Introducing Land Constraints to Macroeconomic Agent-Based Models

Abstract

We present a macroeconomic agent-based model that incorporates land to constrain maximum agricultural production. The model contains three types of agents: firms, households, and land plots. Firms employ households to produce consumer goods which household then buy from firms. The maximum production of firms is limited by the area of land available to them. Varying the availability of land and investigating the results on unemployment, wage rates, firm turnover rates, profits, and inequality among households and firms shows that the interests of firms and households conflict regarding the ideal amount of land available for production.

Full text

Introducing Land Constraints to Macroeconomic
Agent-based Models
Jacob Kelter1, Uri Wilensky1, and Joseph Potvin2
1Northwestern University, 633 Clark St, Evanston, IL 60208,
jacobkelter@u.northwestern.edu and uri@northwestern.edu
2Xalgorithms Foundation, https://xalgorithms.org
jpotvin@xalgorithms.org
License: CC-by 4.0 International
Abstract. We present a macroeconomic agent-based model that incor-
porates land to constrain maximum agricultural production. The model
contains three types of agents: firms, households, and land plots. Firms
employ households to produce consumer goods which household then
buy from firms. The maximum production of firms is limited by the area
of land available to them. Varying the availability of land and inves-
tigating the results on unemployment, wage rates, firm turnover rates,
profits, and inequality among households and firms shows that the inter-
ests of firms and households conflict regarding the ideal amount of land
available for production.
Keywords: macroeconomics, agent-based modeling, land
1 Introduction
Agent-based models (ABMs) can reproduce many stylized facts of the economy
including business cycles, endogenous fluctuations and correlations in macro-
level economic variables, firm-size distributions and more [1–3]. Increasingly over
the past three decades, ABM literature has addressed the intersection of eco-
nomics and the environment, including energy technology markets [4, 5], and
climate-economy interactions [6, 7]. There have also been sector-oriented inves-
tigations of agriculture and land-use practices and their interactions with the
economy [8, 9]. However, these models are oriented to a specific sector. Macroe-
conomic ABMs have not integrated “Land” as a primary factor of production
along with the other three factors emphasized by classical economics: “Labor”,
“Capital” and “Organization” [10].
The ABM presented in this paper lays the groundwork for investigating feed-
back loops between the economy and ecological integrity of land. This is part
of an ongoing modeling project of wider scope, but the aspect of the model
presented here focuses only on how availability of productive Land impacts the
economy.
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

2 Kelter, Wilensky, Potvin
2 The Model
2.1 Model Structure
The model3, implemented in the NetLogo agent-based modeling environment
[11], contains three classes of agents: firms, households, and land. Figure 1 rep-
resents the relationships between the di↵erent types of agents. We take a pure
agent-based approach—as opposed to representing labor and consumption at an
aggregate level as in the Dosi et al. [3, 12] family of models—so that we can
explore the e↵ects of various parameters on wealth distribution and the e↵ects
of various frictions in the labor market on the overall economy.
Firms produce homogeneous consumer goods on plots of land and sell them to
households. A firm’s output is constrained by the area of land available to it and
the land’s productive capacity per unit area. In the experiments reported here,
the productive capacity of land per unit area is held fixed across all experiments
and the area of land available is held fixed within experiments but varied across
them. In future work, the activity of a firm will increase or decrease the land’s
productive capacity over time depending on the quality of its land management
practices. This is depicted in Figure 1 with a grey dotted line since it is part of
the model design but not used in this preliminary paper.
The model is initialized with a set of households, firms, and plots of land.
Each firm is associated with a single plot of land. We abstract from land markets,
and therefore land is not bought or sold. Households begin with an initial amount
of currency liquidity, they each supply their Labor to a single firm, and they have
links to several firms as consumers. Firms raise initial start-up funds from one
or more households as investors (described in more detail below). To keep the
model simple, there is no banking sector, finance, or credit in the model. Firm
entry is funded directly by households rather than through banks as in the Delli
Gatti et al. [2] family of models. The present model holds money supply fixed,
resulting in zero inflation; a subsequent version of the model will have a dynamic
monetary system.
2.2 Sequence of Events
The model uses discrete time steps, and the sequence of events in each period is
as follows:
1. At the start of the period, firms without enough liquidity to pay a single
worker go bankrupt and are replaced.
2. Firms, to plan the upcoming period:
(a) Estimate demand in the upcoming period
(b) Adjust wage rates based on success/failure in filling job positions
(c) Plan output based on previous sales (constrained by land) and adjust
price
3Model code available at https://github.com/jzkelter/tabular-standards/wiki/
How-do-download-and-use- the-model
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

Introducing Land Constraints to Macroeconomic Agent-based Models 3
(d) Adjust desired labor (lay o↵workers or posts job openings)
3. Households:
(a) Update consumer links
(b) Adjust reservation wage
(c) Search for employment
4. Firms:
(a) Pay wages (happens at the beginning of the month to simplify firm
planning and after households have searched for employment so new
hires get paid)
(b) Distribute profits from prior period (this can only happen after paying
wages)
5. Households set consumption for the month
6. Firms produce output goods
7. Households buy and consume goods
Fig. 1. The structure of
the model. Firms each
have a plot of land.
Households and firms
have three types of rela-
tionships: employment,
consumption, and equity
(investment).
Each of these events is described in more detail below. In most cases, when
describing procedures that involve parameters, we use the parameter name from
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

4 Kelter, Wilensky, Potvin
the model’s code directly instead of translating the parameter names from the
code to mathematical symbols, which would then need to be translated back
again by anyone who reads the code. To further ease reading, we usually de-
scribe the model using the concrete values for parameters used in simulations,
and footnote what the relevant parameter name is in the code. We do use math-
ematical notation in select cases when we think it facilitates communication.
For full specification of the model, see the online supplemental documentation4
based on the Dahlem guidelines [13].
2.3 Bankruptcy and Firm Replacement (step 1)
At the beginning of each period, if a firm does not have enough liquidity to pay
a single worker, it goes bankrupt. Any remaining liquidity it has is returned
to shareholders. A new firm is then created which raises start-up capital from
households5. The new firm asks households in a random order for funds. House-
holds are willing to invest up to half of their current liquidity in the new firm,
but their investment is only accepted if it represents at least 10% of the total
value the firm is raising. In this way, there is an emergent class of capitalists
based on wealth rather than a hard-coded class of capitalists as in [1]. House-
holds own a fraction of the firm in proportion to what fraction of the startup
funds they provided. The firm is initialized with 20 consumer-links6, otherwise,
it would usually fail to sell anything the first period and immediately go out
of business. These initial consumer links can be thought of as being due to an
initial advertising campaign. The new firm is also automatically given the plot
of the land that the bankrupt firm vacated.
2.4 Firms Estimate Demand (step 2a)
Firms estimate demand based on a rolling average of past sales, Save(t)which
is updated by the sales of the previous period by the following equation:
Save(t)=mSav e(t1) + (1 m)s(t)
where s(t) is sales in the time period, t, that just passed and mis a ”mem-
ory” constant7between 0 and 1 that determines how much the firm remem-
bers/weights previous average sales compared to the prior period’s sales.
2.5 Firms Adjust Wage Rates (step 2b)
Firms adjust wages based on their success or failure in hiring. If a firm wanted to
hire a worker last month and failed, it increases its wage to attract workers. On
4https://github.com/jzkelter/tabular-standards/blob/main/Main%20Model/
Dahlem Description.md
5The amount is equal to the parameter STARTUP-LIQUIDITY
6INITIAL-CONSUMER-LINKS
7FIRM-MEMORY-CONSTANT
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

Introducing Land Constraints to Macroeconomic Agent-based Models 5
the other hand, if a firm has had no vacancies for the past 12 months8, the firm
decreases wages. In either case, the increase/decrease is by a random fraction
chosen uniformly between 0 and 20%9.
2.6 Firms adjust planned output and price (step 2c)
Ideally, firms want to fully satisfy their expected demand. Since demand may ex-
ceed expectations, firms try to keep a bu↵er stock of 50% of expected demand10 .
So, after production and prior to selling goods, firms aim to have 150% of ex-
pected demand in stock. Goods are non-perishable in the model. This means
that if a firm has already built up its bu↵er, it rarely has to produce much
more than expected demand. Firms may not be able to produce enough to have
150% of expected demand in stock before sales begin due to limited land, liq-
uidity, or failure to hire adequate workers. As this is the planning stage, only
the first two limitations come into play, and they determine how many workers
the firm will aim to have this period. The number of workers a firm desires is
equal to target production divided by the “tech-parameter” which determines
labor productivity. This assumes that production is a linear function of labor
(no changing returns to scale). For the purpose of this paper, tech-parameter
is uniform across firms and held constant (i.e., there is no technological inno-
vation). The pseudo-code in Algorithm 1 describes the process of firms to plan
output and their desired number of workers.
Algorithm 1 Firm process to plan output
1: set target production = 1.5 * expected demand – current inventory
2: if target production >total productive capacity of land then
3: set target production = total productive capacity of land
4: end if
5: set target n workers = target production / tech parameter
6: if liquidity <target n workers * wage-rate then
7: reduce target n workers to maximum that can be a↵orded given liquidity
8: end if
After planned output has been decided, the firm adjusts its price. Following
[2], a firm will not increase both output and price. A firm will raise prices only
if all three of the following conditions are met:
1. Demand was higher than expected last period (which means expected de-
mand this period is higher than last period)
2. The firm is unable to satisfy expected demand this period (either due to
lack of liquidity to hire workers, or due to reaching the maximum productive
capacity of the land).
8MONTHS-TO-LOWER-WAGE
9MAX-WAGE-CHANGE
10DESIRED-BUFFER-FRAC
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

6 Kelter, Wilensky, Potvin
3. The firm’s price is less than the average price of other firms
The rationale for these conditions is that firms aim to increase market share
before increasing unit profits. If condition 1 is met but not 2, this means the
firm will try to meet the increased expected demand at the current price. If
conditions 1 and 2 are both true but not 3, the firm will not risk losing market
share by raising prices further above the average price of other firms.
A firm will decrease price if the following three conditions are met:
1. Demand was significantly less than expected last period, as measured by
inventory being 120%11 or more of the ideal bu↵er amount
2. The firm has enough liquidity to meet expected demand this period
3. The firm’s price is more than the average price of other firms
Condition 1 guarantees there is surplus. Condition 2 checks that the firm
is able to fulfill expected demand, which suggests there will probably still be
surplus. If this is true and the firm’s price is above the current average, the firm
decreases price to try gain market share.
In the case of either raising or lowering price, the firm increases or decreases
its price by a random percentage between 0 and 20%12.
2.7 Firms adjust labor (step 2d)
If a firm has fewer workers than desired, it automatically has a job opening(s)
available. It is then left to households searching for jobs to find these firms. If
a firm has more workers than desired, it will attempt to lay o↵workers. Rather
than keep track of labor contract lengths for each worker, we instead allow firms
to lay o↵workers probabilistically. The firm attempts to lay o↵each worker it
does not want and succeeds with probability equal to the parameter LAYOFF-
PROBABILITY. A low layo↵probability is equivalent to long labor contracts
and a high layo↵probability is equivalent to short labor contracts.
In addition, a firm that cannot a↵ord to pay its current number of workers
lays o↵as many workers as needed so that it will be able to a↵ord the wage bill.
2.8 Households update consumer links (step 3a)
Households have 7 consumer links13. If they have fewer than this (due to a firm
going out of business) they create new trading links. If a household has more
than 7 consumer links, it randomly deletes one.
After guaranteeing they have enough consumer links, households probabilis-
tically search for more desirable trading links. With a 25% probability14,the
household will pick a random firm and, if its price is cheaper than the household’s
11BUFFER%-TO-LOWER-PRICE
12MAX-PRICE-CHANGE
13N-TRADING-LINKS
14PROB-REPLACE-FIRM-PRICE
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

Introducing Land Constraints to Macroeconomic Agent-based Models 7
most expensive current consumer link, will delete its most expensive consumer
link and create one with the cheaper firm. If a household has a consumer link
with a firm that failed to satisfy its demand last period, with a 50% probabil-
ity15 the household will replace it with the randomly selected firm. In both cases,
randomly selected firms are chosen from the set of firms the household does not
currently have a consumer link with and their chance of selection is weighted
by the number of consumer links they have (larger firms are more likely to be
selected).
2.9 Households adjust reservation wage (step 3b)
A household’s reservation wage is the minimum wage it is willing to accept for
employment. If a household is unemployed, it decreases its current reservation
wage to 90%16 of its current value. If the household is employed and its current
wage is above its reservation wage, it increases its reservation wage to equal its
current wage.
2.10 Households search for employment (step 3c)
An unemployed household checks five17 randomly chosen firms for job openings
and takes a job with the first one that o↵ers a wage above the household’s
reservation wage. An employed household will check one random firm for a better
paying job if its wage is below its reservation wage or with probability equal to
10%18. If the randomly chosen firm has a job opening at a better wage, the
household switches jobs.
2.11 Firms pay wages and distribute profits (step 4)
At this point, all employment is set for the month and firms pay their workers
at their current wages. In case sales are lower than expected, firms keep some
liquidity in reserve equal to 30% of current labor costs. Whatever liquidity re-
mains is distributed to households with equity in the firm in proportion to their
equity.
2.12 Households set consumption for the month (step 5)
Households, having been paid, set their consumption for the month based on the
equation:
C=L↵
15PROB-REPLACE-FIRM-QUANT
16RES-WAGE-CHANGE
17SEARCH-N
18SEARCH-BETTER-JOB-PROB
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

8 Kelter, Wilensky, Potvin
where Cis planned consumption, Lis the household’s current liquidity, and
↵is a parameter19 determining diminishing marginal utility of consumption.
Based on this equation, consumption always increases with increased liquidity,
but unless ↵= 1, the increase is sub-linear. The simulations in this paper use
↵=0.6.
2.13 Firms produce output goods (step 6)
Firms produce output based on the number of workers they have and land con-
straints. Output is linear with the number of workers (output = n-workers *
tech-parameter) up until the total productive capacity of the land which equals
land area times productive capacity per area. This is equivalent to an assump-
tion that each worker can work a certain area of land and there is no benefit
from additional labor applied to the land. We assume productive capacity per
area of 1 such that the total productive capacity is equivalent to the area of land
available.
2.14 Households buy and consume goods (step 7)
Households visit the firms they have consumer links with one at a time in a
random order20 and attempt to satisfy their demand by buying from that firm.
If the firm runs out of inventory, the household visits the next consumer firm up
until it either satisfies its demand or runs out of consumer links.
3 Simulations
3.1 Simulation Setup
The simulations were designed to explore the e↵ect of altering the availability of
land on the aggregate economy. The main variable varied is available land area
per capita21 , denoted here as Apc. It is expressed per capita (per household) to
make simulations with di↵erent numbers of households comparable. Each run
can be thought of as a parallel universe with more or less available land. In the
real world, available land per capita can vary due to population growth/decline,
ecological degradation/restoration, or changes in land-use.
Productivity of land per unit area was assumed to be 1. The tech-parameter
was chosen to be 1 as well—meaning that one unit of labor can work one unit of
land and produce one unit of output per period. So, when Apc = 1 all households
could theoretically be productively employed. When Apc <1, the total amount
of land is less than the overall labor capacity. Apc was varied between 0.2 and 10
19DIMINISHING-UTILITY-CONSTANT
20PICK-CHEAPEST-FIRM? = false. If households do sort firms based on price,
results are similar in most parameter settings. A full discussion of the impact of house-
holds sorting firm by price is beyond the scope of this paper.
21LAND-AREA-PER-CAPITA
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

Introducing Land Constraints to Macroeconomic Agent-based Models 9
with varying increments22 depending on how close the value of Apc was to the
critical value of 1. The simulations were run with 1000 households and 60 firms
for a total of 500 periods after a 500 period “burn-in” time to allow the model
to reach a steady state. Based on visual inspection, the model reaches a steady
state after around 200 periods; a 500-period burn-in was chosen to safely reach
the model’s steady state. The number of households in the model is arbitrary
but similar to prior macroeconomic ABM research [2, 14]. The ratio of firms to
household is chosen to roughly match a typical firm in the United States. In the
United States, 85% of employers have 19 or fewer employees [15], and the average
firm in the model will have around 15 workers when unemployment is low. Of
course, in the real economy there is a small number of massive employers, but
this is ignored in the model. The appendix lists all the parameters used in the
simulations and their values.
3.2 Results
To give an initial sense of the output of the model, Figure 2 shows time series
of unemployment, mean wage, and mean profits for Apc = 1 (blue) and Apc =3
(orange). Wages and profits are expressed as a fraction of average consumer good
price because monetary units in the model are arbitrary. Dividing by the price
of goods gives a measure of actual purchasing power. The time series show 500
periods of model execution starting after the 500 period “burn-in” to allow the
model to stabilize. For both values of Apc, all three variables fluctuate around
a mean. When Apc = 1, unemployment is high, wages are low, and profits are
high. When Apc = 3, unemployment is low, wages are high, and profits are low.
We now turn to a systematic analysis of how changing Apc a↵ects households
and firms according to several important variables.
Increasing Apc benefits both household and firms up to around Apc = 1. Af-
ter that, increasing Apc favors households over the interests of firms. When Apc
is low, unemployment is very high, as can be seen in Figure 3A. High unemploy-
ment correlates with low wages as seen in Figure 3B, and inequality between
households is very high as seen in the graph of the Gini coefficient in Figure 3E
(blue line). The extremely high inequality when Apc is low results from three
factors: (1) many households are unemployed, (2) even among those who are
employed, wages are very low, and (3) a very small number of households end
up owning the majority of the equity in firms and therefore collect the vast
majority of profits. As Apc increases from below 1 to slightly above 1, unem-
ployment drops rapidly and then slowly rises again with higher Apc, leveling o↵
at around 13%. This slow increase in unemployment with higher Apc is likely
due to more firms going out of business, as discussed in the next paragraph.
Wage rate increases rapidly with increasing Apc, leveling o↵at around 0.75.
This means that under conditions of high Apc, households are paid about 75%
of what they produce, the remaining 25% being kept as profits. Due to relatively
low unemployment and higher wages, the Gini coefficient for households drops
22All values used are: [0.2 0.4 0.6 0.8 1 1.1 1.2 1.3 1.4 1.6 1.8 2 3 4 6 8 10]
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

10 Kelter, Wilensky, Potvin
Fig. 2. Time series of un-
employment, mean wage
rate, and mean profits for
two values of Apc .
with increasing Apc and levels o↵around 0.42. To summarize, as Apc increases,
unemployment decreases rapidly and then slowly increases, wages increase, and
household inequality decreases.
The story is di↵erent for firms. They also benefit as Apc increases from below
1 to slightly above 1 as seen in the decreasing turnover rate (Figure 3C) and
increasing profits (Figure 3D). However, as Apc continues to increase, turnover
rates increase and profits decrease. As a result, inequality among firms increases
with increasing Apc (Figure 3E). The reason for these patterns is that when
Apc is low, even if demand increases, a firm cannot produce more due to the
limited amount of land. So, a firm will not hire workers away from other firms,
and competition between firms remains low. As Apc increases, firms can produce
more and therefore compete more for market share. Such competition decreases
their rates of profit and increases turnover rate. To summarize, firm profits and
turnover rate improve as Apc increases to around 1 when all households can be
productively employed but as Apc increases further, profits sag and turnover
increases.
4 Conclusion
The results of the model show a basic conflict between the interests of households
and firms regarding the availability of productive land. When there is a surplus
of labor compared to available land, wages are suppressed, and unemployment
is high. When labor and land availability approximately balance, unemployment
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

Introducing Land Constraints to Macroeconomic Agent-based Models 11
Fig. 3. The e↵ect of changing Apc on both
households and firms. The solid lines represent
the mean and the shaded area one standard
deviation. Households can each work one unit
of land per period, so Apc =1meansthat,in
theory, all households could be productively
employed. Wages and Profits are expressed as
a fraction of average price of goods to give a
measure of actual purchasing power.
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

12 Kelter, Wilensky, Potvin
is low, but wages are low as well. This is because firms have no incentive to
compete for workers since they do not have extra land with which to increase
production. Such a situation favors firms, resulting in low turnover rates and
high profits. This results in low inequality between firms but extreme inequality
among households, because over time a small number of households come to own
the large majority of firm equity and wages for remaining households are low.
As the available land increases, firms have an incentive to increase production
to compete for market share, leading them to compete for workers which drives
up wages. Inequality among households decreases while inequality and turnover
among firms increases.
The model presented here fills a gap in extant macroeconomic ABMs by con-
necting the productive capacity of the economy to the underlying productive
capacity of the Earth. As such, it lays the foundation for a number of future
dynamic modeling interactions between the economy and ecosystems. The im-
mediate next step is to model firms impacting the productivity of the land due to
either degenerative or regenerative farming practices as discussed in section 2.1.
Beyond this, we plan to model more complex economies with a supply network
of multiple types of firms and introduce indexed pricing schemes designed to
stabilize economic fluctuations and align free market incentives with sustained
ecological integrity as a passive emergent e↵ect.
5 Acknowledgements
Thank you to Jacob Wit and William Conboy for their contributions to the code
of the ABM presented in this paper.
References
1. Assenza, T., Delli Gatti, D. & Grazzini, J. Emergent Dynamics of a Macroe-
conomic Agent Based Model with Capital and Credit. Journal of Economic
Dynamics and Control 50, 5–28. issn: 01651889 (Jan. 2015).
2. Delli Gatti, D., Desiderio, S., Ga↵eo, E., Cirillo, P. & Gallegati, M. Macroe-
conomics from the Bottom-up isbn: 978-88-470-1971-3 (Springer Science &
Business Media, Apr. 2011).
3. Dosi, G., Fagiolo, G. & Roventini, A. An Evolutionary Model of Endoge-
nous Business Cycles. Computational Economics 27, 3–34. issn: 0927-7099,
1572-9974 (May 2006).
4. Hesselink, L. X. W. & Chappin, E. J. L. Adoption of Energy Efficient
Technologies by Households – Barriers, Policies and Agent-Based Modelling
Studies. Renewable and Sustainable Energy Reviews 99, 29–41. issn: 1364-
0321 (Jan. 2019).
5. Rai, V. & Robinson, S. A. Agent-Based Modeling of Energy Technology
Adoption: Empirical Integration of Social, Behavioral, Economic, and En-
vironmental Factors. Environmental Modelling & Software 70, 163–177.
issn: 1364-8152 (Aug. 2015).
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

Introducing Land Constraints to Macroeconomic Agent-based Models 13
6. Balint, T. et al. Complexity and the Economics of Climate Change: A
Survey and a Look Forward. Ecological Economics 138, 252–265. issn:
0921-8009 (Aug. 2017).
7. Ciarli, T. & Savona, M. Modelling the Evolution of Economic Structure
and Climate Change: A Review. Ecological Economics 158, 51–64. issn:
0921-8009 (Apr. 2019).
8. Coronese, M., Occelli, M., Lamperti, F. & Roventini, A. AgriLOVE: Agri-
culture, Land-Use and Technical Change in an Evolutionary, Agent-Based
Model SSRN Scholarly Paper. Rochester, NY, Oct. 2021.
9. Shahpari, S., Allison, J., Harrison, M. T. & Stanley, R. An Integrated
Economic, Environmental and Social Approach to Agricultural Land-Use
Planning. Land 10, 364. issn: 2073-445X (Apr. 2021).
10. Marshall, A. Principles of Economics Eighth. isbn: 978-1-343-83441-5 (Lon-
don: Macmillan, London, [1979] 1920).
11. Wilensky, U. NetLogo Center for Connected Learning and Computer-Based
Modeling, Northwestern University. Evanston, IL. 1999.
12. Dosi, G., Fagiolo, G. & Roventini, A. The Microfoundations of Business
Cycles: An Evolutionary, Multi-Agent Model. Journal of Evolutionary Eco-
nomics 18, 413–432. issn: 1432-1386 (Aug. 2008).
13. Wolf, S. et al. Describing Economic Agent-Based Models – Dahlem ABM
Documentation Guidelines. Complexity Economics 2, 63–74. issn: 2210-
4275, 2210-4283 (Jan. 2013).
14. Lengnick, M. Agent-Based Macroeconomics: A Baseline Model. Journal
of Economic Behavior & Organization 86, 102–120. issn: 01672681 (Feb.
2013).
15. Hait, A. W. The Majority of U.S. Businesses Have Fewer Than Five Em-
ployees https://www.census.gov/library/stories/2021/01/what-is-a-small-business.html.
2021.
Appendix - Simulation Parameters
The following is a complete list of the parameters used in the simulations re-
ported in this paper as they would appear in a NetLogo Behavior Space experi-
ment:
["setup-structure" "Single-PG&CG-TC=1.json"]
["LAND-AREA-PER-CAPITA" 0.2 0.4 0.6 0.8 1 1.1 1.2 1.3 1.4 1.6 1.8 2 3 4 6 8 10]
["MIN-WAGE-RATE" 2.5]
["DIMINISHING-UTILITY-CONSTANT" 0.6]
["pick-cheapest-firm?" false]
["delli-gatti-consumer-search?" false]
["N-TRADING-LINKS" 7]
["MONTHS-TO-LOWER-WAGE" 12]
["layoff-probability" 0.5]
["n-households" 1000]
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3

14 Kelter, Wilensky, Potvin
["n-firms" 60]
["SEARCH-N" 5]
["fix-n-framework-agreements?" false]
["PROB-REPLACE-FIRM-PRICE" 0.25]
["framework-duration" 24]
["MAX-PRICE-CHANGE" 0.2]
["index-in-use" "no index"]
["SEARCH-BETTER-JOB-PROB" 0.1]
["mean-new-agreements-per-month" 2]
["firm-memory-constant" 0.8]
["min-wage-80\%-of-tech-param?" false]
["STARTUP-LIQUIDITY" 100]
["primary-good-prod-function" "linear"]
["alpha" 1]
["transactions-per-month" 1]
["DESIRED-BUFFER-FRAC" 0.5]
["RES-WAGE-CHANGE" 0.9]
["N-FRAMEWORK-AGREEMENTS" 7]
["s" 0.1]
["BUFFER-LABOR-FRACTION" 0.3]
["firm-competency" 0]
["MAX-WAGE-CHANGE" 0.2]
["PROB-REPLACE-FIRM-QUANT" 0.5]
["BACKGROUND-IMPROVEMENT" "10"]
This is a preprint of the following chapter: Kelter, Wilensky & Potvin., Introducing Land Constraints to
Macroeconomic Agent-based Models, to be published in the Proceedings of the 2022 Conference of the
Computational Social Science Society of the America, edited by Zining Yang & Elizabeth von Briesen, 2023,
Springer Nature. The final version can be found at https://link.springer.com/chapter/10.1007/978-3-031-37553-8_3