THE growth in the reach and discretionary authority of central government, not only in the United States but everywhere, raises fundamental problems for the control and accountability of agencies charged with serving public purposes.  The “bureaucracy problem” is frequently discussed in terms of “red tape,” inefficiency, unresponsiveness, and resistance to change. Less frequently discussed, but possibly of greater importance, is the extent to which even public agencies that are efficient, responsive, and open to change can produce the desired social effects under an increasingly common condition-namely, when the end in view requires that one government agency modify the behavior of another.