Donald Trump loves drama. His tweets more often provoke feuds than illuminate policies. To many of its observers, the Trump Presidency is also a nail-biter of a different sort: will the separation-of-powers survive him? Or is a constitutional “crisis” (whatever that means) afoot? All this hand-wringing occurs amidst a multi-front assault on the administrative state. An executive order to strangle new regulations. Legislative attempts to “REIN” agencies in. Scholarly invocations of the Star Chamber. Bureaucrats, we are told, are lazy, overpaid, and power-hungry.

Jon Michaels’ new book, Constitutional Coup, couldn’t be a timelier counterweight. It is a full-throated paean to the modern administrative state — an earnest call not only to expand, but to redeem, the bureaucracy. For Michaels, pax administrativa (his term) is far from the headless aberration critics charge. Rather, it is the very buttress to state power contemplated by the Constitution. I’m not writing a traditional book review here, so I’ll dispense with a chapter-by-chapter overview and simply commend a close read. There’s a lot to like: colorful prose, a rich historical critique of privatization, novel jurisprudential insights. Michaels’ main claim, however, is disarmingly simple: the familiar separation-of-powers between the executive, courts, and Congress has been replicated within agencies. Political appointees, career staff, and civil society act as intra-agency rivals: the first group as presidential agents; the second as reason-giving judicial analogues; and the last as a pluralistic legislative assembly of sorts. The conflicts between them ensure the checks-and-balances a constitutional democracy demands.

If this vision sounds Panglossian, that’s likely because we’re in the Trump Age. Constitutional Coup’s themes are timeless, but it is worth asking how this presidency, one year in, complicates and challenges them. And we can do so with the help of Albert Hirschman’s classic meditations on exit, voice, and loyalty within organizations perceived to be in decline. For starters, Michaels’ administrative separation-of-powers depends on civil servants that exercise what Hirschman calls voice: attempts to mobilize, dissent, or protest from within. Indeed, Michaels’ bold bureaucrats defy their bosses, selectively leak, and marshal congressional allies. A lot of that is certainly happening now. But Michael’s theory has less to say about the consequences of bureaucratic exit, when career staff decide to head for the door instead. How will agency executives be checked when there is no one to check them? Or when the few staff that remain are nothing more than loyalists and lackeys?

Indeed, there’s little doubt that a bureaucratic exodus is currently underway. The Washington Post estimates that 71,285 career staff have quit or retired during the administration’s first six months: 20,000 people more than in 2009, Obama’s first year in office. Some of the agencies hardest hit, according to the Post, have been the Departments of Treasury, Education and Labor. The Environmental Protection Agency (EPA) has lost 508 employees. The State Department has seen the departure of a “generation” of career foreign service officers. To be sure, bureaucratic turnover is not new, particularly after a presidential transition. One recent study, for example, found that 1.6% more senior career managers left during the first year after an election relative to the baseline rate. Departures were higher in agencies with ideologies that clashed with those of the incoming president. We’ll thus need more data to determine whether the current trend is historic, but one thing is clear for now: civil servants are bailing.

It is tempting to conclude that Michaels’ tripartite scheme is ill-equipped to grapple with this phenomenon. Two internal rivals are hardly as effective as three. But that conclusion would be too quick. Michaels could point out, for example, that while exit is inevitable, there will always be career staff that remain to challenge their political superiors. Indeed, the previously-mentioned study finds that lower-level staff are slightly more likely to stay after a presidential transition, perhaps in the hopes of a promotion to the now-vacant spots. So perhaps wholesale exit is not a real threat. That’s ultimately an empirical question.

Instead, a more fruitful endeavor — if we want an internal separation-of-powers to thrive — is to ask why bureaucrats are leaving and what can be done to stem the tide. In some ways, though, this will involve mere speculation. One of the defining features of exit relative to voice is that the former doesn’t require reason-giving. Absent a fiery resignation letter or high-profile statement, bureaucrats can leave in silence — and may prefer to in order to keep the option to return open. But, as Hirschman points out, costless exit reduces the ex ante probability that voice will be exercised effectively. In other words, why waste energy protesting when you can just check out? The easier it is to exit, the less likely you’ll serve as the internal check Michaels seeks.

Enter here the importance of what Hirschman calls loyalty. Loyalty includes those factors that “attach” an individual to an organization, engender an “affection” for it. What is critical about loyalty is that it both makes exit less likely and also gives more force to voice. If you care about a place, you’re less likely to abandon it and more invested in improving it. More loyal bureaucrats, that is, are more likely to exercise voice, in part because it raises the cost of exit. Loyalty also “holds exit at bay and activates voice,” in Hirschman’s words, by spurring loyalists to find new and creative ways to exercise voice. And that’s exactly what we’re seeing now. Loyal civil servants are exercising voice through petitions, rogue Twitter accounts, and public demonstrations. This is not, of course, to suggest that those who exit are disloyal. To the contrary, for loyalty to exist, the threat of exit must be real.

So while Michaels has many creative proposals for Congress and the judiciary to reinvigorate his administrative separation-of-powers, Trump also demands further thinking about how to bolster bureaucratic loyalty. For without loyalty, the incentive to resist Trump’s appointees from below is weakened — especially for those civil servants with many exit options. Think lawyers who can migrate to law firms, or scientists who can abscond to industry.

But the concept of bureaucratic loyalty first demands an answer to the question loyalty to whom or what? Different answers have different implications for voice, whether civil servants will use it and what they will say. The first possibility — particularly in executive agencies — is loyalty to the presidency, whether as an institution or an individual. For a president that prizes loyalty above all, Trump would welcome this definition. Such loyalty could be a function of policy agreement or what political scientists and public management scholars refer to as “role perception,” the sense that the job of a civil servant is to serve his elected masters. Perhaps needless to say, such loyalty would not translate into the kind of voice Michaels’ theory demands. Presidential loyalty hardly spurs principled opposition.

Another conception of bureaucratic loyalty, then, might focus on agency statutory mission. In this view, EPA career staff loyalty should be to environmental protection; Health and Human Services employees to public health; Department of Defense civil servants to national security. So when a political appointee crashes into office disavowing these ends, loyal civil servants are more likely to rise up in defiance. But this simplified take immediately gives way to a hornet’s nest of questions, resolutions of which require much more space than available here. Among them: Given that many statutes are ambiguous, who within an agency should have the authoritative interpretation? The appointed agency head, many will say, but what about bad-faith or off-the-wall interpretations? Regardless, for Michaels’ account to work, bureaucratic loyalty must look closer to the second definition, than the first: loyalty to some value orthogonal to that of the elected presidency, whether that be of a statutory mission, a congressional committee, or the “rule-of-law” more broadly. Only then will civil servants be more likely to serve as intra-agency ballast, rather than yes-men (and women).

An important project going forward, then, is to consider how such bureaucratic loyalty could be augmented. Doing so would help to vindicate Michaels’ conception of internal checks-and-balances. Immediate ideas range from the glib to the more high-minded: happy hours, team-building retreats, systematic training about the history of the agency — basically anything that encourages staff association, legal knowledge, and institutional respect. Even if a Congress, like this one, would be hesitant to fund such activities, perhaps they could be taken up or sponsored by other entities with existing resources such as the Government Accountability Office, Council of the Inspectors General, or the Administrative Conference of the United States. Much in the Trump Presidency remains uncertain. We don’t yet know how this episode ends. Fostering a greater sense of bureaucratic loyalty will help to ensure that when the going gets tough, the tough don’t get going.