Since reading educational research changed from being an amateur interest of mine to being a daily professional commitment, nothing has been more consistently striking than the strength with which educational outcomes are correlated with demographics of socioeconomic status, parental education background, and race-- and the relative weakness of pedagogical and educational variables.
Broadly speaking, the demographics of students within a given school are far more determinative of educational outcomes than any pedagogical variable that can be isolated. Given this, people who complain that poor students can't get a good education because they are systematically excluded from the good schools have the line of causation backwards; rather, schools are regarded as good because they have systematically excluded the poor students who contribute to poor metrics. Typical complaints that poor students underperform because they are ineligible to attend the best schools ignore the fact that the best schools would necessarily perform worse with an influx of poorly performing students. The response then should not be to expect education to elevate people out of their current station but rather to elevate them out of their current station in order to improve the outcome of their education.
The racial achievement gap is vexing and necessarily controversial. My own stance is that there is very likely no single cause, but rather a series of causes working together in a vicious cycle. I'm confident that our systems to measure educational outcomes are culturally biased in a way that disadvantages black and Hispanic students; that systematic economic exclusions and inequities undermine parental and community stability which contribute to educational outcomes; and that educational disadvantage is hereditary, in the sense that parents who were disadvantaged are likely to pass that disadvantage down to children, due to an inability to model and engender intellectual skills students need. (Indeed, this hereditary disadvantage is widely acknowledged in language and literacy, which are disproportionately influenced by exposure and modeling during the formative years that precede schooling.)
Given the difficulty in isolating variables that contribute to racial disadvantage, it makes the most sense to address the most obvious and manipulable correlate: parental income. We can likely affect major positive changes in educational outcomes through a brute-force method of giving parents money. Typically, addressing educational disadvantage through spending is focused on schools. But since students spend a large majority of their lives outside of schools, and because of the previously-discussed power of demographic and family factors, this is perhaps a misapplication of funds. Rather, giving parents money is likely to have a more direct and immediate effect. Some people say that you can't just give people money. But as Matthew Yglesias has repeatedly discussed, just giving people money has a strong track record. In addition to preventing an unduly paternalistic system from constraining poor parents from making decisions with the same freedom of the rich parents whose children perform so well, such a system would avoid having to create an onerous and costly bureaucracy that would be necessary for enforcement of particular rules about use. I would pay for such a program through significantly more progressive income taxation, which would likely result in slightly higher tax rates for the middle class and significantly higher tax rates for the upper class. Perhaps more traditional social programs that attempt similar interventions could be shrunk or phased out altogether due to the positive effects of a simple transfer of money. In time, if such a program was successful, it would likely reduce costs of other social programs. I am dedicated to protecting our current system of material support for the elderly, but to enact it to the exclusion of similar redistributive efforts for children and parents seems quite strange.
This would all take time. Poverty, in terms of both real capital and social capital, is a cyclical and effectively hereditary problem. In a country in which it is not uncommon to find families who are in their third or fourth generation of poverty, we cannot expect any programs to make miraculous changes overnight. Standards that are universal are inherently an error, as there will always be those who perform below standard, no matter what the social system. And I have been and remain dedicated to a long-term social overhaul that removes material security from chance and demographics in general. A universal basic income, while far from a panacea, would accomplish a great deal of what I'm advocating for here. But if a universal minimum income is not in the cards politically in this generation, perhaps an explicitly child-and-education focused effort like this one could be justified. If we take education as seriously as we say, and if we are really dedicated to attending to the needs of all children, a program of redistribution is likely a necessity-- particularly in the context of a school reform movement which has proven to be, empirically, one of the greatest failures of modern policy making.
When the correlate is determinative, change the correlate. If you want to make sure that no child gets left behind in a capitalist system, give the parents capital. And then, in the long run, we can address why their parents were ever poor in the first place. Let's turn vicious cycles into virtuous.
Update: Two (preemptive) reasons why this doesn't represent a nihilistic argument for the irrelevance of education. (That would be a particularly odd stance for me to take, as I am an educator, and a researcher whose major research interest is how to make education better.)
First, arguing that educational quality (which includes but is certainly not limited to teacher or school quality) is largely determined by demographic and familial inputs is not the same as arguing that there is no such thing as educational quality. I have little doubt that differences in educational quality are significant, in both the practical sense and the statistical sense (that is, I believe that differences in educational metrics exist that we can say with great confidence are not the result of random quantitative chance). But addressing education as a broad social issue means tackling the largely determinative part of educational outcomes first. Indeed, practically assessing educational quality would be made vastly easier if we had the ability to remove the powerful demographic confounds.
Second, it is plausible (although for the time being largely unverified empirically, to my knowledge) that educational quality has an unevenly distributed impact depending on a student's overall educational level. In other words, educational quality could be more determinative of student outcomes for students who have reached a certain threshold, but be largely irrelevant to the success of students who have not. Think about an illiterate 10th grader. Such a student is overwhelmingly likely to have come from a low socioeconomic background. The ability of high school educators to improve that student's learning outcomes is likely severely restricted. A high school is not designed to teach basic reading and writing skills, likely lacks the resources to do so, and likely lacks educators who are specifically trained or experienced in that area. So for such a student, the lack of prerequisite ability could render educational quality irrelevant. But educational quality could still exist for mediocre and high-achieving students. Again, the natural response seems to be to address the demographic and familial factors that are strongly associated with a lack of basic skills.