Discussion
Put the findings together. Schools in areas their principals call high-crime work with more outside organizations, and the added ones are community groups and government agencies in about the proportion the school already had. Five of the seven organizations rise clearly, and juvenile justice agencies rise by too little to separate from zero. The police do not rise, and taking the police out of the count strengthens the association rather than weakening it.
That fits what Patrick Sharkey has shown about American cities: the response to violence over the last thirty years came substantially from community organizations rather than from policing alone. Our result runs the other way round from his, and it is not causal. A school in a violent area holds more ties to the same kind of organization he identifies, and it works with the police less often.
The two explanations come out differently for the people and for the institution. Where principals report more crime, parents turn up to school events considerably less often. That is Wilson's social isolation, and social capital predicts it. The school itself works with more organizations, and resource dependence predicts that. The people and the institution among them run in opposite directions, in the same places at the same time.
The advantage test also runs against the social capital account. If a wide network were a mark of doing well, the schools with the most advantaged students would have the widest networks. They do not. Entered together, the share of students expected to go to college goes with fewer organizations, at -0.057, and the share testing in the bottom fifteen per cent goes with more, at +0.039. Both separate from zero, and both run the wrong way for that account. Average daily attendance, at +0.027, does not separate from zero. With all five student-body measures in the model the first two shrink to -0.038 and +0.019, and neither separates from zero.
We know that the police stay behind. We do not know why
The police result holds up. Among schools with the same number of partners it is negative in all six surveys, the pandemic year included, and at every grade level, and it survives leaving any one survey out. The reason behind it is missing, and this page says so plainly.
Four explanations are ruled out, each by its own test on this page, and a fifth is weakened.
It is not the police being so common that there is nowhere to go. Cut schools by how many organizations they work with and the result is strongest in the group where the police are least common, at -0.040 where 74% work with them. In the group where 98% already work with the police it shrinks to -0.008 and does not separate from zero, in a group where almost every school already holds the tie.
It is not a recent argument about police in schools. The swap is already in the two oldest surveys, at +0.062 in 2005-06 and +0.052 in 2007-08.
It is not trouble inside the building. Put the trouble staff report into the model and it washes out. The principal's reading of the outside keeps predicting.
It is not policing arriving under a different label. Take the police out of the count and the association gets stronger, +0.124 against +0.104.
The fifth, that the schools we compare were picked by the crime reading, is weakened and not ruled out. Whether a school ends up in the either-or comparison at all goes slightly with the crime its principal reports, at -0.016 and -0.013, and both separate from zero. So the comparison is selected on the crime reading to that small extent.
That still leaves four explanations standing, and this survey cannot tell them apart. The police may cost a school control, so a school under pressure buys help that costs it less. The police may be claiming a job the school treats as its own, which is the idea set out on the theory page. The choice may not be the school's at all, since these agreements are often signed by a district or a city. Or the counselors and social workers may simply have become easier to get in poorer places, which would put more of them in rougher-area schools whatever any principal wanted.
Resource dependence, the theory this study tests, gives a reason for the first of these. It runs in five steps.
First, a school under pressure looks outside for help. That is why schools in rougher areas work with more organizations.
Second, every partner that helps also gets some say in what the school does. The police get the most say: an officer can arrest a student whatever the school thinks.
Third, a school with more partners has more places to get help, so it does not have to rely on any one of them, the police included.
Fourth, so among schools with the same number of partners, the ones in rougher areas less often count the police as one of them.
Fifth, this is about a standing partnership, not about calling the police when something happens. Schools in rougher areas pass about the same share of the incidents they record to the police as other schools do (+0.003), and slightly more when they record the same number of incidents (+0.012). What they do less often is give the police a part in running the school's safety efforts. That way the school gets outside help and keeps control of its own safety work.
Philip Selznick, Richard Emerson and Karen Cook, set out on the theory page, each describe one of these steps. The survey shows a pattern that fits them. It cannot show that principals reasoned this way.
Three pieces of information would sort that out, and the survey holds none of them: who signed the agreement, what the agreement says, and whether the school deals with its own discipline cases in-house. So the pattern is solid and the reason for it is still open.