> If your value increases linearly with education
I am not assuming that at all. The requirements to get a diploma are more than just learning in class. It involves navigating a complex bureaucracy, long term planning, and a certain level of maturity. It is entirely possible that someone could plod along for 7 semesters while having no better chance at graduating than someone who dropped out in the first year. But someone who struggles in school but still manages to push through and get a degree has demonstrated a sufficient level of growth and maturity to warrant extra consideration from future employers. Caplan thinks the degree is a meritless signal, and he thinks his "sheepskin effect" analysis proves this. And I am saying that his "sheepskin effect" requires analysis data from the college dropouts, but the fact that college dropouts exist at all implies more social value to the degree than he lets on. In fairness, he does try and address the potential "ability bias" of the sheepskin effect in his book, but I found it not particularly convincing (the study he referenced controls for standardized test scores, which seem tautological as college students have already been "sorted" to a high degree by those same scores).
> And some signaling is present
I am not arguing that signaling is not present. I am arguing against Caplan's theory that +90% of the value of the degree is signaling. If you read his book, he spends some time explaining that the ratio of the value of "human capital" vs "signaling" of a college degree is a key factor in whether or not college has social value. He claims that college is almost entirely "signaling", and therefore has negative social value. And he tries to use his spreadsheet analysis about the "sheepskin effect" to prove this. But what we are saying in this thread is that he is hiding some ideologically driven assumptions in his analysis. Someone else hypothesized that the effect could be due to an inefficient hiring market. I am arguing that the obtaining the degree in and of it self is a merit worthy accomplishment valued highly by the market. And others still could argue that maybe job candidates do not post partial educational experiences on their resume for fear of negative signaling. All of these could be partially true, or none of these could be true. It doesn't really matter. They all are plausible arguments for why the sheepskin effect exists and their isn't sufficient evidence to disprove their plausibility.
Feel free to counter anything I said with ideological arguments. But be warned that doing so merely proves the point I am trying to make. The entire point of this thread was to figure out why no one audited Caplan's spreadsheet. The most common argument, was that the spreadsheet is irrelevant if the analysis was stitched together with ideological assumptions. Your math can be correct, but someone with a different world view will interpret the meaning of those numbers differently than you will. And some of us took a look at his spreadsheet, and instead of a "smoking gun" we got exactly what we expected. A bunch of squishy formulas using squishy data trying to control for squishy variables and at the end a fairly large leap in how to interpret the results.