Information Technology Reference
In-Depth Information
farmers always accept formal rules is thus misleading. On the other hand, many
farmers do follow formal rules, although it would be profitable not to do so and the
chance of being controlled and fined is very low in the UGB. In order to incorporate
this situation, the model considers a motivation to comply with formal rules, which is
independent of the risk of getting a penalty for non-compliance. Another issue is that
family farms and part-time farms which are mostly run with family labor have some
natural limit on labor capacity. The labor force of big farms which belong to land-
owners considering the land as capital is not restricted to family members. For those
farmers, constraints on the applied labor force may however still arise from other
sources like the availability of skilled workers or from the organizational structure of
a farm. In the models the labor intensity of various land-use patterns can hence
influence farmers' decision. Labor intensity also influences profits because family
labor is considered available at no cost, but labor force beyond family labor causes
costs. Several processes that are part of farmers' decision-making are identical in both
models. In each time step:
1.
For each farmer
f
a set of "considered options" is identified. The set includes those
options that are combinations of crops and technologies both known to
f
. Further,
unknown options are randomly considered, the probability increasing with usage of
an option by other farmers, what induces a self-reinforcing diffusion process.
2.
Based on
f
's land-use-pattern in the previous step
p
0
, a set
2
of "considered patterns"
p
i
is developed. Each
p
i
is created through (randomly) iterating three basic
operations on
p
0
several times (the original
p
0
is also considered further): a) add
one considered option which is not yet part of this pattern and associate some area
to it, forming a new land-use. The respective area is subtracted from a random
other land-use. b) Remove one land-use (if not the last), the respective area is
added to a random other land-use. c) Re-scale two land-uses: exchange some
random area between two land-uses.
3.
Land-use patterns
p
i
which are too different from
p
0
are discarded. "Difference" is
thereby evaluated based on two types of distances
d
between
p
0
and
p
i
: a) a
distance
d
s
related to skills representing uncertainty and learning efforts associated
with a change in the area of land-uses (especially implementation of new crops and
technologies) and b) distance
d
c
related to capital representing investments
necessary and capital loss arising from the change in the area of land-uses. Both
distances depend on the history of this farm because individual farms build up
stocks of irrigation technologies and accumulate knowledge on crops based on
their land-use decisions. For each land-use pattern
p
i
,
f
keeps
p
i
for further
consideration randomly with a probability
p = (1-d)
α
. This random filtering is done
for both types of distances, α
s
(skills) and α
c
(capital) being respective parameters
that can be adjusted to explore the influence on model behaviour. That is, those
patterns that are "close" enough to
f
's current land-use in both regards are likely
kept for further consideration while more distant
p
i
are likely discarded.
4.
Consequences arising from regulations (subsidies, penalties, legality) are evaluated
for each of the remaining patterns.
2
In the implementations used in this article for each farmer 1000 "considered patterns" are
developed in each time-step.
