We are heading into divided party government in Washington after an unproductive unified Republican period. Will a Democratic House bring even less productivity and more government shutdowns? We use the history of Congress and US state governments for an updated look at what party divisions between the legislative and executive branches bring us in terms of policy output. Patricia Kirkland finds that divided government at the state level increases the chance of budget delays that can lead to shutdowns. But she says some states are perennially late and others that face disastrous outcomes for impasse do get their work done. Benjamin Schneer finds that divided government historically does reduce the number of major landmark laws passed by Congress, but only by a few each term. There are also huge differences across time and presidencies, regardless of partisanship. But both confirm the conventional wisdom that divided government will not help improve our governance.

Studies: “Is Divided Government a Cause of Legislative Delay?” and “Divided Government and Significant Legislation: A History of Congress from 1789 to 2010.”

Interviews: Patricia Kirkland of Princeton University; Benjamin Schneer, Harvard University


Grossmann: This week on Political Research Digest, do divided governments limit lawmaking and bring budget impasse? For the Niskanen Center, I’m Matt Grossmann. We are heading into divided government in Washington after an unproductive unified session. Will a Democratic House bring even less productivity and more government shutdowns? Today we use the history of Congress and US state governments for an updated look at what party divisions between the legislative and executive branches bring us in terms of policy output.

I talk to Patricia Kirkland of Princeton University about her new Quarterly Journal of Political Science article with Justin Phillips, “Is Divided Government a Cause of Legislative Delay?” They find that divided government at the state level increases the chance of budget delays that can lead to shutdowns, but she says some states are perennially late and others face more disastrous outcomes and get their work done.

I also talk to Benjamin Schneer of Harvard University about his new Social Science History article with Stephen Ansolabehere and Maxwell Palmer, “Divided Government and Significant Legislation: A History of Congress from 1789 to 2010.” They find that divided government does reduce the number of major landmark laws passed by Congress but only by a few. He says they also uncovered big differences across time and presidencies.

When one party controls the presidency or governor’s office and does not fully control the legislature, we call it divided government, and the conventional wisdom with some confirmation by both Kirkland and Schneer is that it won’t be good for governance.

Kirkland: The congressional wisdom is certainly that divided government can create legislative delays or slow down the lawmaking process. The idea obviously is that when different parties control different branches, it’s harder for lawmakers to agree on new legislation. That said, some of the work that scholars in political science have done over the years has challenged this notion and looked at maybe some other factors that might be more important in legislative performance.

Grossmann: While Kirkland looked at the states, Schneer assessed the conventional wisdom over the full span of history.

Schneer: The conventional wisdom is definitely that divided government would lead to less legislative productivity. That certainly seems to be what you would see in the media, and a lot of political scientists actually I think agree with that as well. I think that in some sense our paper is confirming that conventional wisdom but just using as the context a much broader span of history since we’re going back to the first Congress.

A lot of the work on this question has been post-World War II, but at the same time what I would say is that while we do find this effect on the margin, in some sense the message of our paper is that the other side is in some sense right too in that it doesn’t seem like divided control of government is a really particularly important predictor in explaining variation that we see in legislative productivity. There are a lot of other more important things. I would say that while we do find support for this finding that divided government matters, but part of the paper is pretty sympathetic to the other side of the argument.

Grossmann: Kirkland’s main finding is that divided government leads to budget delays, but that can be mitigated with consequences.

Kirkland: My co-author, Justin Phillips, and I set out to try and gain a better understanding of the link between divided government and legislative performance. To do this we focused on state budgets. In the US, states need to pass a new budget every year or every other year depending on the state. When a state experiences divided government as opposed to when the same party controls the legislature and the governor’s office, we find that lawmakers are more likely to miss the deadline to pass a state budget before the start of the new fiscal year.

One of the things that’s important about this is if lawmakers fail to pass the budget in time, there could be negative consequences. Some of them aren’t too terrible, maybe some small limitations in services or some short furloughs, but in some cases they can be sort of major furloughs of state workers, significant limitations in public services. And in some states if lawmakers don’t pass a budget before the new fiscal year starts, state government actually shuts down. We find that when government is divided, it’s more likely that there’s going to be a delay and a late budget. These findings are consistent with the notion that divided government slows down legislation and maybe even contributes to gridlock.

However, I think one of the most interesting things about this paper is that we also find that when there are rules in place or institutions that make missing the deadline very costly for lawmakers, divided government doesn’t seem to have much of an effect on the timing of budget adoption at all. We think that this is because if, for example, a state government will shut down if the budget’s late, we know that voters really don’t like that, and so we think that lawmakers are probably concerned that voters might punish them at the polls in the next election. In states that have government shutdown rules, a late budget is actually no more likely to occur under divided government or when there’s unified government.

Grossmann: Schneer similarly found that divided government matters but is by no means the most important factor.

Schneer: We gathered data on significant legislation that was passed from the first Congress onwards, so 1789 onwards, and we were able to answer those questions. First, what we found was that divided government was associated with about three fewer acts per Congress as compared to unified government. It didn’t, interestingly enough, seem to have a big effect on just total legislation overall. If you break out the results by time period, it looked like before 1900 divided government seemed to lead to about one fewer significant act per Congress. And then post-1900 it was about four acts per Congress fewer, so the effects seem to be larger more recently. That’s the effect that we find on the margin.

The other thing that we look at though is how good of a job does party control of government do in explaining variation in legislative productivity overall. The answer there is it doesn’t seem to do a great job. There’s other things that seem to be much more important. I mean, there are just big time trends overall that aren’t explained by party control of government. Who the president is seems to explain more of the variation. So that while we do think that there is this change on the margin going from divided to unified government or unified to divided government, we don’t think that divided government is necessarily a great explanation to explain the big changes in legislative productivity that we see over time.

Grossmann: The question has a long history in political science stimulated by David Mayhew’s famous claim that divided government does not impede important lawmaking.

Kirkland: Even if we go back before Mayhew, and there’s this famous quote by V.O. Key where he said that common partisan control of the executive and legislature does not assure energetic government, but division of party control precludes it. Even before Mayhew there was this notion that separation of powers made it really difficult to make new laws or change policy and that parties could act as the bridge between branches to get things done.

Kirkland:                One of the things David Mayhew, of course, in Divided We Govern has the first sort of quantitative and systematic study of divided government, and he points out that a lot of the earlier scholars are writing at a time where the norm is unified government. He undertakes this systematic analysis, and he concludes that divided government doesn’t seem to matter very much at all, at least for the passage of landmark legislation. Then of course this seminal work spawns a slew of studies that look at the link between divided government and legislative productivity in particular.

I guess if I were to summarize that work, I would say that what comes out of it is some consensus, although not unanimous, that divided government almost certainly shapes legislative performance, but there are other factors. It might be polarization. It might be the importance of the legislation involved that actually might condition its effects.

Schneer: David Mayhew wrote this famous book, Divided We Govern, where somewhat surprisingly he finds that divided government doesn’t necessarily seem to reduce legislative productivity, and that spawned a ton of research since then with people responding to that claim and trying to approach it from a different angle. Yeah, so one interesting aspect or one notable aspect of that work is that the time period is 1946 onwards, I think to about 1991 or 1990.

Kernel of our idea was to try and do what David Mayhew had done in terms of assembling a dataset of significant legislation but just extend it backwards. Yeah, I think that there is a number of reasons why we’re finding something slightly different from him. First of all, we try to approach the data gathering using the same principles that he used, but we show in the paper even when we look and compare our dataset versus his for the same years, it’s not always exactly the same. Then more importantly, we’re going back further.

Grossmann: Kirkland’s work was stimulated by observing the federal level, where the consequences of impasse clearly mattered.

Kirkland: We landed on this interesting regularity, which was in certain instances Congress could pass something. For example, they couldn’t really pass the regular appropriations bills that are typically associated with the federal budget, but faced with a shutdown, they could get out of it. Or also, when they needed to raise the debt ceiling because we were facing tremendously uncertain consequences of a failure to raise the debt ceiling, and there was a general agreement that we didn’t know what would happen but it would be really, really, really bad. And they could always get together at the last minute and piece an agreement together.

Grossmann: Schneer’s work was stimulated by a class assignment on congressional history.

Schneer:  Origin for this paper was actually that Max and I were teaching assistants for Steve in a course on Congress that he was teaching. One of the main projects for the course was that each student or sometimes a couple students were assigned to become experts on the history of Congress for a particular decade. As part of that process, one of the things they did was begin to gather data on significant legislation passed in their decade.

That actually was the early beginnings that, first, it might be possible to gather this data. That also served as a template that we continued to build on when we pursued this paper more. It ended up being the case that a lot more work had to go into filling out the database of significant legislation over time, but yeah, the origins of this paper were really in this course.

Grossmann: Schneer looked at the traditional outcome, significant legislation, but counting major laws over time is difficult.

Schneer: We really were trying to follow the spirit of what Mayhew had done in his initial work but applying it to a longer stretch of time. I think you can think of significant legislation in terms of its contemporaneous significance, do people at the time think that it’s significant, and also in terms of retrospective significance. David Mayhew in his book is doing that. He’s looking at New York Times end-of-year legislative summaries to get at the contemporary significance and then looking at secondary literature to assess the retrospective significance.

Yeah. Without going into too many details, we didn’t use those exact same sources but tried to basically follow a similar approach. We used the CQ Almanac. We just used a variety of secondary sources and also sources trying to get at the contemporary accounts of what was significant. Ultimately, I would stress that in Mayhew’s approach and in our approach it is in some sense arbitrary ultimately what the researcher decides.

Grossmann: Kirkland found budget delays as a clear measure of success and failure in the states.

Kirkland: Actually, one of the sort of interesting things about this measure is there’s variation in how bad it is to have a late budget. Interestingly enough, on average there’s about 19% of state budgets over the time period that we considered, which was 1968 to 2010, about 19% of state budgets are late, but there’s a lot of variation across states in how frequently the budget’s late. Some states like Wisconsin and New York, for example, very often have late budgets. Other states, it rarely happens. Again, this variation in how bad it is was also important to how we wanted to think about what might condition the effects of divided government.

I think one of the things that was really important to us, because we decided to take a narrow approach with this project because we wanted to try and isolate what we could plausibly claim to be a causal effect of divided government, and so this focus on budget delays was also helpful methodologically. Because one of the biggest challenges with some of the existing empirical studies of divided government is how you measure legislative performance or the production of legislation. Because one option is to count the number of laws that are passed, but there’s a component of that missing. Because if the number is low, it might be because there’s gridlock, or it might also be that there’s just not a lot of demand for legislation.

There are lots of sophisticated ways in which scholars have tried to incorporate both supply and demand, but this is a really thorny issue, and a lot of the differences across studies of divided government actually surround this measurement question. One thing that is really helpful about this budget delay measure is that the supply and the demand are clear. There’s a demand for one state budget a year or one state budget every other year, and it’s clear when it’s passed and that it’s passed.

Grossmann: She was able to get a causal estimate of the effect of divided government by comparing states that were just on the edge of unified versus divided government.

Kirkland: We don’t want to compare just cases of unified and divided government because we might be worried that there are factors that cause divided government, meaning there are factors that determine whether or not a state’s voters elect a unified or divided government. Some of those factors could also influence budget timing, like the timing of budget adoption or how easy it is for lawmakers to adopt a budget or how difficult and prolonged the negotiations might be, however we want to think about that piece of it. Basically, this means we’re worried about some factors that cause states to select into divided government that are related to whether or not the budget is passed on time, so it’s an endogeneity concern.

With the regression discontinuity design, we focus on state years that are about as likely to experience divided or unified government, meaning the chances of experiencing divided government were about 50-50. Among these states we have a sample of state years where all those potential confounders that we might worry about are in theory similar for both states that have unified and divided government, but what’s different about them is whether or not they actually experience divided government.

Grossmann: That was not easy because they needed a regression discontinuity measure that worked for separate elections for the state houses, state senates and gubernatorial elections.

Kirkland: We have a lot of elections to different institutions at different times, so the way that we handle this is we use a series of simulations. What we want to determine from the simulations is the smallest partisan vote shock that would’ve produced a different outcome in terms of divided or unified government. Just to clarify, I mean how much of the vote would have to shift from one party to another for the opposite outcome to have occurred.

If we get a state that has an election and let’s say it’s an on-term election, so we have a governor, the state senate and the state general assembly are all having elections. All the elections take place. Party control is in the hands of Republicans. The question then is what share of the vote at the state level would’ve had to shift from the Republican Party to the Democratic Party to get divided government instead of unified government? The way that we figure this out is with simulations, and we simulate every legislative and gubernatorial election in each state.

Then we look at in each simulation, we shift the vote share from one party to another, and we say, what was the outcome? Then we compare that to the actual outcome, and we determine two things. One, we learn something about the probability that a state will experience divided government, but we also learn what we talk about as the electoral or vote share distance to divided government. Again, that’s the smallest vote share that would have to shift from one party to the other to get divided government instead of unified or unified instead of divided government.

Grossmann: Kirkland and Phillips also found that polarization and the consequences of budget impasses matter for delays alongside divided government.

Kirkland: The idea that the effects of divided government vary with the political context. In particular in our instance, we see that the effect of divided government is mitigated by certain institutions that make it really costly for lawmakers if there’s a delay. We also find some evidence that higher levels of polarization actually make it more likely for divided government to create legislative delay, which is I think very consistent with much of the more recent work on divided government at the federal level.

Grossmann: There are big differences across states in how normal it is to have a delay.

Kirkland: I don’t know if I would necessarily say that something else has to go wrong. The reason I say that is because in some states late budgets are the norm regardless of whether government is divided, so it seems like some of these other factors like the level of polarization and institutions in particular really matter. They really do have an effect on or really do condition the effect of divided government. I think that’s a key takeaway of our work here.

Grossmann: Kirkland says divided government also may increase deficits or affect lawmaking.

Kirkland: We had perfect data. I think we would like to have that perfect measure of legislative performance to look at the relationship between divided government and legislative productivity, but that said, there are lots of ideas of effects of divided government. For example, we have a working paper that develops and refines this method a little bit more, and we have an application to budget deficits because there is some existing scholarship that predicts that there will be larger budget deficits under divided government.

Grossmann: Having only one government to look at, Schneer looked at its productivity over centuries.

Schneer: It’s also the case that looking over time across congresses you could make the case that just the way Congress operates has changed, and so what you observe in one Congress is not directly comparable to another. An example of this is that in the early congresses, a lot of things were done on an ad hoc basis. There was no formal budget process. If you were just to count up legislation, there are lot more individual units. Whereas when the process is rationalized, everything gets put together under one heading. That’s a problem. I don’t know that we have a way of totally confronting that head on other than just to try to be consistent over time in terms of making the same sorts of judgments and applying them in one time period to the next.

Grossmann: But they found major time period differences irrespective of divided government.

Schneer: We observe probably the highest level of productivity, at least according to our measure, in the period from around 1927 to about 1966, so the New Deal through the Great Society years. Then at the same time, that’s the raw number of legislation. If you look at significant legislation as a percentage of total legislation, the first two congresses are huge outliers, which is not really surprising.

Grossmann: And it’s all been downhill since the 1960s in terms of congressional productivity.

Schneer: This coincides with the rise of the conservative coalition between some Southern Democrats and Republicans and generally speaking, the ascendancy of the Democratic Party nationally. As that starts to unravel after the Civil Rights Act, after the Voting Rights Act, we also begin to see legislative productivity decline, and it’s continued to go down ever since according to our measures. Speaking more generally, it’s definitely the case that both in terms of total legislation and significant legislation, the context and the time period matter a lot.

Grossmann: Schneer, Ansolabehere and Palmer look at both immediate changes from unified to divided government and longer differences over time. Most of the effects were immediate.

Schneer: If the question is does the effect seem to be immediate, it does seem to be immediate. I mean, so there were some other things that are related to this that we looked at. One theory is that if there’s been, for example, a long period of divided government, maybe the first Congress where there’s unified government you’d expect to see a big uptick. We didn’t actually really see big evidence in favor of that, though intuitively you could understand why that might be the case.

Our results generally seem to show that it’s not like there’s a process where it’s lagged. It can be a bit memory-less. The context can change and swing back and forth from one Congress to the next, but in every Congress often there are new members. In any case, in order to be reelected, members feel like they need to produce, or if you’re in the minority, maybe stop the other side from producing. So it’s not a shock to me that you wouldn’t see big lags in this process.

Grossmann: They assess some alternative explanations, finding it was not due to supermajorities.

Schneer: We look, for example, at the effect of having supermajorities, the effect of centralization. Unfortunately, we don’t really look at polarization per se, although especially now that’s something we probably should’ve done. The reason we’re concerned with those explanations is, well, take the example of supermajorities. You might think that observing this change in legislative productivity going from divided to unified government, maybe you’re not really observing a difference between divided and unified government; it’s really a change of when there’s a supermajority versus all the other times. By controlling for when there’s a supermajority, you can assess if the process that we think is working, that is that you can basically assess is divided government a necessary condition for the decrease in legislative productivity.

Just to be clear, the reason that we’re interested in supermajorities at all is, for example, in the Senate you might be worried about the ability of the minority party to filibuster. That could block legislation and therefore even under unified government, perhaps you wouldn’t observe high levels of legislative productivity. When we try to control for whether or not a party has a supermajority, we do find that that makes a difference, but it’s in some sense over and above the effects that we observe for unified versus divided government.

When we look at that, the conclusion that we make is that this does seem to matter a bit, but it’s still the case that switching from unified to divided or divided to unified government on its own is a sufficient condition to observe these changes in legislative productivity.

Grossmann: Schneer says the last Congress continued the downward trend.

Schneer: Yeah. I do think that by most measures the current Congress, which is under unified Republican partisan control, would be considered not particularly productive. I do think that this does seem to be part of a general downward trend irrespective of whether or not it’s divided or unified government. I couldn’t really say if it is accelerating or not, and I don’t necessarily think it’s specific to Republicans. This isn’t something we really take up too much in the paper, but I think when you look at polarization at the elite levels within Congress, within national politics as a whole, that has to be a part of it.

Grossmann: As in history, he says presidents do matter, and Trump is not helping.

Schneer: We do find that if you include president fix effects, that does a better job in some sense than just looking at unified versus divided control in terms of explaining overall variation. I think that it’s hard to disentangle how much of that is a period effects versus individual legislative effectiveness of a president. My guess would be it’s some of both.

If we’re willing to take that and run with that, I think yes, and just anecdotally, I think from observing this term of Congress, it does seem clear that the president has a huge role to play in setting the agenda and defining the scope of the debates that are going on. Right now as we’re speaking, a unified Republican Congress may be heading towards a legislative shutdown. That’s mostly I think due to who the president is.

Grossmann: And so we shouldn’t expect things to get better in the next Congress.

Schneer: Barring some sort of unforeseen reversal of fortune, we can probably expect the next Congress to be less productive even than this one. In some sense I think the lesson of our paper is that there may be multiple reasons why we would expect that. First, just because on average moving from unified to divided government we would expect on the margin a decrease in the number of significant acts of Congress passed. But then second, as we show, if you’re able to look at changes in legislative productivity over the broad span of history, right now we’re in the midst of a pretty stark decline. I guess moving to the next Congress, we would expect a less productive Congress on the margin and as part of this continued decline.

Grossmann: The number of states with unified government is actually increasing this year, but Kirkland says that may not save us from budget impasse.

Kirkland: One of the things that’s been notable about state governments, I don’t know, for quite a while now is that as the federal government’s really struggled with gridlock and there have been these periods of divided government, that it’s been less common in the states and they’ve been producing more policy. I think that that will continue with shifts in party and shift from divided to unified government at the state level.

Specifically though on this issue surrounding budgets, I’ll be curious to see what happens. Because I think if you had asked me this question before I did the analyses for this paper, I would’ve predicted that everything would go much more smoothly with divided government. But now that I’ve done them and looked at the data a little bit more carefully, I tend to think that in some states they’re still going to struggle to pass their budgets on time even if they have unified government.

Grossmann: Can policymakers escape their historical circumstances? Schneer thinks that would be an uphill battle.

Schneer: I’m not sure that there’s an optimistic message here for politicians or advocates for particular causes that want to see important legislation coming out of Congress. You might say that one message of the paper in fact is that winning unified control of Congress may result in some policy wins on the margin, which could be really important, but that you can’t necessarily escape the secular trend, in this case downward secular trend, that you’re in based on the time period that you’re in until something pretty rare happens.

Grossmann: But Kirkland has a policy proposal, though it may be unpalatable. She says the worse the outcome for inaction might actually be better for stimulating budgeting and lawmaking.

Kirkland: I think this is actually an interesting policy implication. One thing I keep coming back to is my home state is Pennsylvania, and this is a state that for a long time had a shutdown rule. They took the teeth out of that bill some years ago. In the recent past had an incredibly long budget delay that just went on and on for many, many months. In fact, I believe they were at the point where they needed to pass the next fiscal year’s budget and still didn’t have a budget. I tend to think that these rules really matter.

If it were the case that policymakers wanted to mitigate the potential for divided government to create problems in the future, they could create these critical deadlines, whether it’s a government shutdown for budgeting or whether it’s sunset provisions on laws or bills that need to be or programs that need to be reauthorized or funded at some point in time in the future or they go away. That might be a way of forcing them to get together and come to some sort of agreement to move forward in the future.

Grossmann: Schneer agrees that looking at the states enables a good look at divided government, but he says we also want to understand the institution of Congress itself.

Schneer: Looking across 50 states and looking at variation across states in partisan control probably allows you to make cleaner and maybe even more well-supported claims about, in this case, divided government actually causing budget delays. I think also in terms of understanding legislatures in general, I think being able to do that type of work and make those sorts of inferences is very helpful.

I would say at the same time we also want to understand the US Congress in particular and how things have changed in that institution over time, and so I would say that that approach and this approach are pretty complementary in that they’re telling us slightly different things but about the same overall question or same overall problem.

Grossmann: Kirkland says many of their findings might travel to the federal level in the new divided government.

Kirkland: I think that the insights definitely travel fairly well. Because if we just take the insight that the cost to lawmakers of impasse are an important feature here and we think about some of the instances that have happened in the past where shutdowns have happened or been avoided or the fact that we have managed to continue to raise the debt ceiling, it seems crucial. I think some of these insights might also be helpful for understanding budgeting in particular. It’s been a long time since Congress has routinely passed the set of normal appropriations bills, and so I think that we ought to expect delays in the future.

Grossmann: But there’s more research to be done, and Kirkland says looking to the states has the institutional diversity to see what matters for all kinds of outcomes.

Kirkland: One of the benefits of taking some ideas that come from a foundation at the federal level and looking at them at the state level is that you have variation in institutions across states that you just don’t have when you look at the federal government. The other thing is that when we look at the federal government when we look at something like legislative productivity, the unit of analysis tends to be a congress, and so that creates a small end problem that we can get around by incorporating data from all 50 states or from almost all states. That’s a general statement of some of the things that are helpful. I think there is really interesting work going on now that speaks to some of these theories or some ideas and theories that were developed thinking about the federal government that are now being tested at the state level.

I think some issues surrounding money and finance and lobbying are particularly interesting right now, and it seems like scholars are really interested in this right now and doing some really great work on state politics that draws on these theories. I think that the important focus is to account for the differences between the federal government and the states but also to think about what the politics of this are and what’s the same, and so what insights can we take away from states and apply to the way that we think about federal politics.

Grossmann: Schneer will also be looking at institutional design, going all the way back to the beginning and up to our present polarized era.

Schneer: The biggest remaining questions are probably about how polarization interacts with divided government. I don’t think we have immediate plans to tackle that question with this particular dataset. I think that for me at least, this all points to a broader set of questions about how the institutional design or institutional setup of our government, both in terms of electoral and legislative institutions, influences the ability of the government to respond to whatever the challenges are of the day as well as simply being able to respond to whatever constituents want at that point in time. For me at least, I’m doing more work in that area.

Grossmann: There’s a lot more to learn. Political Research Digest is available biweekly from the Niskanen Center and on iTunes. I’m your host, Matt Grossmann. Thanks to Patricia Kirkland and Benjamin Schneer for joining me. Please check out their articles and then join us next time.