Friday, November 18, 2011

The Modifiable Areal Unit Problem and Administrative Proliferation in Indonesia

Today I'm going to make a modestly ambitious attempt to link together a statistical/theoretical concept found in most geography textbooks with a curious political occurrence here in Indonesia.  I'm going to do this because it's Friday night and there's nothing else to do in Sungai Penuh.  Good Times!

The Big MAUP

Our big concept for the day is deceptively obvious: the conclusions you draw are often variable based on how you classify your data.  For spatial data, which is of great interest to geographers, zones of analysis affect results.  In other words, the way you interpret a particular landscape depends on where you draw the borders.   Geographers call this the Modifiable Areal Unit Problem (MAUP).  Heywood (1998) defined this as "a problem arising from the imposition of artificial units of spatial reporting on continuous geographical phenomena resulting in the generation of artificial spatial patterns".  There are two aspects to the MAUP:

1)  The scale effect:  This refers to the variation in results that occurs due to the number of zones used in analysis.  For example, when thinking about average income in the United States, you can look at the country as a whole, you can look at particular regions (e.g. the Midwest), or you can look at smaller territorial divisions, including states, counties/parishes, on down to neighborhoods.  Not all problems we deal with in geography come with ready-made boundaries, though, and sometimes you have to make your own units.  You have to be careful in selecting your units because if they are too small, the population measured is so small that your results aren't reliable or generalizable.  On the other hand, if the size of the unit is too big you risk losing meaningful geographic variation that is demonstrated with smaller units.  In our previous example, if we just look at the average income of the US as a whole, we won't be able to understand that some places have higher incomes than others.  But if we make our unit of analysis too small, we risk losing site of the bigger picture (and we'd probably have way too much data).  We'll come back to this very important point later, but first let's have a look at some of the statistical effects.  Have a look at the figure below.  I had the crack graphics team here at HQ put this one together.

As you can see, I've got two views of the same block.  In the first one we've created a grid of 16 cells to examine a particular variable, like average income or age.  In the second block we've divided the exact same area, but with a grid of 4 cells.  As you can see, the first is much more detailed.  But in addition to this, we can notice that the standard deviation (SD in my graphic) and variance are different!  This is pretty important for understanding statistical information.  The standard deviation tells you how spread out your numbers are.  The higher the standard variation, the greater the spread of individual numbers off the average.  The variance is a little more esoteric, but you have to have that to get the standard deviation.  The variance is the average of the squared difference of each individual observation from the average.  So the 16-cell grid gives us a lot more detail and helps us to understand the spatial distribution.  But the 4-cell grid is easier to compute. 

From Penn State Geography
2)  The zonation effect:  This aspect of the MAUP refers to where you draw your lines.  The way you divide spatial data almost always changes the picture.  For example, have a look at the diagram to the left that I shamelessly stole from the Penn State geography department's website.  Obviously the thing we're interested in the occurrence of little yellow dots.  We want to know how many there are and how they are distributed.  We indicate the density of the dots by using different colors (2).  However, as you can clearly see, the recorded occurrence of little yellow dots is completely dependent on how we draw our lines.  This lines may be based on topography, on political/administrative boundaries, or anything else.  As I mentioned previously, sometimes the lines are already provided, like if you are looking at discrete administrative units.  But sometimes you have to draw them yourself, and so you have to bear in mind the zonation effect.

Pemekaran Daerah di Indonesia  

"Pemekaran" refers to administrative proliferation in Indonesia.  In a previous post I described the 5 basic levels of government here; they range from the national government, down to the provinces, districts (kabupatens), sub-districts (kecamatan), and village (desa) (3).  After the end of the centralized Suharto regime in 1998, the government granted lots of new powers to the districts as part of a massive raft of government reforms.  One of the reforms associated with this decentralization was a new law that allowed for the creation of new districts and provinces.  There were many justifications for this, including improving efficiency and representativeness of district and provincial governments.

Though the idea was a good one, pemekaran has spun out of control.  Local power brokers have used the law to create scores of new districts.  Before the end of the Suharto government in 1998 there were 26 provinces, 66 administrative municipalities, and 256 districts.  By 2009 there were 7 new provinces, 153 new districts, and 31 administrative municipalities.  In addition to the legitimate reasons for creating a new district or province, there are lots of shady ones as well.  New districts create an official base of power for local elites.  The district headman (bupati) has a tremendous amount of power which can be exercised to make lots and lots of money.  New districts create new civil servant positions, which the bupati can sell (not legally).  New districts also need new capital cities, which means big chunks of cash from the central government. 

So what does the MAUP have to do with all this?  Let's think about it from the perspective of the two aspects of the MAUP discussed above.  From the perspective of the scale effect, pemekaran in a way creates a "reverse-MAUP".  If we think about the "variable" under study as government services and numbers of civil servants, it turns out that the number of civil servants is in part determined by the number of units you have.  Moreover, more districts means more bupatis, which, to put it bluntly, means more corruption.  In addition, physical infrastructure, like district office buildings, have to be constructed in every single district, regardless of the size of the district.  Now it is true that bringing the government closer to the people may enhance the efficiency of government services, but this has to be balanced against the costs.  In many cases larger enterprises are able to do things more efficiently than smaller ones.  This is referred to as economies of scale.  Think about an automotive factory.  It doesn't make much sense to have a factory that makes only 50 cars a year, because each one of the cars would have to be really expensive to cover the cost of all the expensive machinery required to build a car.  But if you build 5,000,000 cars a year, the expenses can be spread out to a much greater degree, allowing you to sell the cars cheaper.  In some ways government works in the same way.  So even though you may be bringing the services closer to the people, if the loss of economies of scale outweighs the increase in services, then everyone is worse off.

On the other hand, we can also see the zonation effect at work.  Drawing new lines changes the apparent distribution of all sorts of things, including poverty and educational levels.  In some cases the new district has a much higher poverty rate than the old district.  While this creates the illusion of more poverty, it has the very real effect of splitting the needy area from the area of relative abundance.  From an administrative perspective, though the numbers haven't changed, the capacity and political will to address problems of poverty have decreased.  People in the richer area are reluctant (and not legally obligated) to provide money through taxes which fund social programs to people in the poorer area.  It also turns out that new districts, as a rule, tend to lag behind the older districts in terms of economic development.  Separating districts might also change the way people relate to one another as well.  For example, the town where I live, Sungai Penuh, has been pemekaraned off from the district of Kerinci.  Now people tend to see themselves as city people or country people.  There is a very low level of cooperation between the two new governments as well. 

So there you have it: the modifiable areal unit problem and the problems of local governance in Indonesia.  They are two sides of the same coin.  Even though pemekaran was meant to solve problems, it has in many cases created a whole new set of difficulties.  Unfortunately the folks benefiting from administrative proliferation aren't the ones that are going to foot the bill in the long run.   


(1)  To get the variance you find the difference of each individual observation.  Then square the differences.  Then average all the squares, and you have the variance.  To get the standard deviation, simply take the square root of the variance.  Or you can simply use the simple statistical calculator found here

(2)  Geographers refer to maps using different colors to show variability as chloropleth maps.  Maps that show density are called dasymetric maps.

(3)  Many cities are administratively independent and are at a level in the hierarchy parallel to the districts.  The cities are called kota administratif.  In urban areas the desa is usually replaced by the sub-sub district, called the kelurahan

References and For Further Reading

Dark, Shawna and Danielle Bram.  2007.  The Modifiable Areal Unit Problem (MAUP) in Physical Geography.  Progress in Physical Geography 31:5 pp471-479

Goodchild, MF, L Anselin and U Deichmann.  1993.  A Framework for the Areal Interpolation of Socioeconomic Data.  Development and Planning A 25, pp383-397

Jelinski, Dennis, and Jianguo Wu.  1996.  The Modifiable Areal Unit Problem and Implications for Landscape Ecology.  Landscape Ecology 11:3 pp129-140

1 comment:

  1. Expansion, not only created small kings, but also small areas which eventually creates also the small character officials, the success of the roots grow back "Divide et impera"