Dating the introduction a new product or technique into the economy is critical for understanding why growth happens. However, there is no perfect measure of “innovation”. Patent records list only the date of invention, not the date an invention was commercialized into a viable product, or used outside of a lab or researcher’s mind. Further, in many sectors, a huge amount of innovation goes unpatented, and much important innovation is simply unpatentable – many business processes, for instance. Other records of innovation used by economists – World’s Fair records, manually compiled innovation data, etc. – tends to either be limited in its scope and duration of coverage, or else lends itself to subjectivity on the part of the compiler. Alexopolous and Cohen show that, in many cases, the number of book publications, annotated with their subject by the Library of Congress, accurately track innovations. For instance, after 1937, neoprene was the subject of a number of technical books and guides, accurately tracking its widespread commercial introduction in that period. Library of Congress book data is also machine-readable, so little effort is needed from researchers who wish to use such data. The measure is not perfect, of course – unlike patent citations, there is no way to measure the geographic spread of an innovation with book publication data – but given the paucity of useful data on invention and innovation, the technique in this paper strikes me as a useful one.