While I’m homeschooling my elder sons, I’ve long wondered about how well it really works.  Sure, the critics are usually angry apriorists.  But advocates seem quick to repeat optimistic statistics without proper scrutiny.  What’s really going on?

The best piece I’ve found so far is Kunzman and Gaither’s “Homeschooling: A Comprehensive Survey of the Research” (2013, Other Education: The Journal of Educational Alternatives).  Highlights:

V. Academic Achievement

The subject of homeschooler academic achievement has received much scholarly attention, but unfortunately most of this work contains serious design flaws that limit its generalizability and reliability. From 1990 to 2010 five large scale studies of academic achievement have been conducted under the sponsorship of HSLDA [Home Schooling Legal Defense Association] (Ray, 1990; Ray, 1994; Ray, 1997a; Ray, 1997b; Ray, 2010). These studies all rely for their data on samples of homeschoolers recruited for the purpose. Volunteers are asked to submit demographic data as well as the results of one or more group of standardized test scores, with promises made that the research will be used for homeschooling advocacy. These self-reported scores (from tests that are typically proctored by the parent in the home) are then compared against national averages and the results reported. In every case homeschooled students have consistently scored in the 80th percentile or above on nearly every measure.

The original studies are always clear that the data being presented do not reflect a random sampling of all homeschoolers, and that they do not control for key variables like race, SES, marital status, or parent educational attainment when comparing against national averages. Such caveats are critical, for the homeschooler sample obtained by this recruitment strategy is not representative of national norms, nor, indeed, of all homeschoolers. For example, in the most recent HSLDA-sponsored study, published in 2010, the sample of 11,739 homeschooled children came from families that were 95% Christian, 91.7% white, 97.7% married, 80% with stay-at-home moms, and 45.9% with incomes over $80,000 per year (Ray, 2010). Though such limitations are noted in the original studies, the less technical versions produced for popular consumption and the press releases put out by HSLDA habitually ignore such caveats and cite these studies as proof that homeschoolers outperform public schoolers by wide margins on standardized tests (Gaither, 2008b; Kunzman, 2009a).

Higher-quality studies find rather different results. 

First, homeschooling has opposite effects on verbal versus math scores:

Given this persistent corroboration across two decades we might conclude, tentatively, that there may be at least a modest homeschooling effect on academic achievement–namely that it tends to improve students’ verbal and weaken their math capacities. Why? Answers here are only speculative, but it could be that the conversational learning style common to homeschooling and the widely-observed phenomenon that homeschoolers often spend significant time reading and being read to contribute to their impressive verbal scores, while math is not given the same priority (Frost & Morris, 1988; Kunzman, 2009a; Thomas & Pattison, 2008).

Second, as usual, what parents do is overrated:

[H]omeschooling does not have much of an effect at all on student achievement once family background variables are controlled for. This conclusion is implicit even in many of the HSLDA-funded studies, which consistently find no relationship between academic achievement and the number of years a child has been homeschooled (Ray & Wartes, 1991; Ray, 2010).

Third, as usual, who parents are is underrated:

[P]arental background matters very much in homeschooler achievement. Belfield (2005) found greater variance in SAT scores by family background among homeschoolers than among institutionally-schooled students. Boulter’s (1999) longitudinal sample of 110 students whose parents averaged only 13 years of education found a consistent pattern of gradual decline in achievement scores the longer a child remained homeschooled, a result she attributed to the relatively low levels of parent education in her sample. Medlin’s (1994) study of 36 homeschoolers found a significant relationship between mother’s educational level and child’s achievement score. Kunzman’s (2009a) qualitative study of several Christian homeschooling families found dramatic differences in instructional quality correlated with parent educational background.

Kunzman and Gaither on college admission, the issue that most concerns me:

IX. Transition to College/Adulthood

[…]

[M]ost of this literature is quantitative, consisting for the most part of surveys of admissions officers. The consistent finding of such studies is that homeschooled applicants are accepted at roughly the same rates as their conventionally schooled peers, that admissions staff generally expect homeschoolers to do as well as or better than their conventionally schooled peers in college, and that while colleges and universities welcome homeschooled applicants, most do not go out of their way to provide special services or admissions procedures for homeschoolers (Duggan, 2010; Haan & Cruickshank, 2006; Jones & Gloeckner, 2004b; Sorey & Duggan, 2008). One qualitative look at attitudes of admissions officers at three institutions, however, found that many officers privately believe that homeschoolers are close-minded religious bigots, suggesting that what such individuals report on surveys might not always tell the whole story (Millman & Millman, 2008).

Unlike most of what I’ve read about homeschooling, Kunzman and Gaither isn’t inspiring.  But it inspires confidence in its accuracy.