CHAPTER 4

CITY SIZE

1. Rank Size Rule

Economists have observed a constant relationship between city size and its rank
(within a country/region)

For instance, if the largest city has a population of 10m, the second largest has 5m,
the third largest has 3.33m and the fourth largest has 2.5m etc.

In general, rank-population relationships more or less follow the rule:

Ranki = N1/Ni,

where N1 is the population of the largest city(i.e., which is rank #1) and Ni the
population of the ith city.

Example 1
The largest city in a region has a population of 20 million. How large is the 5th
largest city? That is, N1=20m, N5=?

5 = 20m/N5 à N5=4m

Example 2
The 10th largest city in a region has a population of 4 million.
How large is the 5th largest city? That is, N10=4m, N5=?

Find N1 first
10=N1/4m à N1=40m
Now find N5
5 = 40m/N5 à N5 = 8m
The “Rank-Size-Rule” is not normative/causal -- is purely descriptive.

2. Utility and City Size

The growth of a city is associated with benefits and costs.

- Agglomeration economies (I) increase labor productivity but at a decreasing rate
- Agglomeration diseconomies (C) are caused by increasing commuting cost, crime
rates, environmental problems. We assume the marginal cost to increase with city
size.

Overall utility (U) is defined as the sum of productivity induced income increases (I)
minus agglomeration costs: U = I – C

At the margin, that is for the last unit, we state that

!" !$ !%
!#
= !#
− !#
, i.e., marginal utility equals marginal income minus marginal cost
(all with respect to city size N)

To maximize Utility we set marginal utility equal to zero

!" !$ !%
!#
= !#
− !#
=0,

which requires that marginal income and marginal costs are equal

!$ !%
!#
= !#
𝑜𝑟 𝑀𝐼 = 𝑀𝐶 .

The Figure below shows a graphical representation of utility maximization:

C,I
I

Max Utility

The difference between I and C is maximized when the slopes of both curves are
identical.

The same relationship can also be expressed by a utility curve where U=I-C:
 U

Max Utility

N
Sopt

Mathematical Example:
Assume income in a city is given by I = 20N – 0.5N2 and agglomeration cost is given
by C = 2N2, where N stands for city size (i.e., its population).

Overall utility is maximized where marginal income (MI) equals marginal cost (MC).
20-N = 4N à N*=4

3. Systems of Cities

Cities may be too large but not too small

Assume a workforce of 6m people and three possibilities of how they can live
(World I, II, and III)
World 1: 6 cities with a population of 1m each with a low utility of $4.5/worker (S)
World 2: 3 cities with a population of 2m each with a utility of $12.5/worker (M)
World 3: 2 cities with a population of 3m each with a utility of $6.5/worker (L)
We will observe the following dynamics:

World 1 is unstable and will not be an equilibrium solution. Suppose there are 6 cities
(A, B, C, D, E, F) of 1m each. There is a strong incentive for a person to move — for
instance, from A to F — since utility per worker will increase in F. Since, on the other
hand, it decreases in A, there are further incentives to leave A. At the end of this process,
A will disappear. Similarly, B and C may disappear, leaving only three cities with 2m
people each. Since World 1 is on the increasing branch of the utility curve (i.e.,
increasing agglomeration results in higher utility), the number of cities and their size will
be driven toward the optimum (World 2).

World 2 is a stable equilibrium. Suppose there are three cities (D, E, F). Any movement
from one to another will make one city larger and the other smaller. Either case will lead
to decreasing utility and is therefore not worthwhile.

World 3 is a stable equilibrium. Suppose there are two cities (E, F). Any movement
from city E to city F will make city F larger and E smaller, and vice versa. Wherever one
moves — from E to F or the other way around — utility will decrease, because one
moves from a shrinking city to a growing one. Although World 3 is suboptimal, it is a
stable equilibrium.

Example: Brasilia
Even cities that were designed to be at the optimal population and maximum utility (point
M in the graph above) cannot maintain this positions. That is markets push the population
above the optimum; e.g., people from poor rural regions move in causing the city’s size
to shoot beyond the planned targe.
The best example of a planned city is Brasília, the new capital of Brazil. Brasilia was
a planned city developed by Lúcio Costa, Oscar Niemeyer and Joaquim Cardozo in 1956
in a scheme to move the capital from Rio de Janeiro to a more central location. It’s
population target was 500,000. The city was laid out in the shape of an airplane (“plano
piloto”) with no plans of any extensions. Below is the plan:

The target population was already reached in 1970 and Brasilia kept growing. Now,
Brasilia has a population of approx. 4million.

One of Brasilia’s features is its inverse population density pattern, i.e., the outer districts
exhibit higher population densities that the core (a trait it shares with Moscow and
Johannesburg).
When segmenting by income groups, we see that while high income earners live in or
near the core, lower income groups are now located the farthest away from the
city the center.

For more details please read,

Alain Bertaud (2010). Brasilia’s spatial structure: Between the Cult of Design and
Markets. (read up to page 9, marked in the text)
There are also a few nice Youtube clips on Brasilia, please watch.
[Link]
[Link]

4. Differences in Size

One way to explain differing city sizes draws on differences in localization

economies (see the Figure below).

First, companies with localization economies form clusters. Depending on the extent
of these external economies the clusters differ in size. This creates towns of
different sizes (in the Figure below, ranging from 120 to 30 jobs)

Second, companies that benefit from urbanization economies move toward the
larger ones of these clusters making larger towns even larger. In the Figure this
relates to an addition of 80 and 20 jobs in large and medium sized cities,
respectively.

Third, these export-oriented workers will be cause “local employment,” in sectors

such as food, education, entertainment, banking or insurance. The ratio of local job
per export job varies with the function of the city. Since larger clusters assume
functions for smaller clusters (e.g., there cannot be a symphony orchestra or a
university in each small town), their ratio is higher than the one for small towns. In
the Figure below, there is one local job for every export job in small cities. This ratio
is 1.5 for mid-sized cities and 2 for large cities. Different local job ratios further
amplify the differences in city size.
The table below explains that large clusters have more local jobs per export jobs because
they serve the population of smaller places. The threshold population for certain
functions varies. While it only takes a population of about 300 people to sustain a church,
a high school requires a population of at least 710 people, a florist needs at least 2188
people. Berry and Garrison (1958)1 computed threshold populations for various
businesses in Snohomish County, WA.

1Berry, B. and Garrison, W. (1958). Functional bases of the central place hierarchy.
Economic Geography, 34,145-154
Function Threshold Population

Service Station 25
Church 300
Restaurant 307
Post Office 358
High School 710
Beauty Shop 788
Dry Cleaners 875
Liquor Store 1287
Optometrist 1800
Florist 2188
Merging these results with the idea of city size and utility from above yields the
following Figure:

Locational equilibrium requires that workers in are indifferent between living in

one of the three cities. Thus, utility per worker (u*) is identical at each of the three
cities. Note that all equilibria (s,m,b) are on the downward sloping branch of the
respective utility curve and are thus stable outcomes.
City-Specific Innovation can’t be fully internalized

The Figure below shows two identical cities with a population of 6m people each. At the
beginning, both cities are at point I and both experience a utility of u=70.

Now, assume one of the two cities is very innovative and shifts up its utility curve
enabling it to reach point j with a utility of U=80. However, this point is not a stable
equilibrium. People in the less innovative city have string incentives moving to the
innovative city and reap its benefits. Due to this, the more innovative city will grow to a
population of 7m and its utility per worker will fall. The less innovative city will shrink
to a population of 5m and its utility per worker will increase. People will move from the
less innovative to the innovative city until the utility per worker is identical at both
locations, which is the case at point s and b.

à the benefits of innovation cannot be internalized but spread into space.