# 1.7: Cause and Effect

$$\newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} }$$

$$\newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}}$$

$$\newcommand{\id}{\mathrm{id}}$$ $$\newcommand{\Span}{\mathrm{span}}$$

( \newcommand{\kernel}{\mathrm{null}\,}\) $$\newcommand{\range}{\mathrm{range}\,}$$

$$\newcommand{\RealPart}{\mathrm{Re}}$$ $$\newcommand{\ImaginaryPart}{\mathrm{Im}}$$

$$\newcommand{\Argument}{\mathrm{Arg}}$$ $$\newcommand{\norm}[1]{\| #1 \|}$$

$$\newcommand{\inner}[2]{\langle #1, #2 \rangle}$$

$$\newcommand{\Span}{\mathrm{span}}$$

$$\newcommand{\id}{\mathrm{id}}$$

$$\newcommand{\Span}{\mathrm{span}}$$

$$\newcommand{\kernel}{\mathrm{null}\,}$$

$$\newcommand{\range}{\mathrm{range}\,}$$

$$\newcommand{\RealPart}{\mathrm{Re}}$$

$$\newcommand{\ImaginaryPart}{\mathrm{Im}}$$

$$\newcommand{\Argument}{\mathrm{Arg}}$$

$$\newcommand{\norm}[1]{\| #1 \|}$$

$$\newcommand{\inner}[2]{\langle #1, #2 \rangle}$$

$$\newcommand{\Span}{\mathrm{span}}$$ $$\newcommand{\AA}{\unicode[.8,0]{x212B}}$$

$$\newcommand{\vectorA}[1]{\vec{#1}} % arrow$$

$$\newcommand{\vectorAt}[1]{\vec{\text{#1}}} % arrow$$

$$\newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} }$$

$$\newcommand{\vectorC}[1]{\textbf{#1}}$$

$$\newcommand{\vectorD}[1]{\overrightarrow{#1}}$$

$$\newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}}$$

$$\newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}}$$

$$\newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} }$$

$$\newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}}$$

$$\newcommand{\avec}{\mathbf a}$$ $$\newcommand{\bvec}{\mathbf b}$$ $$\newcommand{\cvec}{\mathbf c}$$ $$\newcommand{\dvec}{\mathbf d}$$ $$\newcommand{\dtil}{\widetilde{\mathbf d}}$$ $$\newcommand{\evec}{\mathbf e}$$ $$\newcommand{\fvec}{\mathbf f}$$ $$\newcommand{\nvec}{\mathbf n}$$ $$\newcommand{\pvec}{\mathbf p}$$ $$\newcommand{\qvec}{\mathbf q}$$ $$\newcommand{\svec}{\mathbf s}$$ $$\newcommand{\tvec}{\mathbf t}$$ $$\newcommand{\uvec}{\mathbf u}$$ $$\newcommand{\vvec}{\mathbf v}$$ $$\newcommand{\wvec}{\mathbf w}$$ $$\newcommand{\xvec}{\mathbf x}$$ $$\newcommand{\yvec}{\mathbf y}$$ $$\newcommand{\zvec}{\mathbf z}$$ $$\newcommand{\rvec}{\mathbf r}$$ $$\newcommand{\mvec}{\mathbf m}$$ $$\newcommand{\zerovec}{\mathbf 0}$$ $$\newcommand{\onevec}{\mathbf 1}$$ $$\newcommand{\real}{\mathbb R}$$ $$\newcommand{\twovec}[2]{\left[\begin{array}{r}#1 \\ #2 \end{array}\right]}$$ $$\newcommand{\ctwovec}[2]{\left[\begin{array}{c}#1 \\ #2 \end{array}\right]}$$ $$\newcommand{\threevec}[3]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \end{array}\right]}$$ $$\newcommand{\cthreevec}[3]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \end{array}\right]}$$ $$\newcommand{\fourvec}[4]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}$$ $$\newcommand{\cfourvec}[4]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}$$ $$\newcommand{\fivevec}[5]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}$$ $$\newcommand{\cfivevec}[5]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}$$ $$\newcommand{\mattwo}[4]{\left[\begin{array}{rr}#1 \amp #2 \\ #3 \amp #4 \\ \end{array}\right]}$$ $$\newcommand{\laspan}[1]{\text{Span}\{#1\}}$$ $$\newcommand{\bcal}{\cal B}$$ $$\newcommand{\ccal}{\cal C}$$ $$\newcommand{\scal}{\cal S}$$ $$\newcommand{\wcal}{\cal W}$$ $$\newcommand{\ecal}{\cal E}$$ $$\newcommand{\coords}[2]{\left\{#1\right\}_{#2}}$$ $$\newcommand{\gray}[1]{\color{gray}{#1}}$$ $$\newcommand{\lgray}[1]{\color{lightgray}{#1}}$$ $$\newcommand{\rank}{\operatorname{rank}}$$ $$\newcommand{\row}{\text{Row}}$$ $$\newcommand{\col}{\text{Col}}$$ $$\renewcommand{\row}{\text{Row}}$$ $$\newcommand{\nul}{\text{Nul}}$$ $$\newcommand{\var}{\text{Var}}$$ $$\newcommand{\corr}{\text{corr}}$$ $$\newcommand{\len}[1]{\left|#1\right|}$$ $$\newcommand{\bbar}{\overline{\bvec}}$$ $$\newcommand{\bhat}{\widehat{\bvec}}$$ $$\newcommand{\bperp}{\bvec^\perp}$$ $$\newcommand{\xhat}{\widehat{\xvec}}$$ $$\newcommand{\vhat}{\widehat{\vvec}}$$ $$\newcommand{\uhat}{\widehat{\uvec}}$$ $$\newcommand{\what}{\widehat{\wvec}}$$ $$\newcommand{\Sighat}{\widehat{\Sigma}}$$ $$\newcommand{\lt}{<}$$ $$\newcommand{\gt}{>}$$ $$\newcommand{\amp}{&}$$ $$\definecolor{fillinmathshade}{gray}{0.9}$$

# Cause and Effect

All scientists look for regularities and patterns while trying to figure out the causal forces behind them. One of the main points of science is that by understanding why things happen the way they do, we may better understand how to prevent what we dislike or maintain what we love. Geographers are no different, but we tend to look for regularities in the processes that create or destroy patterns on the landscape. Sometimes geographers simply observe phenomena, and try to make sense of it. Other times, geographers use maps to plot that which they observe, or are trying to observe. Once the map is constructed, then we begin to try to identify what is causing the pattern on the map.

 Want to know more? Check out the following links! News Magazine article about Hip-Hop music in Iran

## Cultural Ecology

One of the oldest strategies geographers use to seek causality is to carefully consider the role of the physical environment in which phenomena occur. Geographers ask, “What is the effect of climate, weather, water, rocks, soils, etc. on politics, religion, language, culture, etc.?” The study of the interaction between various cultural practices and the physical environment is referred to as cultural ecology. Sometimes the relationships between cultural practices and the natural environment are easy to identify, such as the relationship between the sport of ice hockey and cold winters. More often though, relationships between cultural practices and the environment are difficult to fully understand. Recall the discussion about Environmental Determinism from earlier in the chapter.

## Cultural Integration

Often there is no identifiable causal relationship between cultural practices and the environment, so geographers check for interrelationships between various cultural practices within a place or region. An understanding that many ideas, practices, and traditions from one cultural realm (religion, e.g.) may have an effect on other cultural practices (politics, e.g.) is the basic principle of the concept of cultural integration. All cultural practices function like gears in a complex machine where movement by one gear, wheel or spring is likely to turn other gears, wheels and springs elsewhere So, weather, religion, politics, language, health, and crime all affect each other.

## Plankton & Obama

Figure In the map on the left, the red line indicates the approximate shoreline of the ancient Gulf of Mexico responsible for the rich soil in the region today. On the right is an overlay map showing both intense cotton production during the 1800s (dots) and the election results of 2008 by county. Blue counties went for Obama and Red for McCain.

An excellent example of the complex nature of a culturally integrated system was uncovered and explained by clever geographer-types during the presidential elections of 2008 and 2012. The stunning finding of this study was that long-dead plankton helped Barack Obama get elected! Preposterous? Not to geographers. Consider that more than 100 million years ago, the shoreline of the Atlantic Ocean/Gulf of Mexico was far inland from where it is today. In the shallow seas that once covered what are today parts of Alabama, Mississippi, Georgia, and the Carolinas, there floated trillions of tiny plankton that upon dying, created rich, chalky soils along ancient beaches (the red line in the map below). This dirt, thousands of years later became a key element in a narrow region of agriculturally rich soils stretching across the American South, dividing a region where the soils are otherwise poor. Because these rich soils were very dark, farmers labeled the region the Black Belt. During the 1800s, people also called this area the Cotton Belt, because it had become the richest cotton farming region in the United States. Because African-Americans, enslaved on plantations grew most of the cotton, the term, black belt, acquired a new layer of meaning that it maintains today. In those counties, where the descendants of slaves still outnumber white people by a large margin, Barack Obama won more votes than his challengers John McCain and Mitt Romney.

## Process and Pattern

Geographers, especially in the last 40 years or so, have sought to not only explain why patterns emerge on the landscape but also to make predictions about when and how they will change. Geographers who make predictions based on evidentiary trends in data are, by their actions, scientists. One of the first steps in spatial science projects involves identifying clustering within spatial patterns evident either on maps or the landscape. Many phenomena cluster in space because the friction of distance affects everything, as suggested by the First Law of Geography. One way to verify the clustering of some phenomena is to plot data on a map with points, as is done in the accompanying figures below.

 Want to know more? Check out the following links! A website dedicated to teaching and learning spatial concepts.

Figure Map- Fort Lewis , WA. This map shows significant clustering of payday lenders near the gates of this military facility in 2003. The map helped convince legislators of problems with the payday lending industry.

In the map below, payday lenders, represented by red triangles, appear to cluster in large numbers near the entrance of McChord Air Force Base/Fort Lewis in Washington. The pattern on the map strongly suggests that the payday lending industry focused on military personnel as an attractive target demographic. Representatives of the payday lending industry denied they were targeting soldiers and sailors before Congress, but the intense clustering of payday loan stores near military bases across the U.S. helped convince legislators in both Washington D.C. and Seattle, Washington that this industry was indeed targeting the military, and that stricter regulation of the payday lending industry was necessary to protect service members from the often-dangerous effects of short-term, high-cost loans.

Figure 1-12: The pattern of phenomena on the landscape may help us understand causality. Clustered, random and dispersed patterns can be measured stratistically by GIS software.

Often, it is easy to visually identify the clustering of points on a map with, as is the case in the map above. Sometimes though, it is not so easy to determine if points are clustering beyond what might happen randomly, or if a pattern of points is more clustered than another.

When patterns are difficult to see, geographers turn to spatial statistics to measure the distribution of points in space. One cool technique, known as nearest neighbor analysis, compares the average distance between actual points on a map against an equal number of hypothetical points randomly dispersed on the same map. GIS software then compares the two patterns and calculates the likelihood that the distribution of the actual (observed) points is more (or less) clustered than a random distribution of the same points. See figures below:

Figure GIS Software output window indicating a statistically significant clustering of points. Clusters of crime, disease or businesses provide decision makers with important tools.

Random, clustered and dispersed patterns each tell a story about the processes that are acting upon the mapped phenomena. The image in Figure 1:14 is a graphical report from a statistical test of payday lenders in Los Angeles’ San Fernando Valley. As you can see, the level of clustering of this type of business is far greater than a randomly distributed set of points in the same space. Therefore, we can be quite certain that the locations of payday loan shops are not due to random processes. Clustered patterns of diseases, crimes, tornados, or any other phenomena, are of great interest because clustering helps us identify causes, and hopefully, solutions to problems. Highly dispersed point patterns are nearly as interesting. Dispersion suggests that for some reason points are being pushed away from one another. Schools, Fire Stations, and 7-Eleven stores would all likely show spatial dispersion.

Regions also may exhibit clustering. You can observe clustering on a choropleth map when data values associated with neighboring polygons (e.g. states) are more similar than the data values associated with non-adjacent polygons. Put more simply, clustering is evident when neighboring regions are more similar than non-neighboring regions. A special term, spatial autocorrelation is often applied to such patterns. Like the point clustering patterns discussed above, spatial statistics are available to geographers seeking to measure the degree of clustering, dispersion or randomness in a choropleth (polygon) map. Moran’s I is a common statistic used to measure spatial autocorrelation, or clustering, on a choropleth map.

FIgure Electoral maps like the one above show clustering of like states. The Moran’s I value would indicate significant clustering of like values.

Figure Graphic - This series of checkerboard images represents various levels of spatial autocorrelation for polygonal features (regions). The Moran's I statistic could be used to determine the degree of clustering of spatial autocorrelation.

The well-known election map of 2000 shows a clear pattern of clustering that does not need statistical analysis to notice. However, if you wanted to compare the degree of clustering evident in that map against another election map, or against a map of something unrelated, like cancer rates, you would need a statistical tool. Using Moran’s I, you could determine which pattern was more clustered; or if you were comparing voting patterns over many years, trends could be analyzed, perhaps allowing you to make predictions about future elections.

Figure Scatter Plot diagram - This diagram displays the same data in the map below. The correlation coefficient is r=.55,indicating a moderately strong relationship.

## Co-location

Clustering is a type of co-location. When things, behaviors or ideas (e.g. factories, payday lenders, auto dealerships) cluster, that is a form of co-location called agglomeration. Co-location can also characterize the situation in which seemingly different things, behaviors or ideas are found in the same location. For example, Evangelical Christians and Trump Voters are found in the same states. Payday lenders and military personnel cluster in the same towns; and night clubs and college students are often found in the same neighborhoods.

When co-location among phenomena occurs then there exists a spatial relationship between the phenomena. Occasionally, persistent co-location indicates a causal relationship; where one thing in a location (an increase in air pollution) causes the other thing (an increase in lung cancer) to occur in the same location. Uncovering, measuring and explaining causal relationships is a major goal of geographers.

Figure 1-17: These maps represent income inequality (left) and murder rate (right) by states. You can see that about half the time, as one variable increases, so does the other. The correlation coefficient is r=.55.

Hypothetically, you may notice that there seems to be an unusual number of obese people living in neighborhoods where there are also many fast-food restaurants. If you were a geographer, you might hypothesize that living near fast-food restaurants increases residents’ chances of gaining weight. You could test this hypothesis by collecting data from the local health department. Next, you could map the obesity rate by neighborhood (census tract or ZIP code perhaps). Then, you could map all the fast-food restaurants in each neighborhood so that a count of fast food outlets per neighborhood was possible. Then you could run statistical tests on the data to test your hypothesis.

Figure Los Angeles - The map on the left displays the density of fast food outlets per business. A regression model was used to analyze the relationship of fast food availability to the percentage of children rating “ healthy” on school fitness test. The map on the right shows the location where the regression model under-predicted to over-predicted healthy levels among school aged children in each Zip code. Other variables like income, ethnicity etc. were beheld constant by the regression model.

Often, geographers begin analyzing relationships by testing for the degree of correlation between variables (e.g., fast food vs. obesity rates per ZIP code), using a test statistic like Pearson’s Correlation Coefficient. This test, and others like it, measure the amount of covariance, or dependence, between two variables. Put simply, correlation tests report how much one variable (like obesity rates) rises or falls as a second variable (restaurant density) rises and/or falls. You might find that as the density of fast-food restaurants goes up in neighborhoods around town, so do obesity rates in those neighborhoods, indicating a positive correlation. Negative correlations are possible too. You may find that as the miles of bike paths per city increases, the obesity rate goes down. Often you find that there’s not much correlation at all. If you find strong negative or positive correlations between two variables in space, then you may have grounds to argue there is a causal relationship.

 Want to know more? Check out the following links! GIS PopSci.org Research and Instructional Materials for those interested in spatial statistics.

Unfortunately, correlation can be misleading. It’s easy to accidentally misinterpret correlation statistics. You can mix up the direction of causality: maybe the fast-food restaurants are in certain neighborhoods because their owners build them where the population is known to love french fries. Maybe the density of fast-food restaurants and obesity rates are caused by a third less obvious thing, like poverty. These unknown variables are called confounding factors. Maybe the factors are completely unrelated, exhibiting a relationship that is purely by chance. These random relationships are known as spurious relationships and happen when phenomena rise and fall together but are causally unrelated.

The nature of relationships between cause and effect variables is most often measured using regression analysis, a more complex statistical technique used by geographers to determine the strength and direction of causality between a dependent (effect) and one or more independent (causal) variables. So, for example, regression would help you understand not only if having lots of fast food joints in a neighborhood had an effect on obesity, but it would also allow you to better analyze the effect of ethnicity, income, access to parks, etc. on the dependent variable (obesity). Regression analysis, done with GIS software, also allows geographers to see quickly where trends are well predicted by the variables used in the analysis, and where there is less or more of the predicted variable.

No doubt, you have seen patterns on the landscape and wondered, “Why is that there?” The chapters that follow should help you answer those questions. Some of the techniques may seem challenging to you, but college students who have been exposed to geography should be able to observe patterns, ask questions about the observed patterns, and do a few basic analyses on data. Geographers have effective techniques for answering questions and solving problems. It is the major goal of this text to expose students to some of these techniques.