PrintIn Lesson 1's reading we learned about some of the reasons why spatial data is special, including spatial autocorrelation, spatial dependence, spatial scale, and the ecological fallacy.
This week in our project we will look closely at another pitfall, the Modifiable Areal Unit Problem (MAUP).
The Modifiable Areal Unit Problem (MAUP)
Often, MAUP is considered to consist of two separate effects:
- A shape or zonation effect
- A scale or aggregation effect
Both effects are evident in the example in Figure 1.2 and further emphasized in Figure 1.3. The shape effect refers to the difference that may be observed in a statistic as a result of different zoning schemes at the same geographic scale. This is the difference between the 'north-south' and 'east-west' schemes. The scale or aggregation effect is observable in the difference between the original data and either of the two aggregation schemes.
Aggregation:
- combines smaller units into bigger units
- affects results!
- variance (Figure 1.3, left) decreases, although the mean stays the same.
MAUP is, if anything, more problematic than spatial autocorrelation. It is worth emphasizing just how serious the MAUP effect can be: in a 1979 paper, Openshaw and Taylor demonstrated by simulation that different aggregation (i.e., zoning) schemes could lead to variation in the apparent correlation between two variables from -1 to +1, in other words, the total range of variation possible in the correlation between two variables.
In practice, very little research has been done on how to cope with MAUP, even though the problem is very real. MAUP is familiar to politicians, who often seek to redistrict areas to their spatial advantage in a practice commonly referred to as "gerrymandering." In the practical work associated with this lesson, you will take a closer look at this issue in the context of redistricting in the United States.