Will Netflix’s shareholders be sorry that they voted to let Reed Hastings carry on as both CEO and chairman? A look through the research on combining and separating out the two roles suggests that, much like most splits in life, the answer is… complicated.

Netflix’s shareholders notwithstanding, most people assume the right answer is to keep the two roles separate, in the interests of diversity of thinking and proper CEO oversight. And that’s how pretty much every research report starts — before going on to explain why it probably isn’t so.

Back in 2006, for instance, in an article in our magazine entitled “Before You Split that CEO/Chair…,” Robert Pozen, chairman (but not CEO) of a Boston-based investment management firm, cited three studies from three different countries (the U.S., the U.K., and Switzerland) which each found no statistically significant difference in terms of stock price or accounting income between companies that split the roles and those that combined them. These findings echoed dozens of previous ones going as far back as 1996.

Yet over the same period, a steady stream of business thinkers and practitioners offered up reasons (if not data) for why splitting the two roles could cause trouble. In 2003, for instance, in “In Defense of the CEO Chair,”Harvard Law’s William Allen and William Berkeley (who was chairman and CEO of the eponymous insurance holding company) argued that doing so would create two armed camps that would interfere with productivity. Two years later, Jay W. Lorsch and Andy Zelleke similarly argued in the Sloan Management Review that splitting the two roles blurs lines of responsibility, distracts both parties, and creates power struggles.

Maybe that’s why the 2012 research from Matthew Semadeni and Ryan Krause at the University of Indiana’s Kelley School was so widely reported as suggesting that the roles should not be split unless the company is doing badly. But a closer look shows the findings have more in common with the “it doesn’t make any difference” camp than the headlines would suggest.

The study looked at three scenarios, all of which involved what happens when a combined CEO/chair is split. In the first, a sitting CEO/chair gives up the CEO role but remains chairman, essentially making the incoming CEO an apprentice. In the second, the incumbent CEO/chair leaves (voluntarily or not), and the positions are filled with two separate people. In the third, a CEO/chair remains CEO but gives up the chair to another (what the researchers referred to as a demotion). In that last circumstance, if the company was in difficulties, the data indicated that splitting the two roles, so that someone could ride herd over a less-than-ideal CEO, made a positive difference. But in the first two cases, once again, the data found no difference in company fortunes. From this, Krause drew a general if-ain’t-broke-don’t-fix-it conclusion, not because splitting the roles would cause harm but because when a company is doing well, it doesn’t appear to matter which model you follow.

That same year, though, a more obscure study from GovernanceMetrics International, highlighted by the Harvard Law School Forum, approached the question from a different angle, tracking not just the effects but the costs of splitting the two roles.

The 2012 study looked at 180 North American corporations with a market capitalization of $20 billion or more. Given the complexities of running such large businesses, it was thought, differences in cost and performance of different leadership structures would be especially marked.

And differences there were. Surprisingly, combining the two roles cost more than splitting them – much more. Median total compensation (base salary, bonus, incentives, perks, stock, stock options, and retirement benefits) of executives holding both positions was $16 million. That was nearly 60% more than the median combined compensation ($10.6 million) awarded to the two individuals in companies in which the positions were split.  Considering that median income for the chair-only position was just under $500,000, it’s hard not to avoid drawing the conclusion that the order-of-magnitude $5.4 million extra the CEO receives for taking on the additional chairmanship duties is rather a lot. This impression is further strengthened by the fact that the median income for non-independent chairs was $630,930 — more than 50% higher than the $417,910 for independent chairs (making a company run by a separate CEO and independent chair a real bargain).

One might argue that differences in compensation reflect differences in corporate performance, and if that’s the case, it would be a strong argument for combining the two roles. That was so in this study – but only in the short term. Median one-year shareholder returns for companies in which the roles are combined were an impressive 11.65%, compared with a distinctly anemic 2.27% for those in which the roles were separate. Over time, however, performance for the first group lagged and the other improved such that shareholders shelling out for a combined CEO/chair received five-year returns of 31.3% while their counterparts, paying considerably less to both their CEO and chair, were enjoying an even more impressive 39.96% return.

Put all of these finding together, and perhaps the only conclusion one can draw is that the higher long-term returns for companies in which the roles are separate come from struggling companies that take steps to address the inadequacies of their CEOs (or that lower returns in companies where the roles are combined come from allowing a poor CEO too much latitude for too long).

Otherwise, most of the research suggests that Netflix’s fortunes, like most companies, will be what they will be — regardless of whichever governance system the shareholders vote in.

This blog first appeared on Harvard Business Review on 06/25/2014.

View our complete listing of Strategic HR, Leadership Development, and Compensation and Benefits blogs.