing. Law Regarding Certificates of Birth. Kalamazoo Evening Telegraph. Kalamazoo, Michigan. 6 January 1906, page 10, column 4.
- Secretary of State of Michigan. Thirty-Second Annual Report . . . 1898., ix.