Microeconomics I

University of Cape Town

Unit 4: Social Interactions

Content Notes
2023

written by Robert Hill

Table of Contents
Lesson 1: Introducing Game Theory and the Nash Equilibrium ....................................... 1
Lesson 2: Solving games and some common characteristics ............................................. 8
Lesson 3: Preferences as Indifference Curves and Altruism ..............................................16
Lesson 4: Public goods games and experimental economics ........................................... 24
Lesson 5: Ultimatum games and sequential-move games ................................................31
Lesson 6: Expected payoffs and competition in the ultimatum game ............................. 37
 Lesson 1: Introducing Game Theory and the Nash Equilibrium
1.1 Game theory in context

TheBefore
icewe caps are
begin with the melting!
economic theory, let me tell you a story about something different: a
story about climate change. Did you know that global carbon dioxide levels are higher than they
Globalhave been at any point
atmospheric in history?
carbon Looklevels
dioxide at the graph
are inhigher
Figure 1, below.
than The have
they carbonbeen
dioxideat level
any
point in history.
has really spiked in the last 70 years or so, and it has had dire consequences: the average
temperature of the Earth’s surface has increased by about 0.9°C since the 1800s1 and we are
Earth’scurrently
average in asurface temperature
situation where hasAntarctica
Greenland and risen byare0.9°C
losing since the late
approximately 4001800s;
billion tons
Greenland andyear.
of ice per Antarctica have lostof ice
That’s the equivalent aboutsheets of the
285 billion motororder of 400
cars’ worth billion
of weight tons per
in ice.
year.
The likelihood of climateFigure
change arising
1: Global from
carbon human
dioxide activity
levels over time is above 95% - i.e. we
are almost certainly to blame for this!

Source: [Link] 2

Source: [Link]

We know that humans are almost certainly to blame for this change in global climate patterns and
the Stern Review (a report written by Nicholas Stern in 2006) tells us that “[t]he scientific evidence
is overwhelming: climate change presents very serious global risks, and it demands an urgent global
response.” The Stern Review takes the position that the benefits of early action to slow climate
change will far outweigh the costs of neglecting the problem – in other words, by taking action
now, we may well be able to save the Earth and maybe repair the damage that has been done.

The recommended course of action is to try cut greenhouse gasses significantly, so that we can
slow climate change. Of course, if everyone just goes on driving their cars, using coal-based energy
sources, and manufacturing in the way we are, then nothing will happen – we need something
more; we need wholesale change to occur. Enter agreements like the Paris Agreement (as part of
the United Nations Framework Convention on Climate Change), which mandates countries to
commit to reducing their carbon emissions significantly in order to keep global temperatures from
rising to above 2°C above pre-industrial levels.

But what, you may ask, does climate change have to do with economics?

Well, the idea that underlies climate change – the idea that people make decisions about their
consumption (e.g. driving a car), but fail to take into account how this decision impacts on
someone else’s utility (e.g. through climate change) – is an idea that is important in our study of

1 This, of course, coincides with the onset of the industrial revolution.

1
economics. The idea that people can make decisions about their consumption of a good or service,
and not consider how this decision impacts on others can cause what we call a social dilemma.

One example of a social dilemma is a tragedy of the commons – a term coined by Garrett Hardin
in his 1968 paper called “The Tragedy of the Commons”. In this paper, Hardin explains how if a
resource is not owned by anyone in particular, then it is easily overexploited by society, when
people act in their selfish interest. The canonical example of this is to think of a field that is shared
amongst a group of farmers who want to graze their cows on the field. All farmers want to let
their cows graze for as long as possible to make their cattle as fat and healthy as possible. But if
everybody lets their cattle graze unchecked, then the field will be degraded, the grass won’t regrow
and the topsoil may be eroded through rainfall. In order to fix this problem, you need to control
access to the resource in some way, or make people accountable for the resource they use.

Other examples of the tragedy of the commons are prevalent everywhere: the atmosphere being
polluted is one since the atmosphere is a common resource and people acting in their own
immediate interest by driving their cars, is polluting and degrading the air around us. Traffic jams
are another example: Freeways are often the most direct routes in and out of a city, and people
would use them to get to work because they would be more convenient than taking back routes
and spending extra money on fuel because you took a longer route to work. But everyone thinks
this, and ends up clogging the freeways with traffic to the extent that they are a nightmare to use
for all involved. Even your brother, sister or roommate leaving dirty dishes in the sink is an
example of the tragedy of the commons! Think of your home as a common resource, and by
leaving dirty dishes in the sink, you would save yourself some time because you don’t have to wash
the dishes yourself … but you are simultaneously degrading the atmosphere of cleanliness around
your house.

Of course, in each of these cases, we want to solve the problem. Maybe you let your cows graze
on your own field, or you take a back road to work to ease congestion; maybe you decide to
convert your home to solar power, or maybe you wash your roommate’s dishes. Invariably, there
will be someone who benefits from this – someone whose cows will be able to graze longer, or
someone whose dishes get washed through no effort on their part. This person who gets the
benefit without putting in the work is called the free rider.2

So, in this unit of the course, we ask ourselves how to solve this social dilemma or tragedy of the
commons. It is a problem that has been faced for thousands of years – Aesop (of Tortoise and the
Hare fame) wrote a fable about mice who had to tie a bell around a sleeping cat’s neck 2500 years
ago. One mouse had to take the risk to solve the dilemma of the mice turning into lunchmeat, but
the other mice benefitted from this without lifting a finger. Maybe the mouse who ties the bell
around the cat’s neck has altruistic preferences – i.e. preferences that mean he or she cares for
others and may be willing to sacrifice their happiness for others.

There are, of course, modern day examples of society trying to solve social dilemmas, such as the
Tribunas de las Aguas – a water court in Spain, where farmers report their peers who use more than
their fair share of water to a tribunal; or the Montreal Protocol, which systematically banned CFCs3
from production; or the City of Cape Town trying to avert Day Zero during the drought. More
recent – and relatable – examples of social dilemmas can be found during the COVID-19

2 You may be familiar with free riders from your own academic work. Have you ever had a group project where
someone just didn’t pull their weight, but at the end of the project, they still got their name on the report, and the
same mark as you? Well, that person was a free rider – they did nothing, but still got the benefit of someone else’s
work.
3
Full name: chlorofluorocarbons (quite a mouthful!)

2
pandemic: decisions made by individuals regarding wearing face masks, isolating after testing
positive for the virus, or getting vaccinated could all be considered social dilemmas.

Unit 4 starts to analyse these kinds of situations: where people make decisions that may have an
impact on other people. We start to investigate how people may make decisions interactively, and
how they may be able to work together to avert social dilemmas. The tools we use for this are part
of something called game theory, and this is what we will be studying in this unit.

1.2 What makes a game?

Let’s start off by asking a simple question: which side of the road do you drive on? The majority
of you probably answered “the left” because you are in South Africa, and that’s what we’re told to
do. But what if you were in the U.S.? You would have answered “the right”, because that’s what
you do there. Why did you answer that question the way you did? Well, maybe there are laws telling
you that’s the way it has to be – sure. But where did those laws come from? At one point there
wasn’t a law dictating which side to drive on. Maybe back in the day, it was simply a case of riding
your horse and carriage where you liked. Maybe we just gravitated towards these rules because of
something more primal: the instinct of “don’t hit the guy coming at you”.

Once you understand that people in the olden days may have simply chosen to drive on the side
of the road that other people drove on – i.e. if everyone else is on the left, then I had better go to
the left too – then you understand the basic concept of how decisions made by people may be
interactive, and you understand the notion of a strategic interaction. A strategic interaction is
one where people making a decision are aware that their choices have an impact on someone else,
and vice versa.

A game is then the model that we use to analyse these strategic interactions, and these games are
made up of a number of different components. These components are as follows:

1. The players – i.e. who is playing the game

2. The strategies – i.e. what complete plans of action are available to the players throughout
the course of the game
3. The information – i.e. what each player knows when they make their decisions (this boils
down to asking whether the game is sequential or simultaneous in nature)
4. The payoffs – i.e. the utility the player gets from a particular combination of their own
and the other player’s actions.

Once you have identified all of the above components, you should be able to put a game together
for analysis.

Box 1: Distinguishing strategies from moves

Although it is not necessarily clear why at this point, remember that a strategy is defined as a
complete plan of action for the game, while a move/action is the next action taken by a
player at that point in time.

An intuitive way to think about this is if you think about a game of chess: your move/action is
the way you move your piece in your turn – i.e. you have moved your queen forward three
spaces. Your strategy, on the other hand, is the full plan of all your moves from the start of the
chess game to the end – i.e. move 4th pawn forward two spaces, move right bishop out three

3
 squares, move 6th pawn out 1 square, etc. all the way to the checkmate or stale mate that ends
the game.

This distinction blurs in the case of a simultaneous move game (since in simultaneous move
games, you can often think of a strategy as your action and vice versa), but it is very important
to understand this difference when we study sequential move games later in the course and in
second year game theory courses.

Box 2: Sequential and simultaneous move games

When asking yourself about the type of information available to the players in a game, you are
essentially asking yourself whether the players are involved in a simultaneous move game, or
a sequential move game.

Remember, a simultaneous move game is one where the two players are moving with the
same amount of information. This means that they are making their decisions with no
knowledge of what their opponent will do. They may believe their opponent will act in a certain
way, but there is no way of knowing for sure until after the game has been played. This does
not mean that players have to move at the same time, although this is one interpretation of
this type of game. A standard game of Rock, Paper, Scissors would fit this definition.

Simultaneous move games can actually have asynchronous moves: for example, if we play a
game of Rock, Paper, Scissors, I can send Player 2 outside. While Player 2 is outside, I ask Player
1 what their move will be, and I note this down. After doing so, I call Player 2 in from outside
and ask what their move is and note this down. I can then determine who won the game, or
whether there was a draw. Even though the moves were asynchronous (i.e. did not happen at
the same time), the game is still simultaneous because Player 2 made their decision of what they
would do without knowing what Player 1 did, and (obviously) Player 1 made their decision of
what to do without knowing what Player 2 was going to do. This satisfies the criteria of people
moving with the same amount of information.

On the other hand, a sequential move game is one where players update their information in
between moves. In other words, one player makes a move, the other player observes this and
reacts accordingly. For example, in a game of noughts and crosses, if I place my O in the centre
square, you observe this and where you put your X will be influenced by my move – you are
reacting to the information of the placement of my O. Similarly, in a game of chess, you would
observe my move, and base your next move on the updated positions of my pieces – i.e. the
new information you have received.

1.3 Representing a basic game

In order to understand how to use game theory effectively, let’s start with a very basic scenario
that we want to examine using game theory. Here is a story for you to read:

Story time

Imagine that there are two farmers, named Amandla and Betty. Each of
them has to choose what crop they are going to grow and take to market.
The choices available to them are maize and sugar cane. Now, Amandla
and Betty both know how to grow both crops, but they can only grow one

4
 at a time. Furthermore, due to the nutrients in the soil, Amandla’s land is
better-suited to growing sugar cane, while Betty’s land is better-suited to
growing maize. Amandla and Betty cannot talk to one another (maybe
WhatsApp is down again), so they must make the decision of what crop to
grow independently. So, what should they each grow?

Your initial gut reaction may be that it makes sense for Amandla to grow sugar cane and Betty to
grow maize. That may well be the case (and we will work towards proving that this is what they
should do using game theory), but how do we prove that? Would the answer be so clear-cut if we
knew that the price of maize was particularly high this year, or that South Africans were desperate
for sugar cane this season? Maybe not.

So, let’s use game theory to try and model this situation in such a way that we can solve it
methodically. To do this, we make some simplifying assumptions4:

• There are no other people involved in any way; only Amandla and Betty are affected by
this decision.
• The decision of what crop to grow is the only decision that has to be made. There is no
need to decide on the fertilizer, or whether to hire extra labour, or anything like that. Only
the crop.
• Amandla and Betty only interact once. (We call this a one-shot game because there is
only one shot at playing the game. We will talk about repeated games later in the chapter)
• Amandla and Betty decide simultaneously what they will grow (in other words, they decide
without the other person knowing what they will do – remember “simultaneous” in game
theory doesn’t mean “same time” … it means “same information”!)

So, to start representing a game, we need to identify the first three components of the game: the
players and the strategies. Let’s get to it then.

Who are the players in this game? Well, the only two people who are involved in making
decisions are Amandla and Betty, so they must be the players. If you don’t understand this, ask
yourself “Who are the people making decisions in this scenario?” The answer will give you your
players.

What are the strategies available to the players? Remember strategies have to plan out the
players’ actions from the start of the interaction to the end of it. Since this is a simultaneous-move
game, a single decision will take the player from the start of the game to the end of it. But what
options are available to the players making the decisions? Well, in this case, it would be the options
of the crops available for the players to grow. So, Amandla can choose maize or sugar cane, and
Betty can choose maize or sugar cane. This means that we can write Amandla’s strategies as
“Maize” or “Sugar Cane” because these are the options available to him when he makes his
decision. Similarly for Betty.

What type of game is being played – simultaneous or sequential? This question deals with
the information aspect of the game, and hopefully you can see from the story that the game here

4 Now, you may not like these assumptions. That’s fair enough – they aren’t very realistic for the most part, but bear

with me. We are just starting our journey through this content, and it’s always easier to start simple and build up from
a solid foundation. Making these assumptions will make the analysis simpler to begin with. This means we can really
nail the basics, and when you understand those, we can start relaxing the assumptions (which we actually will start
doing later in the chapter).

5
is a simultaneous one. Simultaneous-move games are represented differently from sequential-
move games. We will start off looking at simultaneous-move games (which are represented in
game tables) as opposed to sequential-move games (which we will look at in a later lesson).

Now, with tall his information, we can put together a table that looks like this:

Betty
Maize Sugar Cane
Maize
Amandla
Sugar Cane

In this table (and we know it’s a table by virtue of the fact that we have a simultaneous-move
game), we have what we call a row player – the player whose information is displayed along the
rows of the table (in this case, that is Amandla) – and a column player – the player whose
information is displayed along the columns (in this case, that is Betty). We also have a row for
each of Amandla’s strategies (because he is the row player). Note that there is a row labelled
“Maize” and a row labelled “Sugar Cane”. This tells us that Amandla can choose to plant either
Maize or Sugar Cane. Similarly, we have a column for each of Betty’s strategies (because she is
the column player). One column labelled “Maize”, one column labelled “Sugar Cane”, indicating
that these are Betty’s available choices.

As a very important aside: We have a very specific naming convention that applies to cells in a
game table. We always label them as (row player strategy, column player strategy). So, look at
the cell in the top-right corner of the table – in the row where Amandla plants Maize, and Betty
plants Sugar Cane. This cell would be labelled (Maize, Sugar Cane) because to get to it, Amandla
(the row player) must play his strategy of Maize, and Betty (the column player) must play her
strategy of Sugar Cane. Note that the order is important! Writing (Sugar Cane, Maize) refers to the
bottom-left cell of the table, because we are now saying that Amandla must play Sugar Cane and
Betty must play Maize. Make sure you are comfortable identifying cells because getting the
name of a cell wrong in a test situation will lose you marks!

Now onto the last aspect of a game – payoffs. Essentially, a payoff is a representation of the utility
that an individual gets in a given situation. We use payoffs in game tables because writing out
whole stories in the table is just not practical. Payoffs represent the utility gained by an individual
from the actions of all players in a given situation. For example – if Amandla plants Maize and
Betty plants Sugar Cane, then Amandla will get a certain amount of utility, call it X; and Betty will
get a certain amount of utility, call it Y. The number X is what we call Amandla’s payoff in this
situation, while Y is Betty’s payoff.

Payoffs can be what is known as ordinal – in other words, only the order of the payoffs matters,
but not the size of them – or cardinal – where the order and the size of the payoff matters. For
example, think of ice cream flavours. If you like chocolate ice cream more than vanilla ice cream,
then we could say that you get a payoff of 10 from chocolate ice cream, and a payoff of 5 from
vanilla ice cream. If these payoffs are ordinal, then all that matters is that the payoff for chocolate
ice cream is higher than that for vanilla – in other words, a payoff for chocolate ice cream of 1
000, and a payoff for vanilla of -23.6 would actually be equivalent to 10 and 5 IF THE PAYOFFS
WERE ORDINAL.

On the other hand, if payoffs are cardinal then the order and size matter. Think about money –
you know that if you have two payoffs – say R10 and R5 – that the R10 is better than the R5 and

6
you can say that the R10 is twice as good as the R5. This ability to compare the size of the payoffs
with meaning makes the payoffs cardinal.

In games like we are examining now, we generally let the payoffs be ordinal – all that matters is
that we can rank the outcomes. Whether we represent the outcomes with payoffs 4, 3, 2 and 1, or
1000, 90, 0.6, -5, all that matters is that we know that the best outcome must give the highest
payoff.

Once you know the payoffs for each situation in a game – i.e. you know what utility Amandla and
Betty get in each scenario of the game we drew out above – then we can fill them into the game
table. To do this, there are some rules again: the row player’s payoff (in this case, Amandla’s)
ALWAYS goes first in a cell, followed by the column player’s payoff. So, imagine I have magicked
up the payoffs from Amandla and Betty’s utility functions for now and put them into a game table
like this:

Betty
Maize Sugar Cane
Maize 1, 3 2, 2
Amandla
Sugar Cane 4, 4 3, 1

I have colour coded the payoffs for your convenience here – normally they are just written in the
same colour. Amandla is the row player (and I have highlighted him in blue). As a result, in every
cell, Amandla’s payoff is put down first – look at the blue numbers in each cell. These blue
numbers are the first number in each cell, and so they correspond to the payoff Amandla receives
in each scenario. Similarly, Betty, as the column player, gets her payoff written second in the cell.
I have highlighted Betty’s payoffs in red.

Just to make sure that you can understand what is happening: imagine I ask you to locate
Amandla’s payoff when he plants maize and Betty plants maize as well. Where do you look? Well,
Amandla planted maize, so go to the row that says “Maize”. Betty planted maize too, so go to the
column that says “Maize”. Find where this row and column intersect – it is the top-left cell
(remember that we call this cell (Maize, Maize)?). Now, you were asked for Amandla’s payoff in
this scenario. Well, Amandla is the row player, so his payoff is first. This means that Amandla’s
payoff is 1.

Now, what is Betty’s payoff when Amandla plants sugar cane and Betty plants maize? See if you
can work out why the answer is 4.

As a quick aside – you may notice that the notation I show you here is slightly different to the
main CORE text. Although the two are absolutely equivalent in terms of the information shown,
the notation I show you (where the payoffs are alongside one another, instead of in coloured
triangles) is more commonplace and is the more normal way of representing games. You will see
the notation I introduce you to in future courses in game theory offered at UCT as well as in many
international textbooks and journal articles, so I figure why not show you the standard notation
from the beginning?

Now you have worked through how to represent a basic game, but we have yet to solve the game.
That will be the focus of the next lesson. It is crucial that you understand this lesson thoroughly
before moving on, so please take time to look over anything you are confused about before you
move on to Lesson 2.

7
Lesson 2: Solving games and some common characteristics
2.1 The Nash Equilibrium and how to solve for it

For now, we have only studied simultaneous move games (and relatively simple ones at that),
however, we will introduce sequential move games, and variations on simultaneous move games
as we progress through this chapter.

The process of solving a game requires the identification of a Nash Equilibrium, which we
understand to be a “mutual best response” by players in a game. This concept was popularized by
a man named John Nash5 who proved that every game has an equilibrium in either pure or mixed
strategies. Now, we only care about pure strategies at the moment (mixed strategies is more
complicated, where you start allowing people to choose certain strategies with various probabilities
… something for second-year ecos!) Technically, there are many definitions, but this one works
well in our context. A more detailed definition of a Nash Equilibrium is:

“A Nash Equilibrium is the combination of strategies in a game such that neither player
has an incentive to change their strategies, given the strategy choice of their opponent.”

In order to identify a Nash Equilibrium, we use what is known as the Best Response method of
analysis. Given the following game table which represents Betty and Amandla’s interaction to
choose which crop to grow, we can identify the Nash Equilibrium as follows.

Betty
Maize Sugar Cane
Maize 1, 3 2, 2
Amandla
Sugar Cane 4, 4 3, 1

The Best Response method of analysis always gets us to ask the question “If Player 1 did X, what
would Player 2’s best response be?” For example, we could ask “If Amandla planted Maize, what
would Betty’s best response be?” We use the payoffs in the game table to answer this question.
Recall that the first payoff in a cell represents the payoff to the row player (in this case, Amandla)
and the second payoff represents the payoff to the column player (in this case, Betty). So, if
Amandla is planting Maize, then we are looking at the top row of the game table. We then ask
whether Betty would prefer to plant Maize and get a payoff of 3, or whether she would prefer to
plant Sugar Cane and get a payoff of 2. We can see that 3 is greater than 2, meaning that Betty gets
greater utility from planting Maize, and thus that is what she will do. We indicate this decision on
Betty’s part by circling/highlighting/underlining Betty’s payoff corresponding to her best
response.

This question should be asked iteratively until you have gone through all possible strategies for
both players, and highlighted all the best responses for both players. So, let’s do that:

If Amandla played Sugar Cane, what would Betty do? To answer: look at the row where
Amandla plays Sugar Cane (the bottom row). Would Betty prefer to plant Maize and get a payoff
of 4, or plant Sugar Cane and get a payoff of 1? She would prefer to plant Maize to get a payoff of
4, so we highlight (in pink) Betty’s best response (which is the payoff of 4 in the bottom-left cell).

5 You may know about him from the film “A Beautiful Mind”. Yup, that’s him!

8
Now that we know what Betty would do for each of Amandla’s possible choices, we swap roles
and ask what Amandla would do for each of Betty’s possible choices.

So, if Betty played Maize, what would Amandla do? To answer: look at the column where
Betty plays Maize. Would Amandla prefer to plant Maize and get a payoff of 1, or plant Sugar
Cane and get a payoff of 4? He would prefer to plant Sugar Cane and get a payoff of 4, so we
highlight (in green) Amandla’s best response (which is the payoff of 4 in the bottom-left cell).

Lastly, if Betty played Sugar Cane, what would Amandla do? To answer: look at the column
where Betty plays Sugar Cane. Would Amandla prefer to plant Maize and get a payoff of 2, or
plant Sugar Cane and get a payoff of 3? He would prefer to plant Sugar Cane and get a payoff of
3, so we highlight (in green) Amandla’s best response (which is the payoff of 3 in the bottom-left
cell).

This would look something like this in our example:

Betty
Maize Sugar Cane
Maize 1, 3 2, 2
Amandla
Sugar Cane 4, 4 3, 1

Based on our definition of a Nash Equilibrium being a “mutual best response”, we can then
conclude that the Nash Equilibrium must be the situation where both players are best responding
– i.e. the cell where both players’ payoffs are highlighted (remember a highlighted payoff means it
is a player’s best response in the given situation, so two highlighted payoffs must mean both players
have a best response in that cell). In this case, that cell is the one corresponding to where Amandla
plants Sugar Cane and Betty plants Maize. This cell is denoted (Sugar Cane, Maize) – remember
that a cell is always labelled as (row player strategy, column player strategy).

Now, since the Nash Equilibrium is the set of strategies that result in a mutual best response, we
denote the Nash Equilibrium as:

NE = (Sugar Cane, Maize)

The payoffs received in the Nash Equilibrium are called the Nash Equilibrium Outcome (or NEO,
for short). Thus, if you are asked for the Nash Equilibrium Outcome, you would denote it as:

NEO = (4, 4)

IMPORTANT NOTE: Please, please, please make careful note of the difference between a Nash
Equilibrium and a Nash Equilibrium Outcome! Remember that the Nash Equilibrium will look
like you are naming a cell – i.e. you have something like (row player strategy, column player
strategy) – while a Nash Equilibrium Outcome talks about the payoffs in the cell that we found
to be the Nash Equilibrium. Getting these two concepts mixed up WILL mean you get marked
wrong, so make 100% sure you are comfortable with which is which.

9
 Box 3: The case of equal payoffs

It isn’t always the case that when solving for a Nash Equilibrium, one strategy always gives a
player a strictly higher payoff than the alternative. For example, what if Betty and Amandla’s
game looked slightly different …?

Betty
Maize Sugar Cane
Maize 1, 3 2, 3
Amandla
Sugar Cane 4, 4 3, 1

Well, now when we ask what Betty would do if Amandla planted Maize, the answer isn’t as clear:
If Amandla plants Maize, Betty gets a payoff of 3 from planting Maize, and a payoff of 3 from
planting Sugar Cane. In other words, she is indifferent between her two options. So, how do
you use Best Response Analysis here? Which of the two options is Betty’s best response?

Actually, both strategies in this case become Betty’s best response under the situation when
Amandla plants Maize. This means you should circle/highlight/underline both of Betty’s
payoffs in this scenario, as both strategies are technically her best response (as they both give
her the highest payoff she can achieve under this scenario). This would lead to your best
response analysis looking something like this:

Betty
Maize Sugar Cane
Maize 1, 3 2, 3
Amandla
Sugar Cane 4, 4 3, 1

Note that the way of identifying your Nash Equilibrium has not changed – it is still the cell with
all the payoffs highlighted. In other words, NE = (Sugar Cane, Maize).

2.2 Identifying important concepts in a simple game

Now that you understand how to draw up a simple game and how to solve for the Nash
Equilibrium, we zoom out a little bit and look at some characteristics of games that we may use to
describe different types of games. This section is easiest to explain in chunks, so each box below
will describe a concept that is important for you to get comfortable with.

Box 4: Dominant strategies

Sometimes, in a game, one particular strategy will always provide a player with their best
response. If this is the case, we say that the player has a dominant strategy. We focus on the
case of strictly dominant strategies, which are defined as follows:

Strategy X is a strictly dominant strategy for Player A if it provides a strictly greater payoff
than all of Player A’s other strategies, no matter what Player A’s opponent does.

Considering the game between Betty and Amandla again, let’s ask ourselves whether Betty has
a dominant strategy.

10
 Betty
Maize Sugar Cane
Maize 1, 3 2, 2
Amandla
Sugar Cane 4, 4 3, 1

To do this, we ask ourselves whether there is one strategy which always gives Betty her best
response, irrespective of what Amandla is doing, and the answer is yes – notice that no matter
what Amandla does, Betty’s best response is always to plant Maize (if you can’t see this, ask
yourself what Betty would do if Amandla planted Maize, and then ask yourself what Betty would
do if Amandla planted Sugar Cane).

Since Betty has a strategy that is always her best response, we can say that Betty has a strictly
dominant strategy of Maize. The strategy that always provides the lower payoff in every case
– in this case, Sugar Cane – is then termed a strictly dominated strategy.

In exactly the same way, see if you can argue that Amandla has a strictly dominant strategy
of Sugar Cane, and that his strictly dominated strategy is Maize.

Side note: Note the use of the phrase “strictly greater payoff” when we defined a strictly
dominant strategy. If strategy X provides payoffs that are greater than or equal to the payoffs
from all other strategies, then we say that X is a weakly dominant strategy. The use of the
term “weakly” as opposed to “strictly” when describing dominance describes whether we allow
payoffs to be equal or not. In this course, we care more about strictly dominant strategies,
but if you want an easy way to remember these rules, think about the inequality signs: strict
inequalities describe strict dominance (i.e. > or <), while weak inequalities describe weak
dominance (i.e. ≥ or ≤).

Box 5: Dominant strategy equilibria

This is a very short diversion to discuss the concept of a dominant strategy equilibrium. We
know how to find a Nash Equilibrium in a game using the best response method (we discussed
this earlier). However, there is another way to find the Nash Equilibrium: if both players have a
strictly dominant strategy, then we know both players have a strategy they will always play, no
matter what the other person does. This means that the only place where the players’ best
responses will overlap and result in an equilibrium is when both players play their dominant
strategies – thus the Nash Equilibrium occurs when players play their dominant strategies.

When this situation occurs, we have what is known as a dominant strategy equilibrium – i.e.
a Nash Equilibrium where players are both playing their dominant strategies.

In our game between Amandla and Betty, the Nash Equilibrium (Sugar Cane, Maize) is actually
a dominant strategy equilibrium, because the Nash Equilibrium occurs when Amandla plays his
dominant strategy (Sugar Cane) and Betty plays her dominant strategy (Maize).

Betty
Maize Sugar Cane
Maize 1, 3 2, 2
Amandla
Sugar Cane 4, 4 3, 1

11
Nash Equilibria do not always have to be dominant strategy equilibria. For example, look at the
game below where Carl and Dumi are making the same decision as Amandla and Betty had to
make:

Dumi
Maize Sugar Cane
Maize 1, 3 2, 4
Carl
Sugar Cane 4, 4 3, 1

Notice that although Carl has a dominant strategy of Sugar Cane, Dumi does not have a
dominant strategy: when Carl plays Maize, Dumi will choose to play Sugar Cane; but when Carl
plays Sugar Cane, Dumi will choose to play Maize – there is no one strategy that always gives
Dumi a best response. The Nash Equilibrium in this case is still (Sugar Cane, Maize) but in this
case, it is not a dominant strategy equilibrium, because it isn’t achieved by all players playing
their dominant strategies.

Box 6: Socially optimal outcomes

In a game, you may be tasked with determining which outcome is most beneficial to society.
This will require you to find the socially optimal outcome in a game table. In order to do this,
we rely on a couple of assumptions: firstly, we assume that society is comprised of only the
people involved in playing the game at hand; and secondly, we assume that the payoff to society
is simply the sum of payoffs to all members of society.

So, to find the socially optimal outcome in a game like the one between Amandla and Betty, you
would have to add up the payoffs in each cell to work out what the total payoff to society is.
This sum is shown by the red value in each cell below. For example, in the cell (Maize, Maize),
we see that 1 + 3 = 4, giving the red 4 in brackets.

Betty
Maize Sugar Cane
Maize 1, 3 (4) 2, 2 (4)
Amandla
Sugar Cane 4, 4 (8) 3, 1 (4)

The socially optimal cell would then be the one where the sum of payoffs to Amandla and Betty
is highest – i.e. in the cell (Sugar Cane, Maize). There may also be multiple socially optimal cells
if more than one cell gives the highest possible payoff, such as in the driving game below:

Me
Left Right
Left 2, 2 (4) 0, 0 (0)
You
Right 0, 0 (0) 2, 2 (4)

In this case, there are two socially optimal outcomes, both of which give society the highest
possible payoff of 4. These cells are (Left, Left) and (Right, Right).

12
 2.3 Games with multiple equilibria

It is not always the case that there is only one Nash Equilibrium in a game. For example, think
about a game where people have to choose which side of the road to drive on. In some countries
around the world, like South Africa, people drive on the left-hand side of the road; in other
countries, like the USA, people drive on the right-hand side of the road.

In both of these cases, people are happy to continue driving on the current side of the road, given
what everybody else is doing: if everyone around you is driving on the right, you’re happy to drive
on the right. This means that everyone driving on the right-hand side of the road is a Nash
Equilibrium. Similarly, though, if everyone is driving on the left, then you have no reason to change
away from driving on the left, making this another Nash Equilibrium.

In our framework of a game table, we could put the game together as follows:

Me
Left Right
Left 2, 2 0, 0
You
Right 0, 0 2, 2

We can use our technique of Best Response analysis to find the Nash Equilibrium of this game in
exactly the same way as we did before. If we do, we will see something like this:

Me
Left Right
Left 2, 2 0, 0
You
Right 0, 0 2, 2

Using the rule that a Nash Equilibrium is found in the cell where both payoffs are highlighted, we
can see that two cells meet our criteria for a Nash Equilibrium. Thus, these two cells are both
Nash Equilibria and we can report them as follows:

NE1 = (Left, Left) with NEO1 = (2, 2)

NE2 = (Right, Right) with NEO2 = (2, 2)

2.4 Classifying different types of games

There are many different types of games that can be modelled using the tenets of game theory,
but certain games fall into well-known patterns. These types of games have specific names and
characteristics that we should take note of so that we can study them in more detail later.

For the purposes of this course, there are a couple of important games that you should be able to
recognize. These are:

1. The invisible hand game

2. The prisoner’s dilemma
3. Assurance game
4. Battle of the Sexes

13
There are, of course, other types of games as well, but we will not be looking at them in this course
in great detail. Examples of these types of games are the pure coordination game or a chicken
game.

If you are looking to identify the types of games above, then you need to be comfortable with all
the concepts discussed in this lesson up to now. If you are comfortable, then we can move on to
describing each type of game.

2.4.1 The invisible hand game

The invisible hand game is named for the concept of the invisible hand introduced by Adam Smith
in his 1776 book The Wealth of Nations. It arises from the idea that individuals, acting in their own
self-interest, can end up in a socially optimal state of the world.

To identify an invisible hand game, you should look for the following characteristics:

• There is 1 Nash Equilibrium

• The Nash Equilibrium of the game is socially optimal

An example (see if you can identify the characteristics above in this game):

Me
Stay home Go out
Stay home 4, 4 4, 3
You
Go out 3, 4 1,1

2.4.2 The prisoner’s dilemma

The prisoner’s dilemma is a thought experiment which exemplifies the idea that when individuals
act in their best interest, they may end up in an equilibrium that is sub-optimal for society
as a whole. The analogy arises because of the idea that two prisoners who are in separate
interrogation rooms may choose to rat out their partner in order to get a lighter prison sentence
than they would have done by keeping quiet. As an aside: look at slides 31 & 32 of the Unit 4
lecture slides to see how we could model climate change as a prisoner’s dilemma!

To identify a prisoner’s dilemma, you should look for the following characteristics:

• There is 1 Nash Equilibrium

• Each player has a dominant strategy
• If both players were to play their dominated strategy instead, then the outcome that would
result is better than the Nash Equilibrium for both players.

An example (see if you can identify the characteristics above in this game):

Me
Cook at home Get take-aways
Cook at home 5, 5 2, 6
You
Get take-aways 6, 2 3, 3

14
2.4.3 Assurance game

The assurance game is an example of a game with multiple Nash Equilibria, but importantly here,
the players have aligned interests – i.e. they both have the same preference for which equilibrium
they want to see.

To identify an assurance game, you should look for the following characteristics:

• There are 2 Nash Equilibria (in pure strategies)

• In the equilibria, players are playing coordinated (i.e., the same) strategies
• Each player has a preferred equilibrium and the two players prefer the same equilibrium.

An example (see if you can identify the characteristics above in this game):

Me
Study in the morning Study in the evening
Study in the morning 4, 4 1, 2
You
Study in the evening 2, 1 2, 2

2.4.4 Battle of the Sexes

Much like an assurance game, a Battle of the Sexes game is one with multiple Nash Equilibria.
However, in this case, the two players have different interests – i.e. the players prefer different
Nash Equilibria. This gives rise to a conflict in choice over which equilibrium will arise. This
conflict in choice of equilibria is where the notion of the “Battle” comes from in the name of this
game.

To identify a Battle of the Sexes game, you should look for the following characteristics:

• There are 2 Nash Equilibria (in pure strategies)

• In the equilibria, players are playing coordinated (i.e., the same) strategies
• Each player has a preferred equilibrium, however, the two players prefer different
equilibria.

An example (see if you can identify the characteristics above in this game):

Me
Watch Netflix Play board games
Watch Netflix 3, 5 2, 0
You
Play board games 0, 2 5, 3

2.5 Multiple equilibria – which do we see in reality?

In the case of games with a single equilibrium, it is easy to rationalize that we will converge on the
Nash Equilibrium scenario in reality, but when there are multiple Nash Equilibria, it isn’t so easy
to work out which equilibrium we’ll see manifest in reality. As an illustrative example: in the driving
game, which side of the road should people drive on? If you could answer this question definitively,
then surely the whole world would drive on one side of the road, but yet somehow, we have
different countries on different sides of the road.

15
You may argue that in an assurance game, it would make sense to go for the equilibrium that both
people prefer (i.e. the socially optimal equilibrium) – and yes, this would make sense. But it doesn’t
mean that this is necessarily what we see in reality. For example, let’s think about an example to
do with keyboards: compare the QWERTY keyboard we all know to the Dvorak Simplified
Keyboard (DSK). Studies have shown that the DSK keyboard has been found to increase typing
accuracy and speed (by up to 30% over a QWERTY keyboard!), but yet we still use the QWERTY
keyboard for our typing needs.

Why is this the case? Well, in the case of the DSK keyboard, it’s because the person who invented
touch typing developed it on a QWERTY typewriter keyboard; secretaries then wanted to be
competitive in the job market, so they learnt to touch type to better their chances of getting a job,
but they necessarily learnt on a QWERTY keyboard. As a result, firms who wanted to hire these
more efficient secretaries had to invest in QWERTY-typewriters, and so it spiraled … all because
of a chance event.

This is often the case: when there is more than one Nash Equilibrium, it is unclear where we will
end up, because society can be influenced by chance events, historical accidents, and societal
norms. All of these factors can make a big difference in which Nash Equilibrium of a game we see
in reality, but it is impossible for us to predict ex ante (beforehand) which equilibrium will
manifest itself in reality.

Lesson 3: Preferences as Indifference Curves and Altruism

3.1 What is a social dilemma and how might you solve it?

In the previous lesson, we introduced the notion of the prisoner’s dilemma – a situation where
individuals are acting in their own selfish interest, but where these actions actually lead society to
a suboptimal outcome. We face many such interactions in our daily lives: climate change, group
projects, and more recently, the outbreak of COVID-19.

Early in the case of the COVID-19 outbreak, we were faced with the choice to socially distance
ourselves (which is hard to do and comes with a number of psychological costs), or to continue
behaving as normal (which comes with added risk of infection, but with minimal disruption to our
immediate day to day life). Imagine that by staying at home and self-isolating, an individual gets a
benefit of 10 units of utility because they remain healthy and have assisted in flattening the curve
of infections. If one person were to go out and behave as normal while the other practiced social
distancing, then the person practicing social distancing loses 6 units of utility as they feel isolated
and alone, knowing their peers are out living normally. The person who went out, on the other
hand, has very low risk of infection, and they can continue with their life as normal, so they
experience a benefit of 15. If everybody goes out as normal, and does nothing to try flatten the
curve, then the risk of infection is high and they only receive a payoff of 6. This scenario describes
a prisoner’s dilemma, which we put together as follows (as an exercise, try draw this table up
yourself from the information given – it will really test your understanding!):

Person 2
Socially distance Behave as normal
Socially distance 10, 10 4, 15
Person 1
Behave as normal 15, 4 6, 6

Game 1

16
Game 1 presents a very important issue: society is faced with a social dilemma (i.e. a situation
where individuals are acting without considering the effect their actions may have on other people,
and so are harming society as a result). How might we resolve the social dilemma, and find our
way to the socially optimal state of the world?

In this case, we have had the answer given to us – President Cyril Ramaphosa imposed a national
21-day lockdown starting on 27 March 2020. This effectively takes the choice of what to do away
from the individuals and changes the rules of the game through institutional intervention. But
would it have been possible for us to resolve this game without resorting to government
intervention?

To answer this question, think about behaviour before the lockdown began: were there some
people who were socially distancing, even though they were experiencing high psychological costs
from doing so? There definitely were some people socially distancing, although not everybody was
doing this. So, why were there some people behaving in this way? Why would individuals be willing
to choose to do something that is less preferable to them individually, but which could help
society? The answer to this lies in the concept of altruistic preferences, or other-regarding
preferences, which we will discuss further in the next section.

3.2 Social preferences – the case of altruism

There are a number of different ways that one could go about solving a social dilemma: one could
change the rules of the game through institutional intervention; one could allow for games to
be repeated rather than letting them be played one-shot; or one could appeal to people’s social
preferences. This lesson will focus on the introduction of social preferences. If an individual
exhibits social preferences, then we can say that they have an attitude where they don’t only care
for themselves – they also care about the wellbeing or utility of others in society.

Now, social preferences can be positive or negative: you could care about others and want to take
action that benefits them (such as charity donations; self-isolating and helping to flatten the curve;
or giving a gift to someone else), or you could want to take actions that harm others (taking revenge
on a cheating partner; or trolling people online). These negative social preferences could include
things like spite or envy, while positive social preferences include altruism.

We find that in reality, people actually do show tendencies towards social preferences. They aren’t
the robotic, infinitely rational consumers that always act to maximise their own utility with no
concern for others (this type of person is known as Homo economicus). These social preferences
appear in many different places and are often instilled in us as young children. For example, think
about what you would do if you found a lost wallet or cellphone in a public place. Would you hand
it to the lost and found kiosk? If so, then you are exhibiting a social preference, because the Homo
economicus would see the lost wallet/cellphone as an item of value that increases their overall wealth,
and they would keep it for themselves. Similarly, do you tip the waitron at a restaurant you haven’t
been to before? If so, think about why you do it. It can’t be because you have a reputation to
uphold or anything like that, since you have never been to the restaurant before. You might argue
that it’s to reward the waitron for good service, but if you didn’t tip the waitron, they couldn’t
change the way they served you. To the Homo economicus, tipping your waitron would be seen as
giving away money that you didn’t need to – you received your food, have eaten it, and now you
are leaving, so they see no reason to tip their waitron.

In reality, we definitely don’t see people behaving like Homo economicus all the time – and that’s
actually okay. Maybe we could use the fact that people care about one another to solve a social

17
dilemma. But before we can see how this would work, we need to develop a set of tools to analyse
social preferences more rigorously.

3.3 Representing preferences as indifference curves

In Unit 3, you were introduced to the concept of an indifference curve: a set of points which
described bundles, all of which would give you the same level of utility – i.e. a line showing all the
combinations of two goods that gave you the same level of utility/satisfaction. The shape of these
indifference curves also changed depending on your underlying preferences for the two goods in
question – for example, if you were choosing between olives and biscuits, and you don’t really like
olives but love biscuits, you would be willing to give up a lot of olives for a single biscuit. The fact
that you are willing to trade many olives for a single biscuit has an impact on your marginal rate of
substitution (MRS), and thus it must affect the slope (and shape) of the indifference curve.

It is this principle that we will be looking at here – that the shape of an indifference curve can be
used to infer an individual’s preferences. One of the important things that we have to assume in
this section of the work, though, is that payoffs should be thought of as if they are monetary.
We do this because it is generally accepted that people prefer more money to less, and giving some
of your own money away to another person would only happen if you cared about that person. In
other words, thinking about payoffs like money is an easy way to visualize the concepts we are
discussing.

Now, you may well be thinking that indifference curves are a graphical tool, and as such, we need
to be working with graphs if we want to use indifference curve analysis. You would be right – we
do need to introduce a graphical element to our analysis. Just like you did in Unit 3, we will plot
our indifference curves in 2D space, with a good of interest on each axis. In the context of this
unit, we are investigating social preferences, so we want to see how Player A’s utility is affected
by changes in Player B’s payoff (because if Player A cares about Player B, then changes in Player
B’s payoff will affect Player A’s utility).

In order to be able to do this kind of analysis, we will plot our graphs with Player A’s payoff on
the horizontal axis and Player B’s payoff on the vertical axis. Now, it doesn’t matter whose payoff
you put on which axis – the underlying concepts will remain the same. However, what is important
is that you have one player’s payoff on the horizontal axis (for our example, this will be Player A),
and one player’s payoff on the vertical axis (for our example, this will be Player B).

Now that we know what space we’re working in, let’s start by thinking about someone who is
completely selfish and only cares about themselves. What would their indifference curves look
like?

The key here is to recognize that the person cares only for themselves – in other words, if we are
thinking about Player A being selfish, it means that Player B’s payoff can change without it having
any effect on Player A’s payoff. This means that Player A only gets more utility when their payoff
increases, and they don’t care whether Player B has a payoff of 0, 5, 200, or any number in between.
This should start to ring bells for you: if Player A has a certain payoff, call it X, then they would
be indifferent between a bundle where they get X and Player B gets 0; a bundle where they get X
and Player B gets 5; a bundle where they get X and Player B gets 200; etc. If you try to visualize
this set of bundles in the space with Player A’s payoff on the horizontal axis and Player B’s payoff
on the vertical axis, you would hopefully realise that Player A’s indifference curves take the shape
of vertical lines, as you can see in Figure 2.

18
 Figure 2: Representing selfish preferences as indifference curves

Utility increases in this direction

Player B payoff
IC1 IC2

Player A payoff

Hopefully it is clear that in Figure 2, if Player B’s payoff increases while Player A’s payoff remains
the same, then Player A remains on the same indifference curve. Player A only moves to a higher
(further right) indifference curve when their own payoff increases. Take very careful note, though:
these indifference curves represent Player A’s preferences!

But what about the indifference curves you are familiar with – the downward-sloping, convex
curves that you learnt about in Unit 3? If we were to draw these downward-sloping indifference
curves in our context, they would look like Figure 3. But what do they mean in terms of Player A’s
preferences?

To answer this, we can think about the argument in two different ways: first, by considering a
movement from point X to point Y; and second, by considering a movement from point X to
point Z.

Let’s begin with a movement from point X to point Y: This movement occurs along the same
indifference curve IC1. This means that Player A is indifferent between point X and point Y. But
in Point Y, Player A is receiving a lower payoff than in point X, while player B is receiving a higher
payoff in point Y than in point X. If Player A is indifferent between these two points, it means
that even though they are sadder from losing some of their own payoff, there must be something
that offsets that sadness and brings them back to the same level of utility. The only thing this can
be is the increase in Player B’s payoff. This indicates that Player A must care about Player B’s
payoff, and as such, has some form of social preferences.

19
 Figure 3: Downward-sloping indifference curves

Player B payoff
Z

Y
IC2
X

IC1

Player A payoff

The argument centred around a movement from point X to point Z follows a similar logic,
although it may be easier to follow: If Player A moves from point X to point Z, then the only
thing that has changed is Player B’s payoff (note that Player A’s payoff is the same in point X and
point Z). However, we see that point Z is on a higher indifference curve, meaning that Player A is
getting higher utility. The only thing that changed between point X and point Z was Player B’s
payoff, which means that an increase in Player B’s payoff has increased Player A’s utility. This
must mean that Player A has some sort of social preference.

In both cases, the argument comes down to how a change in Player B’s payoffs has fed through
into a change in Player A’s utility. Because this feed-through is positive (i.e. when Player B gets an
increased payoff, Player A’s utility increases, or when Player B’s payoff decreases, Player A’s
utility decreases), this is an example of a positive social preference – i.e. an example of altruism.

Granted, this isn’t pure altruism (if that were the case, then Player A would only care about Player
B’s payoffs and not care about their own), we call the type of preferences exhibited by these
downward-sloping indifference curves somewhat altruistic preferences.

Box 7: The convexity of the indifference curve

Something very important to notice about our discussions about downward-sloping indifference
curves is that the existence of somewhat altruistic preferences does not have anything to do
with the convexity of the indifference curve – all that is required is that the indifference curve
is downward-sloping. (If you cannot see this, try redraw Figure 3, but with the indifference
curves as downward-sloping straight lines and see whether you can still make the arguments
that Player B’s payoff affects Player A’s utility).

The convexity of the indifference curve tells you about the MRS (because the slope of the IC is
the MRS). So, a curved indifference curve tells you about how willing someone is to substitute
the good on the y-axis for the good on the x-axis: In particular, it tells you that if Player A is
getting very little to start off with, they are less willing to give up their payoff than they would
be if they had a lot to start with. This makes sense intuitively: if you have a lot of money, you
are likely to be more willing to buy gifts for friends, or donate to charity; whereas if you have

20
 very little money, although you may still want to do these things, it would come at a much greater
sacrifice than before.

If the rate at which you are willing to exchange your payoff for the other player’s payoff remains
constant, however (i.e. you are just as willing and able to donate to charity when you are rich as
when you are poor), then your MRS is constant. This means that you have an indifference curve
with a constant slope – i.e. it will be a straight line. Note that a downward-sloping straight line
IC would still represent somewhat altruistic preferences; all that will change is the rate of
substitution between goods.

A couple of questions to think over:

1. If Player A had completely altruistic preferences (i.e. they cared only about Player B and
did not care for themselves), what shape would their indifference curves take on?
2. Assuming we still have Player A’s payoffs on the horizontal axis and Player B’s payoffs
on the vertical axis, what shape would Player B’s indifference curves take on if
a) Player B were completely selfish?
b) Player B were completely altruistic?

3.4 Finding the optimal distribution

Just like in Unit 3, there is often an external constraint that limits the payoffs that Player A and
Player B are able to obtain. For example, let’s think about a situation where Amandla wins R10
000 in the lottery, and now he has to decide how best to distribute this money between himself
and Betty. The fact that there is R10 000 to distribute means that Amandla is constrained in his
decisions: he cannot give himself R7 000 and give Betty R4 000 – this is too much! He is
constrained by what is feasible given the total sum of money available to him.

In Unit 3, you studied how individuals found the optimal bundle – this was done by finding the
point where the indifference curve is tangent to the feasible frontier. We will use this same logic,
but we will ask a slightly different question: which is the optimal bundle for Amandla, under
different types of preferences?

Let’s assume Amandla won R10 000 in the lottery and he is trying to decide how we would
distribute this between himself and Betty. If you were to draw the feasible frontier for this scenario,
you would end up with a downward-sloping straight line like the orange line in Error! Reference
source not found..

Now, if Amandla were completely selfish, we know intuitively that he would keep all the money
for himself. Does this correspond with the result we get from the graph? Well, we know that if
Amandla is completely selfish, his indifference curves are vertical lines (like the blue ICs in Error!
Reference source not found.). If we were to push this indifference curve out as far as we could
to the right, the final point at which it touches the feasible frontier is at point A, which corresponds
to Amandla having R10 000 for himself, and giving R0 to Betty.6

6Although point A is strictly not a point of tangency between an IC and the feasible frontier, it is still an optimal point
as it is the point along the feasible frontier that gives maximum utility. We would call this point a “corner solution”
to the problem.

21
 Figure 4: Determining Amandla's optimal split of R10 000

Amandla’s indifference Amandla’s indifference

curves (when curves (when
somewhat altruistic) completely selfish)
12
Feasible
payoffs
10
frontier
Betty’s payoff (thousands of Rands)

B
3

Feasible
payoffs set
A
0
0 7 10 12
Amandla’s payoff (thousands of Rands)

If, on the other hand, Amandla were to be somewhat altruistic, then his indifference curves would
look like the purple ones in Error! Reference source not found.. Notice then that the optimal
point for Amandla is no longer point A, where he keeps all the money for himself, but rather, it is
now at point B, where he gives R3 000 to Betty (of course, this is just an example – where the
numbers come from is a function of the underlying mathematics which you shouldn’t worry about
at this point). This is consistent with what intuition says: if Amandla cares a bit about Betty, he
may share his winnings with her, but he would still keep some of the money for himself.

3.5 Translating a game table into graphical form

Now that we understand how to represent preferences as indifference curves, we could try to use
these tools to analyse our social dilemmas. Let’s look back at Game 1:

Person 2
Socially distance Behave as normal
Socially distance 10, 10 4, 15
Person 1
Behave as normal 15, 4 6, 6

This is a prisoner’s dilemma and we want to be able to solve it somehow, without having to rely
on government intervention. So, we ask whether altruism could be enough to get us to the socially
optimal outcome of (Socially distance, Socially distance). In order to answer this question, we need
to translate the game from a table form into a graphical form.

This is relatively easily done: we start with a set of axes where Person 1’s payoff is on the horizontal
axis and Person 2’s payoff is on the vertical axis. Then we look at each individual outcome in the
table and consider this a point on the graph. So, for example, (Socially distance, Socially distance)

22
 has payoffs (10, 10). If you consider (10, 10) like a point on the Cartesian plane (because the first
coordinate is Player 1’s payoff, which lies on the horizontal axis; and the second coordinate is
Player 2’s payoff, which is on the vertical axis), then this is simply a case of plotting the point
(10,10) on our set of axes.

We can do this for all the cells in Game 1, and we end up with a graph like the one in Panel (a) of
Figure 5, where each point is representative of a particular cell in the game table:
Figure 5: The COVID-19 Prisoner's dilemma graphically
Player 2’s payoff

Player 2’s payoff

IC2

IC1
(Socially distance, Behave as normal ) (Socially distance, Behave as normal )
15 15
(Socially distance, Socially distance) (Socially distance, Socially distance)
10 10

6 (Behave as normal, Behave as normal)

6 (Behave as normal, Behave as normal)

(Behave as normal, Socially distance) (Behave as normal, Socially distance)

4 4

4 6 10 15 Player 1’s payoff 4 6 10 15 Player 1’s payoff

(a) (b)

If we assume then that Player 1 has somewhat altruistic preferences (and the payoffs are considered
as if they were monetary in nature), then we can insert indifference curves that model those
preferences and overlay them on top of the graphical representation of our game. These
indifference curves are depicted in Panel (b) of Figure 5. The Nash Equilibrium, (Behave as
normal, Behave as normal) lies on IC1, but the socially optimal outcome (Socially distance, Socially
distance) lies on IC2. So, by virtue of the fact that the socially optimal outcome lies on the higher
indifference curve, we can see that if people exhibit altruistic preferences, they may coordinate
towards the socially optimal outcome as this is their most preferred outcome.

As a comparison point, consider what happens when Player 1 is selfish, and acts only in their own
self-interest (i.e. has vertical indifference curves). We can see that their most preferred outcome
(the outcome on the highest indifference curve, which in this case is the IC furthest to the right)
would then be (Behave as normal, Socially distance) where they get 15 utils – the highest payoff
available in the game.

23
Lesson 4: Public goods games and experimental economics
4.1 Public goods and the problem of providing them

The prisoner’s dilemma, as we have been studying it, is one way of representing real-life problems
in a relatively simple way. The COVID-19 pandemic, pollution of the atmosphere, climate change
– they can all be modelled as a prisoner’s dilemma.

Another example of something that can be modelled as a prisoner’s dilemma is the provision of
a public good. Now, a public good is a special type of good that fulfills two important
characteristics:

1. A public good must be available for all people in society to use (we say it must be non-
excludable)
2. The use of a public good by one individual does not decrease its availability or usefulness
to others (we say that it is non-rival).

Examples of goods like this would be a community garden, streetlights, Keurboom park in
Claremont, the beach, Rondebosch common, etc. Things that are available for everyone to use,
and where the fact that I use it doesn’t prevent you from using it as well.

These characterstics are different from a private good, which is excludable (i.e. you can prevent
certain people from using it through mechanisms like the price), and rival (i.e. if you use it, then
someone else can’t). Think about a bag of crisps you buy for yourself. You had to have the money
to buy them (i.e. they were excludable to those without money), and you eating the crisps means
someone else can’t (i.e. they are rival – you eating the bag of crisps means that I can’t eat them).

Providing a public good doesn’t require every consumer that uses the good to pay for it: you don’t
necessarily pay anything towards the upkeep of the Rondebosch common, but you can use it to
go for a run any time you want to. So, some will bear the cost of provision, while others simply
enjoy the benefit – this raises the free-rider problem, and our dilemma now becomes how to
maintain the public good.

We can boil down the provision of public goods into a simplified form: a type of public goods
game. As an example, imagine that there are four farmers who all share a large field where they
can plant their crops. The field is serviced by some form of shared infrastructure (let’s say there is
a shared irrigation system). A farmer can choose to contribute towards the irrigation system – this
cost is R10 000 per farmer. For each farmer that does contribute, there is a benefit of R8 000 in
terms of crop value. Now, because the irrigation system is shared, even if only one farmer
contributes to it, all four farmers will benefit from it – this is because the irrigation system will also
water their plants, even though they may not have contributed to it directly.

Question time: Given this scenario, what would you do? Would you contribute to the irrigation,
or not?

If you think about this systematically and purely according to cost-benefit analysis, you probably
wouldn’t contribute. Why? Well, let’s imagine Kim is a farmer in the group and calculate her net
payoff from contributing and not contributing for all possible scenarios: when no other farmers
contribute; when 1 other farmer contributes; when 2 other farmers contribute; and when 3 other
farmers contribute.

24
Let’s start with the scenario where no other farmer is contributing. Kim has to decide whether to
contribute or not. Imagine that Kim does contribute: in this case, there would be a total of 1
farmer contributing (Kim), so there is R8 000 benefit for each farmer. Kim, though, was
contributing, so incurred the cost of R10 000 for contribution. This means the net payoff is:

𝑁𝑒𝑡 𝑝𝑎𝑦𝑜𝑓𝑓 = Benefit − Cost = 𝑅8 000 − 𝑅10 000 = −𝑅2 000

On the other hand, when Kim doesn’t contribute, then there are no farmers contributing, so
there is no benefit. Because Kim didn’t contribute, there is no cost to her, so her net payoff is:

𝑁𝑒𝑡 𝑝𝑎𝑦𝑜𝑓𝑓 = Benefit − Cost = R0 − R0 = R0

So, what would Kim do? Which action gives her the highest payoff? By comparing the payoff
from contributing (-R2 000) to the payoff from not contributing (R0), we can see that Kim would
do better by not contributing.

This logic can extend to any number of farmers contributing, and the choice Kim will make as a
result.

Figure 6 shows graphically the payoffs Kim would get under all the different numbers of
contributing farmers. Check to make sure you can calculate all the payoffs in the Figure. To help
you along, let’s try one more example: What is Kim’s payoff when she contributes, and two other
farmers do?

Well, if Kim contributes along with two other farmers, then there are 3 contributors in total. This
means that there is a benefit of R8 000 per farmer multiplied by the 3 contributing farmers, giving
a total benefit of R24 000. But, because Kim contributed, we know she incurs a cost of R10 000.
This means Kim’s payoff would be:

Net payoff = Benefit − Cost = (R8000 × 3 𝑓𝑎𝑟𝑚𝑒𝑟𝑠) − R10000 = R14000

Figure 6: Kim's payoffs in the simple public goods game

Kim’s payoff (thousands of Rands)

Source: CORE Text, 2020

25
Hopefully, you will notice that Kim’s payoff from not contributing is always higher than her payoff
from contributing, given a specific number of other contributors. This means that Kim has a
dominant strategy to not contribute. But the problem here is that every farmer will experience
the same thing as Kim and they will all also have a dominant strategy of Don’t Contribute. This
means everyone would prefer to free-ride on other farmer’s contributions, and we would expect
the Nash Equilibrium to be where none of the four farmers contributes to the public good.

This is precisely the social dilemma we want to avoid – we have an interaction amongst a number
of people, and they have a dominant strategy which leads to a socially suboptimal outcome (in fact
this model of a simple public goods game is very similar to a prisoner’s dilemma with multiple
players). Now, how might we solve the problem of public good provision?

4.2 Altruism and repeated interactions

In Lesson 2 of Unit 4, we discussed how altruism can be used to solve a prisoner’s dilemma. Now,
in theory, that would work in Kim’s scenario too – maybe Kim is the wealthiest of the four farmers,
and so she doesn’t mind being charitable to help out other small businesses. If that is the case,
then that’s great!

But one major problem that we have is that altruistic tendencies tend to taper off in larger groups.
What does this mean? Well, imagine that it is you and a friend living in your flat. Imagine they
work an evening shift and so they always have to rush off immediately after dinner. You might be
willing to wash their dinner dishes for them each evening to help them out. But would you feel as
charitable if you had 20 roommates who did that? Would you be willing to wash your 20
roommates’ dishes every night? Probably not.

This is not uncommon: people may definitely start to become less altruistic as the group of people
who benefit from their actions gets larger and larger. Maybe they feel taken advantage of if there
are too many people benefitting from their actions, so their tendency towards altruism decreases.
This means that we cannot rely purely on altruism as a mechanism for averting the prisoner’s
dilemma if multiple people are involved. This means altruism can’t be relied upon to solve the
problem of public good provision.

But we do see public goods being provided in reality. Beaches are kept clean, the parks in the City
are well-maintained, people have been trying to cut down their greenhouse gas emissions, and
people have been self-isolating during the COVID-19 pandemic. Why?

Maybe it has to do with something more than altruism. Could it be because people worry about
their reputation, or they worry that someone will punish them for behaviour that is not in
accordance with societal expectations? If the answer is yes, then that means that people are
worrying about how their actions will be perceived by others in the future, meaning that they are
not only interacting with people once.

This raises the issue of whether games being repeated rather than one-shot could help solve a
social dilemma. The answer is yes, because you may well have to interact with someone again, so
you may care about how your behaviour will be perceived, and act accordingly. Thus, this idea of
maintaining your reputation in a repeated interaction could influence behaviour, and avert a
social dilemma as well.

26
 4.3 Experimental economics – what is it?

People are strange creatures, and they have a number of different emotions that can affect their
decision-making process. Upholding their reputation, caring about others … all of these things
can make people behave differently to what we expect them to. But the question is, do they
actually behave differently to what we predict, and if so, how?

This question isn’t easy to answer theoretically. Can you model people’s idiosyncrasies (their odd
behaviours) on paper, or do you need something more? The answer to this is that we need
something more, and this “something more” comes from experimental economics.

Experimental economics is a branch of economics that deals with human behaviour and how it
changes due to different environments. Economic experiments are a very powerful tool that can
be used to measure behavioural traits by observing them under controlled conditions. There are
two main types of experiments that are used by economists to study behavioural changes:

1. Laboratory experiments – these involve subjects arriving at a controlled room (often a

computer lab) and being presented with a number of decision-making tasks to determine
behaviour under certain scenarios. These types of experiments offer the researcher much
more control, as they control their laboratory environment carefully to only alter the
variables they are interested in altering.
2. Field experiments – these involve the researcher changing some aspect of the
environment for certain groups of people out in the real world and observing how
behaviour changes as a result of this. This is often very tricky to get right, and researchers
need to choose exactly how to alter the environment to make sure it is ethically appropriate,
and also academically rigorous.

Note: Don’t get confused between these terms and a natural experiment, which is a scenario
where the environment changes for groups of people, but not because of the researcher. For
example, changes in behaviour before and after COVID-19 would be a natural experiment, but
not a field experiment, because there was no researcher who chose how, when and where to release
the disease for the purposes of academic investigation.

Experimental economics is particularly powerful because the results we get from them describe
what people actually do, as opposed to what they say they do, and “actions speak louder than
words”, right? For example, if I asked you a question in a questionnaire about how altruistic you
are, you may give an inflated estimate because you are trying to self-advertise and impress me with
how caring you are. Economic experiments capture actual behaviour, and often use money as an
incentive to get people to behave as they really would. Money is used as an incentive because
people tend to reveal their true behaviour when they are making decisions with real money
involved.

4.4 Public goods games as experiments

One common form of laboratory experiment that is run models the public goods game from
before. Generally, they work like this: There will be a number of rounds in a public goods game
(often 10, but this can vary depending on the researcher). In each round, the individual is given a
certain number of tokens (e.g. 10 tokens). In each round, individuals must decide how many of
their tokens they would like to contribute towards a public pot, and how many they would like to
keep for themselves in a private account. The number of tokens in the public pot are multiplied
by some number greater than one (and which is less than the number of participants, for reasons

27
we will discuss just now) and shared equally among all participants, whether they contributed or
not. This decision is repeated with a new set of 10 tokens each round.

Now, why do we multiply by a number less than the total number of players? This is a design
feature that makes the individual have a dominant strategy to not contribute to the public pot.
Let’s talk through an example to explain why this is the case: Imagine you are playing a public
goods game with 10 people, and the public pot it multiplied by 5 before dividing it amongst
everyone. Well, then think about what happens to each token you put in the public pot. For each
1 token that goes into the public pot, it will get transformed into 5 tokens. These 5 tokens then
get equally split amongst all 10 players in the game, giving each person 0.5 tokens from the pot.
But if you had simply put that token in your private account, you could have kept an entire token
for yourself! This means that putting tokens in your private account is better for you as an
individual.

Although each round’s decision to contribute to the pot is simultaneous (i.e. players do not know
how much other people are contributing, and we generally prohibit conversation in the labs to
make sure decisions are independent), the total contribution to the pot is shown after each round.
Thus, although a single round of the public goods game is a simultaneous game, this game gets
repeated each round.

If the public pot had been multiplied by some number greater than 10 (e.g. 11), then it becomes
clear that there is an individual incentive to put money into the public pot, because every 1 token
put in the public pot turns into 11 tokens. When you divide 11 tokens amongst 10 people, they
each get 1.1 tokens, which is more than they would get by putting the token in their private
account.

So, we can see that if the pot multiplier is less than the number of players, then there is a dominant
strategy to not contribute. This is beginning to echo our multi-player prisoner’s dilemma from
Section 4.1.

Also, just like in a prisoner’s dilemma, notice that when the multiplier is 5, if everyone were to
contribute, this would be best for society, because the total number of tokens in society would be
multiplied by 5, and everyone would have more tokens that they started off with. So, it is socially
optimal to contribute everything, but the individual incentive is to not contribute anything. This is
a textbook prisoner’s dilemma, and we can see how people behave by running experiments where
individuals play a public goods game.

28
 Box 8: How a public goods game works in practice (example)

Imagine that there are 4 people playing a public goods game. They have 10 tokens and have to
decide how many they would like to contribute to the public pot. The number of tokens
contributed to the public pot will be doubled and divided equally between all players.

Let the four people be represented as Player A, Player B, Player C and Player D. Each of their
decisions regarding how to split their 10 tokens is represented in the table below:

Player Public Account Private Account

A 4 6
B 7 3
C 1 9
D 0 10

We can see that between them, Players A, B, C and D contributed a total of 12 tokens to the
public pot (4+7+1+0). This number will be doubled, making the public pot worth 24 tokens.
Now, these 24 tokens will be equally split over the 4 players (meaning they get 6 tokens each,
which they will add to the tokens in their private account). This is shown in the table below.

Player Private Account Tokens from the Total tokens at

Public Pot end of round
A 6 6 12
B 3 6 9
C 9 6 15
D 10 6 16

This would then be the end of the first round of the game. When the second round starts, the
private account is normally set back to 0 tokens, and each player starts with a fresh set of 10
tokens which they must allocate between the private and public accounts once again.

4.5 Results of public goods games and how to encourage cooperative behaviour

The public goods game experiment that we discussed in Section 4.4 is designed in a way that it
mimics a prisoner’s dilemma. This means that the Nash Equilibrium of this public goods game
experiment would be for all people to contribute nothing to the public pot in each round of the
public goods game. But is this what we see in reality?

The answer – perhaps unsurprisingly, because we know that public goods are in fact maintained
in reality – is no. People behave differently to what the Nash Equilibrium predicts. In fact, results
from actual experiments show that when given 20 tokens to allocate between a private account
and the public pot, people contributed between 8 and 15 tokens to the public pot on average
(depending on which country they’re from). This is shown in Figure 7, below. The difference in
average contribution by country could be explained by different social norms in those countries –
i.e. maybe attitudes towards sharing with others are more prevalent in Denmark (where
Copenhagen – a high contributor – is) relative to Greece (where Athens – a low contributor – is)

29
Figure 7: Contributions to a standard linear public goods game with 20 tokens, by country

Source: CORE Text, 2020

Another interesting thing that we see in Figure 7 is that over time, contributions start to decline.
Although people in Denmark started off contributing 15 tokens to the public pot on average, by
10 rounds in, they were only contributing about 6 tokens on average. This might have to do with
reciprocal preferences – the idea that you will do to others what they do to you. You can think
of reciprocal preferences as preferences that pay others back for their behaviour in the past.

So, since the average contribution of tokens in round 1 for Danish citizens is 15 tokens, there
would likely be some people who contributed more than this. Maybe they feel taken advantage of
since they know that they are making a sacrifice and incurring a cost (of a low return to their
investment), but that their loss is benefitting others who did not contribute as much to the public
pot. So, they feel that actually, since their own generosity is not being reciprocated, it isn’t worth
it for them to contribute as much in round 2, and so they decrease their contribution, dropping
the average contribution of the group. This logic is circular and continues each round, decreasing
the amount of the average contribution in each subsequent round. This gives rise to a rather sad
conclusion: altruism is not enough to maintain a public good over time.

But what if we gave people another option to express their feelings of reciprocity? Instead of just
moving on to the next round, what happens if we allow players to punish other players? For
example, you can see everyone’s contributions to the public pot (anonymously) and for 1 token,
you can take 3 tokens away from someone’s earnings. How do you think this would affect
behaviour?

If you think that it would increase cooperation, you’re 100% right! Look at Figure 8 – this is the
same public goods game as we saw in Figure 7, but now individuals could punish those they
thought deserved it. The punishment was as I described it: for 1 token, you could decrease
somebody’s earnings by 3 tokens. Notice that in Figure 8, now, the average level of contributions
remained relatively high – in fact they even increased after round 1 as low contributors were
punished!

30
Figure 8: Contributions to a public goods game (with punishment) with 20 tokens, by country

Source: CORE Text, 2020

This pattern of behaviour is very common – when punishment is introduced, it increases

contributions in a public goods game tremendously. When I ran an experiment in 2016 using my
Honours class as subjects, the patterns followed these results exactly: without punishment,
contributions started relatively high and then declined; but when punishment was introduced,
contributions stayed relatively high.

Punishment of bad behaviour can take many forms – in fact, studies have shown that when stokvel
(informal savings clubs) groups have played public goods games, posting the group’s results
anonymously on a board and allowing each player 30 seconds to “speak their mind” about the
result also actually increased cooperative behaviour. This was likely due to speakers expressing
their displeasure at the non-cooperative members of the group, and even though the non-
cooperators’ identities weren’t made public, shame and guilt over their behaviour resulted in
greater cooperation going forwards.

Lesson 5: Ultimatum games and sequential-move games

5.1 A second experiment: the ultimatum game

In the previous lesson, we discussed the public goods game and how it is used in experimental
economics to capture individuals’ behaviour as relates to a social dilemma. We now turn our
attention to another type of game that is often used in economic experiments: the ultimatum
game.

The ultimatum game is set up as follows: There are traditionally two players in this game – a
Proposer and a Responder. The Proposer is given a certain amount of money (for the sake of
example, let’s assume this is R30). With this R30, the Proposer is asked to put forward an offer of

31
how to split the money between him/herself and the Responder. (Normally, these offers are made
in increments of R1).

The Responder will observe the offer, and they will decide to either accept or reject the offer from
the Proposer. If the Responder accepts the offer, then whatever split the Proposer put forward is
enacted, and this is how the money is split. However, if the Responder rejects the offer, then both
players receive no money at all (i.e. the R30 is lost).

As an example, think of the R30 being split between the Proposer and the Responder. If the
Proposer offered the Responder R13 and the Responder accepted it, then the payoffs would be:
R17 for the Proposer and R13 for the Responder. If the Responder had rejected this offer, on the
other hand, then the payoffs would have been: R0 for the Proposer and R0 for the Responder.

Hopefully it is clear at this point that the ultimatum game is set up slightly differently to the games
we have been studying up to now. In particular, notice that the ultimatum game requires the
Proposer to make an offer, the Responder to observe this offer, and to respond to it. In other
words, the Responder is making their move after being provided with new information.
This means that the ultimatum game is an example of a sequential move game.

5.2 Depicting a sequential move game

A sequential move game is different to a simultaneous move game because as players make moves,
they move into different states of the world where new information is available to the subsequent
players. To try and depict this difference from a simultaneous move game, we opt for a different
method of depicting sequential move games. We use what is called a game tree or extensive
form to represent a sequential move game (whereas our game tables or normal form
representation was for simultaneous move games).

The game tree of a sequential move game looks exactly like that: a bunch of branches that spread
out from a root and split into more and more sticks and twigs the further down the tree you move.
An example of a game tree is shown in Figure 9, which models a very simple version of the
ultimatum game played over R30, with certain important characteristics of the game labelled with
red arrows.

32
 Figure 9: The ultimatum game in a game tree format

player moving at
this node
Proposer
root node

action
node Offer R1 Split 50:50
branch

Responder Responder
Accept Reject
Accept Reject
payoffs – format:
(1st mover, 2nd mover)

(R29, R1) (R0, R0) (R15, R15) (R0, R0)

In order to put a game together, we begin with what is known as the root node. A node is simply
a dot or circle indicating a state of the world that players are in, and the root node is the state of
the world indicating the state of the world that players start the game from. Every node in a
sequential move game needs to be labelled with a player who is called on to act at this point of the
game. In the ultimatum game, the Proposer starts the game off, so we label the root node with
“Proposer”, indicating that this is the player who is making a move here.

Now, the player at the root node – in our case, the Proposer for an ultimatum game – has options
available for what move he/she will take on. Each action available to the player making a move
gets a branch, which is one of the straight lines extending out of the node. Let’s assume that the
Proposer only has two options available to him (for the sake of simplicity): Offer R1 or Split 50:50.
Since there are two actions available to the Proposer, we draw two branches out from the root
node. Each of these branches is labelled with a particular choice of action available to the Proposer.

Now that we know the Proposer has these two choices available to them, we know that there are
two possible states of the world that exist for the Responder: one where they were offered R1, and
another where they are offered a 50:50 split. This corresponds to the end of the branches extending
from the root node, and because these are now states of the world where the Responder has to
act, we need to indicate this by drawing in nodes and labelling them with “Responder”, showing
this is the player who will act at this node.

In the case of being offered R1 by the Proposer (i.e. we are in the bottom-left node of the game
tree), the Responder has two actions available to them: “Accept” or “Reject”. This means the
Responder has two branches extending from this node – one for “Accept” and another for
“Reject”. We draw these in. Similarly, when the Responder is facing an offer of a 50:50 split, they
have the same options available, hence producing the two branches from the bottom-right node.

Finally, the payoffs: in a sequential move game, the payoffs are always written up in the order that
players are called on to move. In other words, the first player to move will have the first payoff in
the brackets, and the second player to move will have the second payoff in the brackets. This

33
means that for us, the Proposer’s payoff will be first in the brackets, with the Responder’s payoff
second.

To calculate the payoffs, follow the branches down to a particular end-state of the world. For
example, what are the payoffs to the Proposer and the Responder when the Proposer offers R1
and the Responder accepts this. Hopefully it is clear that these specific actions will make you follow
the branches as follows:

Figure 10: Determining payoffs in an ultimatum game

Proposer The arrows in Figure 10, alongside, show

the branches you should follow to get to
the end-state of the world where the
Proposer offered R1 and the Responder
Offer R1 Split 50:50 chose to accept it. The payoffs to the
Proposer and the Responder should be
inserted at the end of this branch, as you
Responder Responder can see has been done in the figure.
Accept Reject
Reject Assuming the ultimatum game was played
Accept over R30, and the Proposer’s offer of R1
was accepted, what are the payoffs to the
proposer and the Responder, respectively?
(R29, R1) (R0, R0) (R15, R15) (R0, R0) Well, since the Proposer offered R1 to the
Responder and this was accepted, we
know this offer will be enforced, meaning that the Responder will get the R1 offered, and the
Proposer will keep the remaining R29. Because the Proposer’s payoff is listed first in the brackets,
we then get the payoff represented as (R29, R1), where the R is the Rands symbol (you could just
as easily have represented the payoffs as (29, 1) and still been correct). See if you can work out
where the remaining payoffs come from.

If you would like to see this game being built up step-by-step, the animations on Unit 4 Powerpoint
slides 69 to 71 show the process of building a sequential move game step-by-step.

5.3 Backward induction and the Subgame Perfect Nash Equilibrium

Now that you know how to represent a sequential move game in game tree format, we need to
think about how to solve for the equilibrium of a sequential move game. The way to do this is
through an algorithm called backwards induction, which uses the logic of looking forward and
thinking backwards. In other words, you work out what people at the end of the game will do first.
Once you have done this, you assume that this is the choice they will make, which allows you to
determine what the previous player will do. This works because once the choices in the final stage
of the game are determined, then the player in the penultimate (or second-last) stage of the game
knows if they choose X, then the next player will react with Y.

This sounds complicated and abstract, but will be much clearer when we do an example (which
we will very soon). Just remember that the process of solving for the equilibrium in a sequential
move game requires you to essentially reason backwards to determine the best response for the
player at every node they may be called on to act at.

34
This will make a lot more sense when we apply it practically. Let’s look at the ultimatum game we
developed earlier, which is shown in Figure 11.

Figure 11: How backward induction works

Proposer

Offer R1 Split 50:50

Responder Responder
Accept Reject
Accept Reject

(R29, R1) (R0, R0) (R15, R15) (R0, R0)

The logic of backward induction requires us to look at the choices made by the player at the end
of the game first – in our case this means the Responder. Let’s start by considering the decision
faced by the Responder in the bottom-left node (the decision inside the red circle of Figure 11).
At this point, the Responder has to choose whether to Accept or Reject the offer that has been
made. In other words, the Responder is making a decision between ending up with the outcome
in the blue oval, or the outcome in the purple oval. Since the Responder’s payoff appears second
in the brackets, this is essentially a decision between accepting the offer and getting R1 or rejecting
the offer and getting R0. Thus, accepting the offer is preferable to the Responder, so we indicate
this by pruning off the less-preferred branch – i.e. you put a line through the branch leading to
the inferior payoff. This is why there is a red line through the “Reject” branch of the tree on the
bottom left.

This same logic can be applied to the node on the bottom right, where the Responder has to decide
whether to accept the offer and get R15, or reject the offer and receive R0. Naturally, since R15 is
better than R0, the Responder would accept the offer, meaning we can prune off the Reject branch.

Now, we can move backwards to the Proposer. Since we know what the Responder will do in each
case, we can work out how the Proposer should behave. We know that if the Proposer were to
Offer R1, the Responder would accept this (from our logic above), leaving the Proposer with a
payoff of R29 (the first payoff in the bracket). Similarly, if the Proposer were to offer a 50:50
split, then the Responder would accept it (from our logic above) and this would leave the Proposer
with a payoff of R15. So, the Proposer is effectively choosing between Offering R1 and getting a
payoff of R29, or splitting 50:50 and getting a payoff of R15. Thus, the Proposer will choose to
Offer R1 and get a payoff of R29. This means we can prune off the Split 50:50 branch for the
Proposer.

35
All this pruning is shown in Figure 12, below. This is done step-by-step in the concept video for
this lesson, as well as on Slides 73-74 of your Unit 4 Powerpoint. Hint: If you are wondering how
to present your working when you solve for the equilibrium of a sequential move game, look at
Figure 12 – this is how you should present your answers.

Figure 12: How to solve a sequential move game

Proposer

Offer R1 Split 50:50

Responder Responder
Accept Reject
Accept Reject

(R29, R1) (R0, R0) (R15, R15) (R0, R0)

On a technical note, the equilibrium we just solved for in this sequential move game is a refined
version of the standard Nash Equilibrium. We call the equilibrium we find using backwards
induction a Subgame Perfect Nash Equilibrium (or SPNE for short). The SPNE of a
sequential move game is the set of strategies that describe the best response for each player
at each node that they COULD be called upon to act at, grouped by player. Notice the word
“could” in the definition – this means that in your SPNE, you should actually list optimal choices
for nodes even if they do not form part of the chain of actions that lead you to your equilibrium
outcome! This is actually just asking you for your complete best response strategy because
remember, a strategy is a plan of action that maps out all possibilities of what you could need to
do in a game.

This means that the SPNE for this game would technically be written out as SPNE = (Offer R1;
(Accept, Accept)). Now, for the purposes of this course, you will not need to worry about how
to correctly specify the SPNE of a sequential move game (you will learn this in more detail in your
second-year game theory course, ECO2007S).

What you do need to be able to do, though is:

• be able to use the logic of backwards induction to solve a sequential move game through
pruning (like you were shown in Figure 12)
• be able to identify the actions that individuals should take in a sequential move game to
reach the equilibrium outcome
• acknowledge and understand that the SPNE of a sequential move game requires a
complete list of best responses, and not only those actions that lead you to the equilibrium
outcome.

As an additional note: in this case, we identified the SPNE of an ultimatum game where the
Proposer only had 2 choices available to them. In general, there are many more choices available

36
to the Proposer (and this is not a problem – it just requires a branch drawn in for each choice),
and so we need to understand how people would behave in the more general version of the
ultimatum game.

Hopefully it will be clear that no matter what the Proposer offers the Responder, there is an
individual incentive for the Responder to Accept the offer rather than Reject it. Why? Well, if the
Responder accepts an offer, then they are given some money, while is they reject it, they get R0.
Something is always better than nothing, so the Responder would be willing to accept any non-
zero offer made by the Proposer. Knowing this, the Proposer would like to get as much money
for themselves as possible, but still induce the Responder to accept (because if the Responder
rejects the offer, then the Proposer loses all their money too). To do this, the Proposer should
offer the smallest non-zero amount of money to the Responder, because they know this offer will
be accepted, based on the fact that some money is always better than none for the Responder.

This logic means that the SPNE of the ultimatum game in general would be something along the
lines of: The Proposer should offer the smallest non-zero amount of money to the
Responder, and the Responder should accept this offer, as well as any other offer made to
them. This is the SPNE of an ultimatum game written out in words – note that it specifies the
best response for each player for every possible state of the world that they could be called on to
act at (including those states of the world that do not form part of the path leading to the
equilibrium outcome).

Lesson 6: Expected payoffs and competition in the ultimatum game

6.1 The ultimatum game in reality

In the previous lesson, we discussed how to solve for the Subgame Perfect Nash Equilibrium of
the ultimatum game. Remembering that the SPNE of a sequential move game needs to list the
best responses at all possible nodes for a player, we concluded that SPNE for an ultimatum game
was “the Proposer should offer the smallest non-zero amount of money to the Responder,
and the Responder should accept this offer, as well as any other offer made to them.”

But do we see people behaving according to this equilibrium in reality? Well, generally not, actually.
Think about yourself – if you knew that someone had R30 to split, and they offered you R1, would
you accept their offer, knowing that they would get R29? Or would you reject the offer, thinking
that if they’re going to be so stingy with their offer, you’re going to punish them for it? For most
people, they would reject the offer, because they feel that being offered R1 while the Proposer gets
R29 is unfair and they are unhappy with the distribution.

So, if the Proposer had R30, how much would you want them to offer you before you deem it
acceptable to you? In other words, how much of the R30 should be offered to you to get you to
accept the offer?

In reality, the amount of money that needs to be offered to a Responder before it becomes
acceptable varies across populations, but generally, people want to have a relatively even split.
Figure 13 plots out the responses from Kenyan farmers and US university students regarding
whether they deem a certain offer acceptable or not. The amount of the pie (total amount of
money) that was offered to the subject is labelled on the horizontal axis, while the proportion of
subjects who deemed such an offer acceptable is plotted on the vertical axis.

37
Figure 13: Acceptance rates of offers in the ultimatum game - Kenyan farmers vs. US students

Source: CORE Text, 2020

Unsurprisingly, as the proportion of the pie offered to subjects increases, the acceptance rate goes
up, until the offer of a 50:50 split gets everybody to accept the offer. (Any offers above 50% of
the pie were considered acceptable to everyone, so they are not recorded in this graph).

Another striking result from Figure 13 is how when Kenyan farmers are offered 10% of the money
or 20% of the money, very few of them will deem this offer acceptable. This stands in stark
contrast to US students, who were substantially more willing to accept these low offers (about
55% of students would accept an offer of 10% of the pie, while about 3% of farmers accept this
same offer). Also, only 10% of farmers would accept a 20% offer, while 63% of students would
accept this … what’s going on?

Perhaps the answer to this lies in something like the social norms of the country. Maybe US
students are generally more desperate for money, and are willing to accept any amount of money
to help them buy their next pack of instant noodles (for example). Otherwise, perhaps Kenyan
farmers live in a very tight-knit community and so they feel very strongly that the right thing to do
is to treat one another as equals, so their feelings of reciprocity are stronger. Perhaps higher
inequality in Kenya than in the US may also impact decisions: if there is higher inequality, then
maybe this means that people feel much more strongly about not perpetuating it. There are many
different reasons for why the American students might have been willing to accept low offers
while Kenyan farmers would not.

Now, instead of just asking individuals whether they would accept an offer of 10%, 20%, etc.,
researchers actually got the Kenyan farmers and the US students to play the ultimatum game. Now,
we know that offering 50% of the pie to a Responder will guarantee them accepting this offer
(whether they are a farmer or a student), so do you think this is what Proposers actually did?

38
Figure 14: Ultimatum game results - Kenyan farmers vs. US students

The answer is no – as we can see in Figure 14. In Figure 14, the results of the ultimatum game are
presented, with the total height of the bar indicating the number of Proposers who made the
relevant offer, and the dark shaded part of the par indicating the proportion of offers rejected by
Responders. So, for argument’s sake: about 50% of Kenyan farmers offered 40% of the pie (red
bar), and only about 4% of these offers were rejected by Responders (the small dark maroon
section of the bar).

One of the very interesting results we see here is that there are actually very few farmers and
students who offer 50% of the pie to their Responders. The most common offer among farmers
is 40% of the pie, while the most common offer among students is 30% of the pie. Why is this the
case? Could it be that although there is a norm of fairness, US Proposers know that there are more
US students willing to accept a low offer, so they take a risk and hope that they come up against a
Responder who would accept a low offer? Because naturally, if a lower offer is accepted, the
Proposer gets more money for themselves, but they are taking a risk … This is what we discuss in
the next section of this lesson.

6.2 Uncertainty and expected payoffs

In reality, we often come across situations where there is some uncertainty over the outcomes.
Like in the case of the ultimatum game, where a Proposer doesn’t know for sure whether their
offer will be accepted or not – there is a level of uncertainty over the outcomes. The Proposer
knows there is a chance that the offer will be accepted, and a chance that the offer will be rejected.
Similarly, when you roll a die, or flip a coin, there is a chance that each of the potential outcomes
will arise: on a fair 6-sided die, there is a one in six chance of getting each of the numbers from 1
to 6; on a coin, there is a 50% chance of it coming up heads, and 50% chance of it coming up tails.

Now, the Proposer in the ultimatum game is trying to do two things with their offer: they want to
try and maximise the amount of money they can get (because you want as much money as possible
out of this game), but they want to do this bearing in mind that different offers are more or less

39
likely to be accepted by a Responder. So, they are balancing the value of the amount of money
they could get out of an offer with the probability of it being accepted.

We are going to thus introduce the concept of an expected payoff. An expected payoff is
essentially a probability-weighted average of the payoffs available in a scenario. In other
words, an expected payoff is a type of average payoff you can expect from taking a particular
action that weights each payoff by the probability that you will actually get it.

For example, imagine that there is a coin to flip, and if it comes up heads, you win R20, while if it
comes up tails, you win R50. What is the expected payoff in this scenario?

Well, since the probability of getting the coin to come up heads is 50% and the probability of the
coin coming up tails is 50%, we know that half the time, you will get a payoff of R20, and the other
half the time you get a payoff of R50. So, we would calculate the expected payoff by multiplying
each payoff by the probability that it occurs, and then adding all these products together
– i.e. it would be calculated like this:

Probability of coin coming up heads Probability of coin coming up tails

!"#$%&$' )*+,-- = 0.5×320 + 0.5×350 = 335

Payoff when the coin shows heads Payoff when the coin shows tails

So, on average, we would expect to get R35 from this scenario … but that isn’t one of the payoffs.
So what exactly is an expected payoff practically? Well, think about a situation where I present
this scenario to every person in the first-year microeconomics class. There are about 1400 of you,
but I say that instead of me paying you out individually, each person’s money they earn from the
scenario gets put into a piggy bank/account for later. Once I have played this game with everyone,
I will then split the total in the account equally amongst everyone. We would expect that the
average amount of money you get from the account would be R35 (this obviously would be
closer to R35, the more people who played the game). The reason this works out is because of a
statistical rule called the law of large numbers, which you will learn more about in statistics.7

So, maybe the Kenyan farmers and US students took the probabilities of their lower offers being
accepted into account, and decided that actually, a lower offer would net them a higher payoff on
average than before. In other words, maybe offering 40% of the pie gave the highest expected
payoff to farmers, rather than offering 50% of the pie.

Imagine that the probabilities of different offers being rejected are given by the numbers reported
in Table 1. These probabilities of rejection could be determined by a number of factors, such as
social norms and community expectations. Let’s imagine that the game was being played over
$100, so $40 represents 40% of the pie, and $50 represents 50% of the pie. Now, let’s ask ourselves
whether a Kenyan farmer should offer $40 or $50 to maximise their expected payoff.

7
In essence, though, this law states that as the number of times an independent event is repeated increases, the
proportion of times that each event outcome occurs while converge to its population probability.

40
Table 1: Probability of rejection of offers in ultimatum game, Kenyan farmers vs US students

Amount offered $0 $10 $20 $30 $40 $50

Kenyan
100% 100% 90% 48% 4% 0%
Proportion farmers
rejected US
100% 40% 35% 15% 10% 0%
students

To do these calculations, let’s consider what the Kenyan farmer can expect when they offer $40:
If the offer is accepted, the Proposer gets $60, and if it is rejected the Proposer gets $0. The offer
of $40 is rejected 4% of the time. So, it is accepted 96% of the time. Thus, the expected payoff
can be calculated:

𝐸(𝑃𝑎𝑦𝑜𝑓𝑓) = (0.04 × 0) + (0.96 × 60) = 57.6

Now, let’s consider what the Kenyan farmer can expect when they offer $50: If the offer is
accepted, the Proposer gets $50, and if it is rejected the Proposer gets $0. The offer of $50 is
rejected 0% of the time. So, it is accepted 100% of the time. Thus, the expected payoff can be
calculated:

𝐸(𝑃𝑎𝑦𝑜𝑓𝑓) = (0 × 0) + (1.00 × 50) = 50

So, notice that actually, Kenyan farmers get a higher expected payoff when they offer $40 – this
means that on average, Kenyan farmers will get more money from offering $40 to their Responders
than they would offering $50, even after factoring in the fact that a $40 offer will be rejected more
often than a $50 offer.

Exercise: See if you can calculate the expected payoffs for US student Proposers for each offer.
There is a separate document that details the working to show the US student expected values. See
whether you can confirm that US student Proposers maximise their expected payoff with an offer
of $30.

So, in short, maybe the farmers and students are taking calculated risks – they recognise that if
they offer less than half the pie to the Responder, yes, it might be rejected, but it might also be
accepted, and when it gets accepted, the Proposer does really well. So, maybe they feel the risk of
having their offer rejected is minimal in comparison to the payout they could get when their offer
is accepted. If we believe that people are willing to take calculated risks, then we can use expected
payoffs to help us look at how people will behave.

6.3 The ultimatum game with many responders

The final thing to discuss in Unit 4 is one that we will touch on very briefly. Basically, we are going
to ask the question “What happens if there is more than one Responder in an ultimatum game?
Does this change what people consider to be acceptable behaviour?”

Very simply, what we look at here is an extension of the ultimatum game where there are two
Responders instead of just one. The set-up is very similar in that a Proposer makes an offer.
However, now the Responders both get to look at this offer and Accept or Reject it. If one
Responder accepts the offer, while the other rejects it, then the split will be enacted over the

41
Propose and the accepting Responder, while the rejecting Responder gets nothing. If both
responders accept, then one of the Responders is randomly chosen to get the accepted split of the
money. It is only if both Responders reject the offer that the Proposer gets no money.

In essence, this experiment has introduced an element of competition into the ultimatum game,
because you now have to consider what another person in your position is doing, and as a result,
your power as a Responder has decreased: you no longer have absolute power over whether the
Proposer gets money or not. If you get a bad offer, you can reject it, but unlike the normal
ultimatum game, the Proposer is not necessarily getting no money … that will only happen if the
other responder also rejects the offer.

So, when there are more Responders in the ultimatum game, your feelings of reciprocity and
revenge may not be directly translated to the Proposer. This will likely make you feel less strongly
about rejecting low offers to punish the Proposer, and you may be willing to accept lower offers
because you know your power has been diminished.

This is exactly what happens, in fact. Look at Figure 15: the proportion of low offers (between 0
and 35% of the pie) that are rejected is much lower when there are 2 Responders than the case
when there is a single Responder. So, perhaps power dynamics and competition can also influence
your behaviour: since you know that you can no longer guarantee punishment for bad behaviour,
you become more accepting of it, and decide to rather go with the flow than try to prove a point
to the Proposer.

Figure 15: Ultimatum games with one responder vs two responders

42