This article presents a puzzle. Namely, why do the internet and search engines make search more like word of mouth? I turn to the data on patents to lay out this puzzle.

The patent data contains a variable called mean forward lag. Suppose Patent A is later cited by another patent, Patent B. From the perspective of Patent A, the difference in time between Patent A and Patent B is called the forward lag.

Documents that cite a patent are usually at least 12 months newer than the cited patent due to a lag in publishing of filed applications. Therefore, patents that are published at the same time as another patent will rarely cite each other. (Source: http://www.acclaimip.com/how-to-use-the-patent-forward-citations-in-acclaimip/ )

I take advantage of the forward lag variable to identify patents that learn about and cite other patents through word-of-mouth. If the mean forward lag is less than or equal to 0.5 years, I label the patent “Word-of-Mouth.” The reasoning is that, due to the lag in publishing, the cited patent will not be published yet by the application date of the citing patent. Furthermore, before the invention of the World Wide Web in 1990, the only way to hear about another patent that is being published at the same time is through word of mouth as opposed to search. If the mean forward lag is more than 0.5 years, I label the patent “Search.”

Figure 1, below, compares word of mouth patents and search patents along a few dimensions. The dimensions are: 1) Originality is high if the the patent cites previous patents that belong to a wide variety of fields, it is low if the patent cites a narrow set of technologies; 2) Generality is high if a patent is cited by subsequent patents that belong to a wide range of fields, it is low if subsequent patents belong to a narrow set of fields; 3) The number of claims represents the “scope” or “width” of an invention; 4) Citations made is indicative of knowledge spillovers. If Patent B cites Patent A, it signifies knowledge flowing from A to B; 5) Citations received is an indicator of the importance of a patent; 6) Self citations represent knowledge transfer that is internalized, as opposed to citing other patents, which represents diffused spillovers of knowledge. (Source: http://people.brandeis.edu/~ajaffe/Copy%20final%20of%20hall-jaffe-traj.pdf )

Figure 1 highlights differences along the above mentioned dimensions between word of mouth patents and search patents, pre-internet, but this article primarily focuses on differences in the generality dimension. The mean score for generality for word of mouth is .05. The mean score for generality for search is .32. The percentage difference in generality is 85%, which is the largest differential between search and word of mouth among the dimensions presented. Therefore, patents that are spread through word of mouth are much more specialized than patents that are spread through search. There is also a large differential in citations received, 81%. This article does not drill down on this dimension, but word of mouth patents are not as “important” as search patents.

Figure 1.  Difference between Word of Mouth and Search Pre-Internet (1975-1990)
	Word of Mouth (Mean)	Search (Mean)	Percentage Difference	T-test Statistic	Significant?
Originality	0.28	0.32	11.72%	5.44	Yes
Generality	0.05	0.35	84.78%	46.74	Yes
Claims	9.98	10.79	7.56%	3.77	Yes
Citations Made	5.02	6.18	18.78%	9.98	Yes
Citations Received	1.31	7.03	81.31%	30.37	Yes
Self Citation Upper Bound	0.20	0.14	-46.73%	-9.28	Yes

The advent of the internet and search engines seems to make search patents more specific and less general, but word of mouth patents seem to have gotten more general over the years. Chart 1, below, demonstrates this. The first red line at 1990 marks the introduction of the World Wide Web. The second red line at 1998 marks the introduction of Google. However, search engines did not start with Google. Yahoo was invented in 1994, four years earlier than Google, but for the most part, this article uses 1998, when Google was invented, as a demarcation date.

In fact, post-Google, word of mouth and search are almost the same across all dimensions. It seems that the internet has made search more like word of mouth. Figure 2, below, shows how similar search patents and word of mouth patents are post-Google. In terms of generality, there is no significant difference between word of mouth and search–both are about 0.08. Recall that, pre-internet, the generality score for search patents was 0.35 and the score for word of mouth patents was .05, an 85% differential. At least in terms of generality, it’s not that word of mouth patents became more general, it’s that search patents became more specific.

Also, citations received fell post-Google. Recall that, pre-internet, there was an 81% differential between search patents and word of mouth patents. This differential fell to 5.8% post-Google. Again, it’s not that word of mouth patents became more “important,” it’s that search patents became less so.

Figure 2.  Difference between Word of Mouth and Search Post Search Engine and Google (1998-1999)
	Word of Mouth (Mean)	Search (Mean)	Percentage Difference	T-test Statistic	Significant?
Originality	0.39	0.41	4.40%	5.57	Yes
Generality	0.08	0.08	1.74%	0.64	No
Claims	16.13	17.22	6.35%	6.29	Yes
Citations Made	10.95	11.79	7.14%	4.65	Yes
Citations Received	1.51	1.60	5.83%	6.15	Yes
Self Citation Upper Bound	0.13	0.13	-6.39%	-2.91	Yes

The puzzle is why has the internet made patents more specific. One could posit that the internet should make patents more general because a more diverse field of people could access the patent through search. This article does not give any answers, but lays out this puzzle.