If you want to get a feel for the political muddles that trap most countries for decades on a sine wave of democratization and de-democratization, and why durable exits from those oscillations are so hard to come by, you might want to take a look at Bangladesh.

Bangladesh won its independence from Pakistan in 1971 after a genocidal struggle that left hundreds of thousands dead and displaced tens of millions. Since then, the country has roughly split its time between democratic and authoritarian rule. As happened in many newly independent states in the twentieth century, the champions of national independence came to power through elections and then refused to leave. Also typically, the one-party regime born of that refusal soon fell to a restive military. Seventeen years passed before another fairly-elected civilian government came to power, starting the longest spell of more or less democratic government in the country’s still-short history.

Over the ensuing two decades, the core feature of politics in Bangladesh has been acute polarization. Whenever elections approach, the rival Awami League (AL) and Bangladesh Nationalist Party (BNP) engage in bitter public showdowns that bring tens of thousands of supporters into the streets and often produce low-level violence on the margins. Unsurprisingly, the two parties carry that same animosity into government. “Once a party is in power in Bangladesh,” the Economist recently noted, “it is the unalterable tradition to declare nearly everything decreed by your opponents to be null and void.”

Meanwhile, the military has continued to play a more active role in politics than democratic theory would allow. In 2007, as elections approached and the cyclical clash between the AL and BNP cranked up, Bangladesh’s military leaders apparently saw intervention as the lesser of a few evils and tossed their civilian rulers. Two years passed under a caretaker government of the military’s choosing. Civilian supremacy returned at the end of 2008, when the AL won elections widely regarded as the fairest in Bangladesh’s history, but according to the International Crisis Group, Bangladesh’s military remains “visibly restive”:

On 19 January [2012, the military] announced it had foiled a coup by mid-level and retired officers who sought to install an Islamist government. This followed an assassination attempt on an AL member of parliament in October 2009 by mid-level officers seething over the deaths of 57 officers in a mutiny by their subordinate paramilitary border guards the previous February. Large-scale dismissals, forced retirements, deepening politicisation and a heavy-handed approach to curb dissent and root out militants have created an unstable and undisciplined force.

The systemic result of this struggle between two political rivals and the military is the familiar “truel,” or Mexican standoff, that characterizes politics in many countries stuck between stable dictatorship and durable democracy. The defining feature of this standoff is each player’s uncertainty about its rivals’ intentions; no one trusts that the others won’t make a grab for power and then shut out or destroy the others. That uncertainty, in turn, sharply increases the odds of undemocratic behavior, because even players fully committed to democracy in principle might feel pressed to cement or usurp power in order to block their distrusted rivals from doing the same to them first.

Now, in late 2013, elections are due again, and Bangladesh seems to be spiraling toward another local climax of this cyclical confrontation. As Reuters reports, the AL and BNP have called competing rallies in the capital this Friday, and at least one party leader has told followers to come “prepared with arms.” Already this year, state security forces have killed scores of protesters  in unrest spawned by the workings of a war-crimes tribunal that many BNP sympathizers see as a political bludgeon directed against them. According to my statistical forecasts, Bangladesh ranks among the 20 countries in the world most susceptible to coup attempts this year, a result that confirms many observers’ concerns that the military might respond to wider disorder as it did in 2007.

So how does a country get off of this roller coaster? Attempts to induce democratic consolidation often focus on institutional design, but Bangladesh shows how this prescription is more easily written than filled.

One of the focal points in the current confrontation is the AL government’s recent decision to dispense with an arrangement whereby a caretaker body would replace the elected government in the run-up to elections. The BNP has cast that decision as an attempt by the ruling AL to tilt the upcoming election in its own favor. Ironically, though, the caretaker arrangement has often been the focal point of mutual recriminations in past elections, as the two parties would fight over whether or not the caretakers were sufficiently unbiased.

In other words, the system that was meant to dampen that mutual distrust only seemed to end up stoking it, but when one party finally made a change, that act is seen through the same lens. The fundamental problem with expecting rule changes to induce democratic consolidation is that the process of institutional design and change is itself political, so it is subject to the same pathologies and touches off the same worries.

Outsiders can also exhort party leaders to negotiate in good faith, but parties aren’t unitary actors. Those leaders sit atop a massive pyramid of principal-agent problems, and internal rivals often respond opportunistically to attempts at compromise by stoking fears of capitulation and offering themselves as the bulwark against it. Aware of this risk, those leaders rarely take the first step.

The histories of Europe and Latin America imply that Bangladesh will eventually find a way out of these oscillations onto a new equilibrium that includes durable democracy. Unfortunately, the history of countries born in the past half-century—never mind a cursory look at the politics on the streets of Dhaka right now—suggests this election cycle probably isn’t the moment that’s going to happen.