Lesson 1: What You Will Learn

Welcome!

This first lesson aims to help you decide whether you want to learn what this website has to teach. It assumes you’re reading this because you’ve been assigned to as part of a college course you’re taking in political science, government, public policy or international relations. Most college courses in these fields are useful for anyone who is broadly interested in engaging with politics as a citizen, activist, journalist, elected official or policymaker. But positive political theory (which we’ll abbreviate from here as PPT) is a collection of tools built for use in one very specific pursuit: political science. So, if you’re here because you want to engage with politics through a practice other than political science, the material taught in these lessons (and maybe the course assigning you to read them) might not be what you’re looking for!

Of course, if you’re like most college students, you haven’t decided yet just how interested you are in any particular approach to politics. So this lesson will try to bring to light the aspects of political science’s and PPT’s approaches to making sense of politics that most sharply distinguish them from the approaches taken by other academic disciplines and fields of practice. The goal is to help you learn just enough to decide whether you want to continue learning more. Along the way you will also pick up a few foundational concepts that will help you make sense of all the lessons to follow, should you choose to continue.

A Crisis, A Mystery

We’ll start by looking at a real-world political crisis and a vexing mystery behind that crisis. We’ll then use that mystery to illustrate the distinct approach that political science and PPT take to making sense of politics.

First, the crisis: Housing is too expensive in many of the U.S.’s most productive metropolitan areas. To get a sense of the extent of the problem, explore the following chart. For each of the U.S.’s 15 most-populated metropolitan areas, the chart shows an estimate of the median value of houses in that metropolitan area in 2019, broken down by the number of bedrooms in the house.

Start by making sure you know how to read the chart by focusing only on houses with two bedrooms. Each colored dot in the left-most of the three rectangular frames in the chart represents the median value of a home with two bedrooms in one metropolitan area. Hover over (or touch, on a touchscreen) one dot clustered over the ‘2’ to see which metropolitan area the dot represents and the median value for a two-bedroom home in that area. For instance, by hovering over the yellow dot at the very bottom of the left-hand frame of the chart, you can see that the median value for a two-bedroom home in the Houston metropolitan area is about $115,000.

$150,000$200,000$250,000$300,000$350,000$400,000$450,000$500,000$550,000$600,000$650,000$700,000$750,000$800,000$850,000$900,000$950,000$1,000,000$1,050,000$1,100,000Metro Area Median Home Value234+Number of Bedrooms New York City Median value of homes with 2 bedrooms: $375,000 Los Angeles Median value of homes with 2 bedrooms: $500,000 Chicago Median value of homes with 2 bedrooms: $192,000 Dallas Median value of homes with 2 bedrooms: $153,200 Houston Median value of homes with 2 bedrooms: $115,000 Philadelphia Median value of homes with 2 bedrooms: $198,000 Washington DC Median value of homes with 2 bedrooms: $260,000 Miami Median value of homes with 2 bedrooms: $200,000 Atlanta Median value of homes with 2 bedrooms: $160,000 Boston Median value of homes with 2 bedrooms: $375,000 San Francisco Median value of homes with 2 bedrooms: $800,000 Phoenix Median value of homes with 2 bedrooms: $200,000 Riverside Median value of homes with 2 bedrooms: $250,000 Detroit Median value of homes with 2 bedrooms: $130,000 Seattle Median value of homes with 2 bedrooms: $350,000 New York City Median value of homes with 3 bedrooms: $400,000 Los Angeles Median value of homes with 3 bedrooms: $620,000 Chicago Median value of homes with 3 bedrooms: $235,000 Dallas Median value of homes with 3 bedrooms: $200,000 Houston Median value of homes with 3 bedrooms: $174,000 Philadelphia Median value of homes with 3 bedrooms: $230,000 Washington DC Median value of homes with 3 bedrooms: $359,000 Miami Median value of homes with 3 bedrooms: $320,000 Atlanta Median value of homes with 3 bedrooms: $200,000 Boston Median value of homes with 3 bedrooms: $460,000 San Francisco Median value of homes with 3 bedrooms: $900,000 Phoenix Median value of homes with 3 bedrooms: $270,000 Riverside Median value of homes with 3 bedrooms: $357,000 Detroit Median value of homes with 3 bedrooms: $180,000 Seattle Median value of homes with 3 bedrooms: $450,000 New York City Median value of homes with 4+ bedrooms: $560,000 Los Angeles Median value of homes with 4+ bedrooms: $800,000 Chicago Median value of homes with 4+ bedrooms: $350,000 Dallas Median value of homes with 4+ bedrooms: $345,000 Houston Median value of homes with 4+ bedrooms: $270,000 Philadelphia Median value of homes with 4+ bedrooms: $400,000 Washington DC Median value of homes with 4+ bedrooms: $500,000 Miami Median value of homes with 4+ bedrooms: $450,000 Atlanta Median value of homes with 4+ bedrooms: $310,000 Boston Median value of homes with 4+ bedrooms: $650,000 San Francisco Median value of homes with 4+ bedrooms: $1,100,000 Phoenix Median value of homes with 4+ bedrooms: $360,000 Riverside Median value of homes with 4+ bedrooms: $450,000 Detroit Median value of homes with 4+ bedrooms: $310,000 Seattle Median value of homes with 4+ bedrooms: $650,000

Estimated median home values in the chart were computed and published by the U.S. Census Bureau using data from the 2019 American Housing Survey. They were collected from the American Housing Survey Table Creator, https://www.census.gov/data/data-tools/ahs-table-creator.html, on July 24, 2022. Chart built using the wonderful and amazing Observable.

Now that you understand how to read the chart, make sure you take account of two critical facts it makes apparent:

First, the values of homes of similar sizes in most of the U.S.’s large metropolitan areas are clustered within a relatively narrow range. But there are a small number of areas where home values (and thus home prices and rents) are far out of line. By hovering over the dots spread thinly across the upper reaches of each frame, you can identify which metro areas have distinctly high housing prices: the Boston, New York City, Los Angeles, San Francisco and Seattle metro areas. Prices in these areas are not only high relative to those in other large U.S. metros. Their absolute levels – even for modestly sized dwellings – are shocking. For instance, while a typical two-bedroom home was valued between $100,000 and $200,000 in most of the U.S.’s large metro areas in 2019, median two-bedroom home values in the Boston, New York City, Los Angeles, San Francisco and Seattle metro areas ranged in 2019 from $350,000 to $800,000.

Second, the five areas where housing prices are extraordinarily high are also places that stand out, relative to other U.S. metros, in the extraordinary economic opportunities they offer to their residents. These expensive metro areas are places where industries are clustered in which the U.S. leads the world in productive capacity – i.e. software, finance, biotech, pharmaceuticals and aerospace. These are places where good jobs, high wages, and profitable business opportunities (both in and out of these areas’ world-leading industries) are abundant. This is why their high housing prices amount to a political crisis. Most U.S. households cannot afford the astronomical costs of housing in these areas, and thus most U.S. households are are excluded from the extraordinary economic opportunities they offer. Moreover, thousands of lower-income, long-term residents of these areas have been displaced during the past 30 years as rents have risen beyond the level they can afford.

How did housing in these particular metropolitan areas get so expensive? And what has kept prices in these places at destructively high levels for over 20 years?

At one level, the explanation is simple: Demand for housing in these places has grown, and not enough new housing has been built in response to that growth. The result has been more households competing for each available housing unit, and increasingly intense bidding wars over the units that come to market. This all started in the late 1990s, when productivity and employment in a few industries clustered in these metro areas (again: in software, finance, biotech, pharmaceuticals and aerospace) began to soar. This raised the value of living in these places and thus both the total number of persons seeking to live in them and the amounts persons were willing to pay to do so.

These areas’ booming economies, however, do not by themselves explain the explosive growth in their housing prices. There are many other U.S. metro areas – such as the Austin, Boise and Salt Lake City areas – have also seen extraordinary growth in economic opportunity and housing demand during the past 30 years, but have experienced only mild increases in home prices and rents. So there’s something distinct going on in the Boston, Los Angeles, New York City, San Francisco and Seattle areas that gets in the way of housing construction sufficient to meet the growth in demand.

And whatever it is that blocks that construction is extraordinarily powerful. In a 2018 study, Glaeser and Gyourko (2018) show that since the late 1990s, the prices of modestly sized homes in these areas have never fallen below 1.45 times the total cost of building them – including the costs of land, labor, materials, profits paid to investors and interest and principal paid on construction loans. This means that the profits developers could have earned from building more housing in these expensive metro areas during each of these years have been extraordinary. Yet Glaeser and Gyourko find that the number of construction permits issued in these areas in each year during this period was only enough to expand total housing supply in each by less than 1 percent. So the process blocking construction, whatever it is, is strong enough to deny real estate developers – actors who are often assumed to be political powerful in urban areas – billions of dollars in profits. What forces might be at work in Boston, Los Angeles, New York City, San Francisco and Seattle that are powerful enough to block, year after year, so much needed and extraordinarily profitable construction?

Thinking Small

When political scientists try to shed light on outcomes like the U.S. housing crisis, they start by thinking small. To get a sense of what it means to think small, consider the following image:

Land Parcels in a Small Part of Cambridge, MA

This and the following images of Cambridge, MA built by combining satellite imagery from Google with tax parcel data from the State of Massachusetts, retrieved from https://www.mass.gov/info-details/massgis-data-property-tax-parcels on June 22, 2022.

This is an arial view of a small piece (about 0.4 miles on each side) of the City of Cambridge, Massachusetts, which is one of about one hundred and fifty cities and towns that make up the Boston metropolitan area. Overlayed on the image are about one thousand quasi-rectangular shapes outlined in blue. Each one of those rectangles is a single “parcel” of land. A parcel (also sometimes called a “lot”) is the smallest unit of land that can be legally bought and sold for the purpose of constructing buildings of any kind, including housing. New housing construction always starts with one or more parcels: A developer buys a parcel and any buildings sitting on it, and then adds new housing by constructing new buildings on the purchased parcels or modifying the buildings on those parcels so that they contain more housing units. Thus, every single one of the about-one-thousand parcels you see in the above image is a site on which additional housing might have been added to the total supply of housing in the Boston metropolitan area during the past 30 years.

So, as you’ve been reading this lesson, you might have been thinking of the U.S. housing crisis as “one big thing”. But the image above shows that that way of thinking about the crisis is incorrect. In fact, the U.S. housing crisis is millions of little things, happening simultaneously. Again, there are about one thousand parcels in the image above, and that image only covers a tiny portion of one of about one hundred and fifty cities and towns in the Boston area. The Boston area as a whole contains about one and a half million parcels. During the past 30 years, each one of those parcels has been the site of scores of decisions determining whether more housing would be built. Those millions of individual decisions about millions of individual parcels (in Boston and the four other metro areas where the crisis has emerged) aggregate to form the U.S. housing crisis.

So, to understand the housing crisis, we have to start by thinking small.

Thinking Small

When trying to shed light on any political outcome or process, political scientists start by thinking small. To think small is to focus one’s attention on the single actions taken by one or just a few persons that combine with many other actions by many other persons to produce political outcomes.

Let’s think small, then, by working with a different kind of image from the one above – one that helps us focus on just a few instances of possible (or possibly forgone) housing construction. like this:

Four Land Parcels in a Small Part of Cambridge, MA

What you see here is the same satellite image of a portion of Cambridge, MA, but this time with only four of all of the parcels in the image outlined in blue. You’ll examine these two pairs of parcels more closely because they each help us to think concretely (pun intended!) about what additional housing construction in one of the U.S.’s very expensive housing markets would entail. Here are close-up views of each of those pairs of parcels:

Two Parcels on Cogswell Avenue

Two Parcels on Creighton Street

Start by considering the right hand-image first, because it’s an especially simple case. Notice that the parcels at 44 and 48 Creighton Street are about the same size, and that the buildings on those parcels have almost exactly the same footprint. But, according to the City of Cambridge Property Database (as of June 22, 2022 when this lesson was written), 44 Creighton Street is a duplex (i.e. it contains two apartments), while 48 Creighton Street is a triplex (i.e. it contains three apartments). Evidently, a developer or building owner could have increased the supply of housing units in the Boston area by one by adding one more housing unit to 44 Creighton Street, but has (thus far) been unable to or unwilling to do so.

Now consider the parcels at 19 and 22 Cogswell Avenue, in the left-hand image above. Notice that these two parcels are very close in size. Yet (again, as of June 22, 2022), the buildings sitting on them are quite different. Six single-family townhouses sit on the 19 Cogswell Avenue parcel. 22 Cogswell, on the other hand, hosts an apartment building containing seventeen apartments. So, evidently, a developer or building owner could have increased housing supply in the Boston area by eleven units by tearing down the townhouses at 19 Cogswell and replacing them with a building with seventeen apartments like the building across the street, but has (thus far) been unable or unwilling to do so.

The failure to add just one housing unit at 44 Creighton Street and the failure to add just eleven units at 19 Cogswell Avenue are each by themselves insignificant. A handful of new units at just one or two parcels will not reduce prevailing rents and prices in the Boston area as a whole. But remember: The housing crisis that has excluded and displaced so many people from the abundant economic opportunities in Boston and the U.S.’s other expensive metropolitan areas is nothing more or less than millions of similarly “insignificant” decisions happening simultaneously. So we cannot understand the crisis without an understanding of individual, insignificant decisions like the failure to add one more unit to 44 Creighton Street or eleven more units to 19 Cogswell Ave.

Pause and complete check of understanding 1 now!

Politics = Aggregation

You’ve now “thought small” by imagining what might have gotten in the way of housing development on two specific parcels of land. Moreover, you’ve imagined different barriers to development at work on the two different parcels you considered.

It’s critical at this point to notice an implication of your capacity to imagine that there are different processes working to block different individual instances of housing development. When we think small – focusing our attention on one or just a few of the millions of individual actions that aggregate up political outcomes – we recognize the rich complexity that characterizes the lives of individual persons, and even the complexity of individual moments within any one person’s life. Behind every human action is a story unique to the person who took that action, and even unique to the moment in that person’s life when that action took place.

This casts doubt on our ability to ever construct a correct explanation for any significant political outcome like the U.S. housing crisis. The image below shows one more view of the one-thousand-or-so land parcels sitting on one tiny portion of Cambridge, Massachusetts. This view emphasizes the overwhelming variability we confront when we hope to explain an outcome that results from the aggregation of millions of individual actions at millions of distinct locations, taken over the course of three decades. It shades each parcel with a randomly-chosen color that is distinct from the color shading every other parcel in picture. (Every parcel really is shaded a unique color, although the difference between the colors of some parcels might be too small for your eye to detect!) At each one of these parcels, there is some aspect of the story behind the failure to develop additional housing on that parcel that is distinct from the story behind the failure of development on every other parcel.

Four Land Parcels in a Small Part of Cambridge, MA

A correct and complete explanation of the housing crisis would be one that accounted for every one of these millions of distinct stories. But of course collecting and knowing every one of these stories in all their relevant details is impossible. Thus no explanation for the crisis that anyone can construct will correctly or fully account for all the events that caused and continue to sustain the crisis. Thus every explanation of the housing crisis – or any political outcome like it – will be incorrect.

The impossibility of constructing any explanation of the U.S. housing crisis that we can know to be correct is not at all unique. In fact it stems from a universal feature of politics.

Politics = Aggregation

Political outcomes result from the aggregation of many distinct actions by many different persons.

This fact about politics is at the heart at everything you will learn in these lessons. So we’ll use an image throughout to help us keep this fundamental fact in mind:

This image walks through, from left-to-right, the universal structure of political processes. At the left side of the image are a large number of narrow rectangles. Each one symbolizes a single action by a single person. The variety of colors of these rectangles emphasize the diversity of circumstances at work in any large collection of distinct human actions.

The square on the right end of the image stands in for a generic political outcome. The persistently high housing prices in the Boston, Los Angeles, New York City, San Francisco and Seattle metro areas is one such outcome. Other political outcomes we’ll consider in these lessons include policies towards international trade adopted by nation states, the widespread under-representation of women in elected offices in the world’s representative democracies, national and sub-national government policies that tax wealth and income and redistribute it in the form of social insurance and welfare programs, outbreaks of protest and non-violent direct action by citizens against repressive state policies and repressive social structures, patterns of state repression and reactionary terrorism that occur in response to protest, and practices by governments and other large institutions that drive the production of environmental externalities, such as atmospheric carbon dioxide.

The purple triangle linking the multiple small rectangles on the left to the single large square on the right depicts the processes (whatever they might be) through which many individual actions aggregate to produce a given political outcome. These processes of aggregation are, ultimately, the things we try to make sense of through political science.

Pause and complete check of understanding 2 now!

Models Not Explanations

So correct explanations of the U.S. housing crisis (and significant political outcomes more generally) aren’t possible. How, then, are we to make sense of politics?

This is not merely an academic question. Political outcomes are determinative of so much that matters to human well-being. The U.S. housing crisis, for instance, prevents millions of persons from participating in some of best economic opportunities in the U.S. economy. And that harm is minor compared to the calamities regularly produced by politics, such as war, the failure to reduce CO2 emissions at a rate fast enough to avert global climate change, and anemic public investments in the prevention of and preparation for outbreaks of pandemic disease. Given the huge stakes of politics, many of us feel compelled to act to try to influence political outcomes for the better. Knowing what actions will effectively shape political outcomes, however, requires effective understanding of the processes that drive politics. Yet complete and accurate knowledge of these processes, as we’ve argued above, is unobtainable. All of us who hope to act effectively to influence politics for the better must, then, decide how to cope with its un-knowability.

There is no single universally, or even widely, adopted view of how best to respond to this dilemma. You’ll see a variety of responses employed as you move across academic fields, and across different institutions outside of academia that attempt to influence politics. What you’ll learn in these lessons is the particular response to the un-knowability of politics used by political science in general and positive political theory more specifically. Models are the core tool in this response.

Model

Because complete explanations of political outcomes aren’t possible, political scientist instead work to develop models of political outcomes. Models are partial representations of political processes. Because models are partial representations, they don’t fully or accurately describe the politics they represent. Instead, they are tools that help us clarify the implications of our ideas about how any given political outcome might occur.

Let’s jump right to an example by building a model of the U.S. housing crisis on the basis of some recent research by political scientists Katherine Einstein, David Glick and Maxwell Palmer (2019a; 2019b). Einstein, Glick and Palmer point out that the laws of most town and city governments in the United States prohibit developers from beginning housing construction on a parcel without first getting a permit from the town or city in which that parcel lies. Further, town and city ordinances usually require developers who wish to increase housing density on a parcel (e.g. convert a duplex to a triplex or replace a set of townhouses with an apartment building) to get what’s called a “special permit”. To get a special permit the developer must first submit an application to the relevant municipal government that describes in detail what they wish to build. The municipal government then issues a public announcement that they are reviewing the developer’s application, and often sends letters to owners of properties near the proposed building site notifying them of the upcoming review. Finally, the municipal government holds a “public hearing” where members of the public may show up and express their views on whether the developer’s application ought to be approved and the permit to begin construction granted. Only then, after the public hearing, does the municipal government make a final decision about whether to grant the developer’s application for a permit.

Einstein, Glick and Palmer collected the transcripts of every public hearing reviewing applications for special permits for housing construction filed between 2015 and 2017 in 97 cities and towns in the Boston metropolitan area. They and their research assistants read every comment by every member of the public that spoke at every one of those hearings, and recorded whether the comment supported, opposed or was neutral on the proposed housing development under review at the hearing. The transcripts also included the names and addresses of every member of the public who commented at each hearing. Using that information, Einstein, Glick and Palmer were able to identify whether each commenter was a homeowner or renter. They were also able to make reliable inferences about each commenter’s age, gender, race and ethnicity.

They found, first of all, that the members of the public who show up and speak at public hearings on proposed housing construction projects express overwhelming opposition to housing development. The figure below shows the number of comments in-support-of and (conversely) in-opposition-to the proposed housing construction projects under review in a random sample of 30 public hearings from all hearings in Einstein’s, Glick’s and Palmer’s dataset in which there were with at least 10 non-neutral comments. Each tiny rectangle in the figure represents one comment from a member of the public in one public hearing. The rectangles are organized into rows according to the proposed construction project they pertain to, and then colored and arrayed to the left and right of the thick vertical line down the middle of the figure according to whether they expressed support of or opposition to the relevant construction project. The projects are sorted from top to bottom in order of the number of comments opposing each project. Comments supporting construction are shaded green and placed to the left of the line. Comments opposing construction are shaded red and placed to the right of the line. You can hover your mouse over (or touch, on a touchscreen), each rectangle to see the name and location of the proposed project, the position of the comment on that project, and the reasons for that position as recorded by Einstein, Glick and Palmer.

Comments in the chart are taken from replication data for Einstein, Palmer, and Glick (2019b), accessed at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/RRF7EU, on July 24, 2022. Chart built using the wonderful and amazing Observable.

Notice that in the overwhelming majority of proposed construction projects, comments opposed to the project substantially outnumber comments in-support-of the project. Also notice the large proportion of projects in the figure in which the all the comments from the public express opposition to the project. Indeed, there are about 4.5 comments opposing housing construction for every one comment supporting construction in Einstein’s, Glick’s and Palmer’s dataset as a whole.

Einstein, Glick and Palmer also find that the group of persons who show up and comment at these public hearings is very different from the population of registered voters in the Boston area as a whole. Specifically, older persons, men, frequent voters, homeowners, and white, non-latino and non-asian persons are all over-represented among persons who show up to public hearings on proposed housing construction relative to registered voters as a whole. Einstein, Glick and Palmer argue that these traits make it likely that attendees at these hearings are more opposed to housing construction than are registered voters in general.

Taking these results as an inspiration, here’s a model of the long-term mis-match between housing prices and construction costs in the Boston, New York City, Los Angeles, San Francisco and Seattle metropolitan areas:

A Model of the Housing Crisis
  • Because prevailing housing prices are very high relative to land and construction costs in these five metro areas, developers would build enough housing to substantially increase housing supply (which would bring prevailing prices down to levels close to those in other U.S. metros) if they were permitted to do so.
  • However, because these areas are already densely developed, substantial increases to housing supply in each area would require substantial increases in housing density. Thus the kinds of construction projects that are needed to substantially increase housing supply – e.g. projects that convert duplexes to triplexes or townhouses to apartment buildings – require “special permits” from the relevant town and city governments. Under current law, these permits can be granted only after public hearings.
  • Developers do occasionally seek permits for projects in these five metro areas that would increase housing density. But when they do, the members of the public who show up to the required public hearings are persons who are unusually hostile to housing development. These persons express overwhelming opposition to the proposed construction projects and as a result, the required permits are usually denied, substantially delayed, or only approved after the developer modifies the proposal so that it increases housing density less or not at all. Knowing all of the above, developers anticipate that construction permits are unlikely to be granted for projects that increase housing density, and rarely bother to seek permits for such projects in these metro areas.
  • If town and city governments changed their laws so that special permits and public hearings were not required for projects that increased housing density, or if by some means persons who were supportive of housing construction were mobilized to appear at public hearings on proposed construction projects that would increase housing density and advocate for the permits for those projects, housing construction sufficient to increase supply would eventually occur and the crisis would abate.

Why do we call the account of the housing crisis above a “model”? Why don’t we call it what it actually appears to be: an explanation of the crisis? It is, after all, an assertion about what’s causing the crisis and what, if it happened, would cause the crisis to stop. Isn’t that just an explanation?

You can take the model above as an explanation for the housing crisis if you want to. But as an explanation, it’s a failure. It makes all kinds of assertions about the processes driving housing development that are at best speculative and at worst demonstrably false. No doubt, for instance, one can point to innumerable instances of foregone housing development in the U.S.’s expensive metro areas that were driven by factors other than public opposition expressed at special permit hearings. Indeed, Einstein’s, Glick’s and Palmer’s findings suggest that opposition in public hearings slows down or blocks housing development in many cases, but they do not establish a causal effect of public opposition at permit hearings on housing development, or give any measure of the magnitude of the effect of that opposition relative to other possible barriers to development. What the “explanation” above does is take one process that is probably at work in many cases of foregone housing development and asserts that all cases of foregone development are driven by that one process.

Political scientists construct models, and are careful to call them that, to help us keep track of the impossibility of constructing completely correct accounts of political events. Remember: the housing crisis (like all significant political outcomes) results from the aggregation of millions of decisions by millions of persons. The full chain of events producing the crisis is un-knowable. While we can offer explanations of political outcomes, none that we can offer will be especially good. Since good explanations aren’t an option, we aim for something else.

“Models”, in contrast to explanations, are not supposed to be completely factual accounts of the things they depict. Thus no one risks mistaking a model for the truth. Think, for instance, of a child’s plastic toy airplane, or of a simulation of a real-world city built in Minecraft, or of Google Map’s interactive depictions of road networks. All of these are models. And models are deliberately incomplete representations – i.e. they are similar to the things they represent in only some respects. Indeed, in a model, unlike an explanation, dissimilarity and even inaccuracy can be good things. Road maps, for instance, are typically much smaller than the physical spaces they represent, and in that respect are inaccurate. But a road map of a city as large as the city itself would useless. And useful – not accurate in every respect – is what models are supposed to be. (Clarke and Primo 2012)

A Sharp Turn

Models of politics help surface critical questions about the processes they represent. These questions often become apparent when we notice aspects of a model that seem to be at odds with reasonable expectations. For instance, in the representation of the U.S. housing crisis offered by the model above, the relatively small numbers of persons who show up at public hearings on proposed construction permits exercise immense political power – power sufficient to damage the economic prospects of the overwhelming proportion of residents of the United States. Could such a small number of persons really be the cause of so much loss borne by so many others? Surely that’s not how it really works!

But with this absurdity in mind, reflect for a moment on who gains and who loses each time a construction project that would increase the density of housing on any given parcel of land in Boston or one of the U.S.’s other expensive metro areas is (by whatever means) blocked. Einstein’s, Glick’s and Palmer’s accounts of the objections voiced at the hearings they study suggest that any harms of an increase in housing density at any given parcel are limited to the persons who live in the immediate vicinity of that parcel. For instance, parking is very scarce on most already-densely-developed blocks in the Boston area. Each additional housing unit added to a given block increases the number of households on the block with one or more cars to park, which increases the competition for the fixed number curbside spots shared by the households living nearby. This imposes a very noticeable inconvenience. But it’s an inconvenience borne only by the persons who live near enough to the parcel in question to need street parking in its immediate vicinity.

In contrast, for the vast majority of persons living outside of the immediate vicinity of any given parcel, additional housing units on that parcel will be beneficial. Most persons in any metro area living outside any parcel’s immediate vicinity are renters, rather than homeowners, and for them, more housing units in the area as a whole means more housing options and (very marginally, since we’re only thinking about one parcel!) lower prevailing rents. Moreover, persons residing outside that metro area have an interest in preserving the option to move to that area at the lowest possible cost to access its jobs and business opportunities. All of these persons, even if they are homeowners, will benefit from greater housing supply and thus lower rents and prices in that metro area.

The key thing to notice here is that in any given instance in which additional housing units might be built on any one parcel of land in one of the U.S.’s expensive metropolitan areas, the persons who stand to benefit from the construction of those additional housing units vastly outnumber (on the order of something like 1,000,000 to 1) the persons who stand to be harmed or inconvenienced by that construction. So, although our model of the housing crisis might incorrectly represent the crisis by attributing it entirely to the activism of the persons who show up and complain at construction permit hearings, it helps us to notice something undeniably true about the crisis: Time after time, housing construction is somehow blocked to the benefit of a tiny number of persons and at the expense of millions of others. How is such an imbalance of power between the few and the many possible?

Encountering this puzzle about the housing crisis, many political scientists will make a sharp turn in the direction of their thinking. We are now going to take the same sharp turn in our focus in order to make sure you notice and take stock of what we’re turning our attention away from and what we’re instead turning our attention towards. The move we make at this point illustrates the most important distinction between the approach to exploring politics taken by political science and PPT and the approach taken by most other fields of study and practice. Before continuing with these lessons, make sure you are game to take this turn yourself.

The turn that we (like many political scientists) will take on encountering this puzzle – again, how is such an imbalance of power between the few and the many possible – is away from the details that distinguish the U.S. housing crisis from other political processes. Notice how weird it is to turn attention away from specific details of the crisis at this point. We just uncovered a truly puzzling and intriguing aspect of the crisis! Doesn’t this make the details unique to the crisis more interesting? Doesn’t this puzzle compel us to learn more about how housing development fails in the Boston, Los Angeles, New York City, San Francisco and Seattle areas?

Yes it does! But it also draws attention to something else that at least some persons will find equally intriguing: The concentration of benefits on just a few people at the expense of millions of others is a widespread feature of political processes. We find instances of it occurring in political processes across the entirety of recorded human history and in political processes taking place at different places across the globe at any one moment in time. So the mysterious imbalance of power at work in the U.S. housing crisis also makes one wonder about politics in general. Might there be some processes at work in all political outcomes that, under the right circumstances, drive those outcomes to benefit tiny groups of persons at the expense of everyone else?

Pursuing that question requires turning our attention away from facts specific to any one political process, and instead focusing on features shared in common across many different contexts. And so, if we take this sharp turn, we stop the effort to learn more specific detail about housing construction in a handful of U.S. cities, and we instead try to think abstractly – about the how the very few features that all political processes have in common might create advantages for small groups relative to large ones.

A Little PPT

This brings us to Positive Political Theory (PPT). PPT is a set of tools political scientists use to build models that represent aspects of politics shared in common across a wide array of political contexts. In this section we will build and analyze a model using a number of the PPT techniques you’ll learn in this course. The goal is not for you to understand every step we take in building and analyzing the model. After all, we’ll use techniques that take some work to get a handle on and that you won’t learn until you work through the following lessons. Instead, the goal is for you get a feel for what PPT typically entails. You should come away with a sense of whether or not you want to learn more. We’ll also highlight a few features common to all models built using PPT. Noticing and understand these features will help you grasp the material in the remaining lessons, should you choose to keep going.

The model we build is based on one of the most influential works of PPT, a book called The Logic of Collective Action, first published by economist Mancur Olson in 1965 (Olson 2009). Olson’s book explores situations that have come to be called collective action problems. A group of persons face a collective action problem when they are in a situation characterized by all the following:

  • Common Interests: The persons in the group have a common interest in some political outcome – with every member of the group benefiting if that political outcome occurs.
  • Effective Action: The extent to which the commonly-desired political outcome occurs depends on the amount of effort or action jointly undertaken by the members of the group, with the outcome occurring with greater likelihood or to a greater extent as the total effort expended or action taken by the group members increases.
  • Costly Action: Exerting effort or taking action is costly for each group member. So every group member would prefer to exert less effort or take less action, if they could somehow do so without diminishing the likelihood or the extent of the commonly-desired political outcome.

For instance, suppose a developer has applied for a permit to build several new housing units on a given parcel of land. Now imagine the group of persons consisting of all those living nearby the parcel who prefer those new housing units not be built. Since these persons all prefer the units not be built, they share a Common Interest in the construction permit being denied. Suppose further that each of these persons has the ability to take action that will make the developer’s permit less likely to be granted. Specifically, each can choose whether to show up at a public hearing and argue against the developer’s permit application. If enough sufficiently persuasive arguments against the permit are made at the hearing, the developer’s permit will be denied. Thus each of these persons has the option to take Effective Action. Finally, suppose that each of these persons would rather not, all else equal, put in the effort and time to prepare a persuasive argument against the construction permit and go to a public hearing to deliver it. Perhaps, for instance, the hearing is scheduled to occur at 7pm on a Tuesday, and the members of the group all have have busy lives. Some of them will have to take time off of work to attend; Some would have to pay a babysitter to watch their kids; Others would rather spend the evening hanging out with friends or otherwise relaxing. In this sense, then, the effective action each member of the group can take is also Costly Action.

While this specific scenario might help to clarify what a collective action problem is, the definition of a collective action problem is highly abstract. It can apply to groups of persons involved in any one of an infinite variety of political processes. So, by exploring the characteristics of collective actions problems in general, rather than any specific collective action problem arising in any one political process, we take the “sharp turn” described above – i.e. we explore a process common to a wide variety of political settings, rather than diving deeply into the details of any one instance of politics.

Moreover, since we want to explore a feature common to many different political processes, a model is exactly what we need. Recall, a model is a deliberately incomplete representation of a political process. And we want to focus our attention on only subset of features of any political process – i.e. common interest, joint action and costly action. Thus we will build a model that represents only these elements, and ignores the other factors and processes that might be at work in any one political situation. Our model then, will not in any way be a completely accurate representation of any one political process. It will simply help us to clarify our thinking about one aspect of politics that we think might be at work in any one of a variety of political contexts.

The Model

With all that in mind, let’s build a model of a collective action problem: Imagine a political process that involves N persons, where N is a positive number. By using a variable (N) we can explore whether and how the model’s implications depend on the number of persons in the group.For instance, we can compare the model when N = 2 to the case where N=\text{2,000,000}.

Because we want this model to represent a collective action problem, we must add elements that depict common interest, effective action and costly action:

To capture common interest, suppose that these N persons are involved in a political process that might result in an outcome that is preferred by each one of these persons. We’ll call that outcome the favorable outcome.

To capture effective action, suppose that each of these N persons must choose a level of effort to put towards activity influencing the outcome, and that the likelihood with which the favorable outcome occurs depends on the sum of the levels of effort chosen by all N persons. More specifically, label the persons “1”, “2” and so on through N, so we can refer of each of them separately. Suppose each person i (where i is 1 or 2 or … and so on through N) must choose an effort level, which is a number e_i greater than or equal to 0. Then, let E denote the sum of the levels of effort chosen by the N persons – i.e. E = e_1 + e_2 + \cdots + e_N. Finally, let L be a number that represents the likelihood that the favorable outcome occurs, and suppose that the likelihood with which the favorable outcome occurs given the sum of the effort levels E is given by L = \frac{E}{E + 10} or, equivalently: L = \frac{e_1 + e_2 + \cdots + e_N}{e_1 + e_2 + \cdots + e_N + 10} Importantly, this means that:

  • The likelihood of the favorable outcome occurring is a number between 0 and 1 regardless of the levels of effort chosen by the N persons;
  • The likelihood increases with each increase in the sum of the levels of effort;
  • The likelihood of the favorable outcome increases more slowly with each additional increase in the sum of the levels of effort.

You can see all these characteristics in the following chart which graphs the likelihood of the favorable outcome as a function of the sum of the N persons’ effort levels:

Adding a representation of costly action will complete our model. We’ll do so using a tool that you’ll learn much more about in subsequent lessons called a utility function. A utility function represents a person’s preferences over a set of events by assigning a number to each event in the set, with higher numbers assigned to events that the person finds more preferable. The number assigned to any given event is called that person’s utility level at that event.

In this case, we want to use a utility function for each person that represents two ideas: First, each person prefers, all else equal, that the favorable outcome occur with greater likelihood. Second, each person prefers, all else equal, to exert less effort. So, for each person i in the group, suppose that if the likelihood of the favorable outcome is the number L and person i exerts effort level e_i, then person i gets utility level 1000\times L - e_i Note that each person i’s utility level increases as the likelihood L of the favorable outcome gets larger. This completes our depiction of these persons’ common interest in achieving the favorable outcome. Further, since each person i’s effort level e_i is subtracted from her utility level, each person’s utility level decreases in her effort, completing our depiction of costly effort.

Here’s a summary of our model:

A Model of Collective Action
  • There are N persons labeled “1”, “2” and so on through “N”.
  • Each person i chooses an effort level e_i, which is a number greater than or equal to 0.
  • Letting E = e_1 + e_2 + \cdots + e_N, an political outcome that is favorable to each of the persons occurs with likelihood L = \frac{E}{E + 10}.
  • Each person i gets utility level 1000\times L - e_i.

What PPT is Like

Before we use the model to clarify our thinking about collective action, it’s worth taking note of a few characteristic of PPT models evident here. If you choose to continue with these lessons, you’ll see many more models with these characteristics.

Notice first of all just how radically abstract this model is. Abstraction is perhaps the most notable feature of models built using PPT. This model, for instance, includes no details at all that might cause us to think that it represents any one specific political context, or even any one category of contexts. It focuses purely on a few aspects of politics that might be present in any political situation.

Our model’s extreme abstraction, moreover, makes its usefulness for gaining understanding of any particular political situation limited and contingent – Limited because the model only represents some aspects that might be at work in any given political process; Contingent because the relevance of any insights we might gain from the model to any given political context depends on how the limited aspects the model represents interact with other processes at work in that context that the model ignores. For instance, it seems clear that the model we’ve built here is at least a plausible representation of the common interests of groups of persons who are similarly affected by any one instance of housing construction, and represents the potential for those persons to take effective and costly action in pursuit of their common interests. But that describes only one of a myriad of processes at work driving the politics of the housing crisis. The extent to which any insights we might gain from this model will help us to understanding the housing crisis depends on the importance of collective action in the crisis relative to other processes. We cannot know that relative importance without examining the specifics of the housing crisis more closely.

PPT models, then, are like tools in a toolbox. When we have a big job to do – for instance, assembling a piece of furniture from dozens of separate parts – no one tool will be by itself sufficient for completing the entire job. We’ll use a screwdriver for one part of the job, a hammer for another part, and a wrench for another. Moreover, some tools won’t be useful at all for some jobs. Building effective understanding of any political context is a complicated, multi-step process. Any one PPT model will be useful for only some of the steps involved, if at all.

You probably also noticed that our model uses math. In general, PPT uses mathematics to build models. You’ll see as you continue with these lessons that although PPT models are almost all mathematical, the math isn’t valued in-and-of-itself. Instead, math is used as a way to try to be as unambiguous as possible in specifying the features of a model. This in turn allows us to be absolutely clear about what any given PPT model depicts.

Finally, and most importantly, PPT models depict the aggregation of multiple actions by multiple persons into single political outcomes. You can see this by mapping the elements of the model we constructed above (for the case in which the number of persons N is equal to 15) onto our graphical representation of aggregation:

Discovering What We Think

Having used PPT to create a model of collective action, what can we use the model for? Recall that, above all else, this is a model, not an accurate rendering or recording of any actual politics. Indeed, building this model entailed a sequence of acts of imagination. Look back at the words we used to construct it:

  • Imagine a political process that involves N persons…”,
  • Suppose each person i must choose an effort level…”,
  • Suppose that the likelihood of the favorable outcome given the sum of the effort levels E is given by L = \frac{E}{E+10}…”.

Evidently, the model amounts to a collection of thoughts. It depicts an imagined situation, every detail of which we invented. Whatever we can learn from the model, then, ultimately is a lesson about our own thinking. Put more simply, this model, like all models built using PPT, is useful for discovering the implications of what we think.

What are those implications? Collective action models like this one are extraordinarily fertile. They have inspired and are still inspiring all kinds of intriguing ideas and novel questions. These questions in turn have motivated thousands of studies – not just in political science, but in economics, psychology, sociology, anthropology, philosophy and even in life sciences such as evolutionary biology. We’ll develop just one of these implications here, so that you can get a sense of what working with these models entails.

Recall that we built this model specifically to depict a group of persons who share a common political objective. By exerting effort, each person in the group can raise the likelihood of an outcome that every member of the group desires. In this sense, this model is one that depicts a group persons who might be described as mutual allies. And yet, when we examine the model carefully, we discover a conflict between group members that might not otherwise be apparent.

To see the conflict, recall how the effort levels of the group members determine the favorable outcome. If the sum of their effort levels is E, then the likelihood L of the favorable outcome is given by L = \frac{E}{E+10} The conflict within the group becomes apparent when we think of the relationship between the group members’ efforts and the likelihood of the favorable outcome “in reverse”. First, we imagine the favorable outcome occuring with some particular level of likelihood. Then, we ask what effort levels the group members would have to exert to cause the outcome to occur with that particular level of likelihood.

So, thinking in reverse: By rearranging the terms of the above equality, we can see that for the outcome to occur with any particular level of likelihood \hat{L}, the sum of the group members effort levels E must satisfy E = \frac{10\hat{L}}{1-\hat{L}} Looking at the relationship between the total effort of the group and the likelihood of the outcome in this way reveals a question that we did not contemplate when building the model: How can the burden of achieving any particular likelihood of the favorable outcome be distributed between the group members?

The answer is: There are an infinity of ways to distribute the burden of any particular likelihood of the favorable outcome between group members. This is true regardless of the number N of members of the group. But it’s easiest to grasp in the simplest case in which N = 2, so there are only two persons in the group. If we take a particular level \hat{L} of likelihood of the favorable outcome, then any combination of effort levels e_1 and e_2 for persons 1 and 2 that sum to \frac{10\hat{L}}{1-\hat{L}} causes the favorable outcome to occur with likelihood \hat{L}. This includes each of the following distributions of the burden:

  1. Person 1 exerts effort level 0 and person 2 exerts effort level \frac{10\hat{L}}{1-\hat{L}}.
  2. Person 1 exerts effort level \frac{10\hat{L}}{1-\hat{L}} and person 2 exerts effort level 0.
  3. Person 1 exerts effort level \frac{1}{2}\frac{10\hat{L}}{1-\hat{L}} and person 2 exerts effort level \frac{1}{2}\frac{10\hat{L}}{1-\hat{L}}.

Notice how radically different these ways three ways of distributing the work of pursuing the group’s common objective are from the point of view of the group members. (a) puts the entire burden on person 1 and none of it on person 2. (b) does the opposite. (c) distributes the burden equally.

You can explore all of the possible distributions of the burden between persons 1 and 2 using the interactive chart below. The chart depicts the situation in which the sum of person 1’s and 2’s effort levels causes the outcome to occur with a particular likelihood level \hat{L} – i.e. e_1 + e_2 = \frac{\hat{L}}{1-\hat{L}}. It then allows you to use the slider at the top to pick the percentage of the total amount of effort required that person 1 exerts. The left-hand chart shows the resulting effort levels persons 1 and 2. The right hand chart shows their resulting utility levels.

Pause and complete check of understanding 3 now!

So what? Why does it matter that the burden of achieving any given likelihood of the favorable outcome can be distributed between group members in such a wide variety of ways?

Think about the choice of effort level e_i by an arbitrary person i in the group. Let E_{-i} stand for the sum of the effort levels of the group members other than person i. For instance, if i = 1, then E_{-i} = e_2 + e_3 + \cdots + e_N. Using that notation, we can write the likelihood L that the favorable outcome occurs as a function of group member i’s effort level e_i and the sum of the efforts E_{-i} of all the other persons as: L = \frac{e_i + E_{-i}}{e_i + E_{-i} + 10} Now recall that person i’s utility level if the likelihood of the favorable outcome is L and her effort level is e_i is 1000 \times L - e_i Substituting the former expression into the latter, we see that person i’s utility level can be written both as a function of her own effort level and the sum of the effort levels of the persons other than i: U(e_i, E_{-i}) = 1000 \times \frac{e_i + E_{-i}}{e_i + E_{-i} + 10} - e_i

This implies that no member of the group i can make a reasonable decision about what effort level to exert without first forming an expectation about how much effort her fellow group members will contribute to the cause. You can get a sense of this by playing with the interactive graph below. The graph displays the utility level of any given person i as a function of her effort e_i, taking the sum E_{-i} of all of other persons’ effort levels as given. The inputs to the left of the graph allow you to set E_{-i} and to choose whether the graph displays the level of effort that maximizes person \text{$i$'s} utility level, given the selected value of E_{-i}. Play with the graph in order to get a sense of how the effort level that results in the highest level of utlity for person i depends on the sum E_{-i} of the efforts of the other group members. Notice that as the effort levels of person i’s fellow group members decrease, she has an incentive to “step up” and contribute more effort to the group’s common objective.

With this in mind, imagine a person i who, for whatever reason, expects her fellow group members to exert very little effort. For instance, imagine she expects that they will exert no effort at all, so that E_{-i} = 0. This puts person i “on the hook” for taking on all of the work of pursing the group’s common interest. You can check using the graph above that if E_{-i} = 0, then the effort level that maximizes i’s utility is 90. Imagine person i contemplating putting out all that work, but then thinking…

Wait…if I exert effort level 90 and everyone else exerts level 0, our total effort will be 90. But why should all of the burden be on me??? We could achieve a total effort level of 90 in any number of ways! Someone besides me ought to step up and contribute! Maybe I should refuse to put out any effort unless at least a few other people in the group agree to chip in!

What this imaginary reflection demonstrates is that this model implies that group members have strong incentives to bargain, negotiate, and contest with one another over the distribution of the burden of advancing the group’s common interest. In fact, the total amount of effort put towards the common interest by the group as a whole – and thus the extent of the group’s political power relative to other interests – might be a result of such contestation between group members! Thus this model implies that the way a group navigates internal conflict over the distribution of the burdens of political action might be a determinant of a group’s external political power.

Wondering Not Knowing

Many of us think of science as a practice that above all else involves creating definitive knowledge about what the world consists of and how it works. We wrote at the outset of this lesson that PPT is a collection of tools built specifically for use in political science. So if PPT is used for political science, and science involves creating definitive knowledge about the world, does that mean that the models we build with PPT create knowledge of how the world really is? For instance, does our model of collective action tell us what will happen when a group of persons faces a collective action problem?

Let’s see. Consider once again Einstein’s, Glick’s and Palmer’s data recording comments made by persons at public hearings in the Boston metro area opposing or supporting permits for new housing construction. Recall again that each row of rectangles in the chart below corresponds to one public hearing on one construction project, that each rectangle in each row is a comment on that project, and that each red rectangle to the right of the thick vertical line is a comment opposed to construction and each green rectangle to the left of the line is a comment supporting construction.

This data displays the outcomes of many different collective action problems. At each hearing (i.e. each row in the chart) two collective action problems took place: In one, each person in the group of persons who approved of the construction project under consideration chose whether to put the effort in to show up at the hearing and plead their case. In the other, each person in the group of persons who opposed the construction project under consideration chose whether to put in the effort to show up at the hearing and plead their case. Each rectangle, then, is one member of one of these two groups who put in effort on behalf of their groups’ cause.

So if our model of collective action problem gives us knowledge about what happens in collective action problems and why, we have a real test of its power right here! By virtue of our analysis of the model, we should now know who turned out to speak at each of those public hearings and why.

Has our model of collective action created that knowledge?

Absolutely not!!!!!!!

Remember the basic fact about politics that we recognized by “thinking small” at the beginning of this lesson: Political outcomes result from the aggregation of the actions of many, many persons. Persons vary immensely in their circumstances. The story behind the actions taken by each person represented by a rectangle in the chart above is unique. Everyone showed up (or didn’t!) for a distinct set of reasons, and through a distinct combination of circumstances.

Because of the diversity and complexity underlying this and all other significant political outcomes, we know without any doubt that our model of collective action is incorrect as a description of the process determining who showed up to speak at those hearings, who stayed home, and why. Think back to the model and remember that it depicts a set of persons who are in every way identical to one another. In effect, the model allows us to explore what happens in the completely imaginary world in which a group consisting of multiple copies of one person all relate to a collective action problem in exactly the same way. No such world has, does or ever will exist. More generally, recall the point we made above that models are acts of imagination. Their implications carry information about our own thoughts, which may or may not accurately correspond to any actual politics.

Yet PPT models are tools for political science, and science is supposed to generate knowledge. How do we square that circle?

It’s true that sciences like political science try to produce definitive knowledge about the world. But that’s only half the story. Scientists can’t produce knowledge of anything without first wondering: Wondering what is happening, how it’s happening and why. Knowledge is the end product of a very long process, most of which involves discovering the extent of what one doesn’t know and wondering what the answers might be.

And that – wondering, not knowing – is what PPT models help us do.

Take, for instance, what we learned by examining our model of collective action: Despite sharing a common political objective, any group of persons in a collective action problem described by the model necessarily face an internal conflict over the distribution of the burdens of pursing the group’s common objective. Because the model is a model, it’s not a fully accurate description of any collective action problem that has ever occurred in real politics. Its implication about internal conflict merely makes us aware that such an internal conflict might be an important feature of any given real-world collective action problem. And so, having analyzed the model, when we encounter such a real-world problem, we wonder: Do the persons involved in the problem think or talk about how the burdens of action should be distributed between them? If so, what do they think and what do they say? When they become aware that choices must be or are being made about how those burdens are distributed, what do they do? How do they contest with one another over the distribution of the burdens of action and how is that conflict ultimately resolved? How do their ways of resolving that conflict affect the extent of their political power as a group?

Now those are questions worth pursuing by closely investigating the details of actual politics! The model helped us become aware of those questions and wonder about their answers.

If you choose to continue these lessons, this is the nature of the work you’ll be doing. PPT models are tools for political science, and more specifically not for knowing, but wondering. PPT models help us discover the implication of our ideas about politics, and by doing so help us identify questions we need to ask to fully understand politics.

Pause and complete check of understanding 4 now!

References

Clarke, Kevin A, and David M Primo. 2012. A Model Discipline : Political Science and the Logic of Representations. New York: Oxford University Press.
Einstein, Katherine Levine, David M Glick, and Maxwell Palmer. 2019a. Neighborhood Defenders: Participatory Politics and America’s Housing Crisis. Cambridge University Press.
Einstein, Katherine Levine, Maxwell Palmer, and David M. Glick. 2019b. “Who Participates in Local Government? Evidence from Meeting Minutes.” Perspectives on Politics 17 (1): 28–46.
Glaeser, Edward, and Joseph Gyourko. 2018. “The Economic Implications of Housing Supply.” Journal of Economic Perspectives 32 (1): 3–30.
Olson, Mancur. 2009. The Logic of Collective Action: Public Goods and the Theory of Groups, Revised Edition. Harvard University Press.