Where is precedent set? An exploratory geovisualization of State Supreme Court cases

Abstract

Vast data warehouses of legal documents and court decisions present an opportunity for data visualization and analysis, yet court decisions have rarely been visualized geographically. We explore the potential for geographic visualization of judicial case data by mapping a state-level standardized score calculated from 135 State Supreme Court abortion cases published in the USA between 1973 and 2013, and again from 228 redistricting cases published between 1962 and 2013. We observe substantial variation in Z-scores across the USA, with geographic patterning of high values observed for redistricting cases, but not for abortion cases. The resulting maps suggest that the geovisualization of court decisions may aid the generation and testing of hypotheses about whether or not certain states disproportionately set legal precedent and subsequently influence the national discourse on a given issue. This type of geovisualization may also have implications for forum shopping and other geographically explicit legal strategies.

Keywords:

• visualization
• State Supreme Court
• abortion
• redistricting

1. Introduction and review

In the current era of Big Data, vast electronic databases of legal documents present opportunities for new approaches to the visualization and analysis of US case law. Recent efforts have focused on visualizing the results of data mining (Gaur, 2011) and concept searches (Uijttenbroek, Klein, Lodder, & Harmelen, 2007), while others are exploring ways to visualize legal data such as citation networks (Bommarito, Katz, Zelner, & Fowler, 2010) and categories of court cases (Hook, 2007), and temporally visualizing case histories (C. Harris, Allen, Plaisant, & Shneiderman, 1999). The University of Baltimore School of Law is developing new software to plot relationships between various court opinions (University of Baltimore School of Law, 2013), and Ravel has grown into a popular online legal search, analytics, and visualization platform (Ravel, 2014). Yet, despite a modest literature in the geography of law within the social sciences (e.g. Economides, Blacksell, & Watkins, 1986; Taylor, 2006), there have been very few, if any, efforts to map case decisions geographically. In this paper, we take an exploratory approach to putting published State Supreme Court (SSC) decisions on the map for two contemporary issues: redistricting and abortion. Our intent is to create a starting point for further analysis of the geographic origin of published cases. We chose redistricting and abortion cases because they represent the kinds of issues that have reshaped national politics and divided our nation, and because each had a defining ruling by the U.S. Supreme Court in the latter half of the twentieth century that provided a nice bookend for the analysis.

On 26 March 1962, the U.S. Supreme Court ruling in Baker v. Carr determined that in issues of redistricting or reapportionment (the attempt to change the delineation of voting districts), federal courts may intervene to resolve these cases. This decision was the forbearer of the ‘one person, one vote’ standard for redistricting and the Voting Rights Act of 1965, and fundamentally changed the political representation process in the USA. Nearly every state was forced to redistrict during the 1960s, and our country experienced a subsequent shift in political power from rural to urban areas, an effect that is still felt in today's so-called culture wars (Loomis & Chanay, 2008). Similarly, on 22 January 1973, the U.S. Supreme Court handed down the Roe v. Wade decision which established a woman's right to privacy under the due process clause of the 14th Amendment. It further recognized that a person has a right to abortion until ‘viability’ or the ‘potential to live outside the mother's womb, albeit with artificial aid’ (Wood & Hawkins, 1980). This decision restricted numerous state and federal abortion regulations, and is still a divisive issue in modern America.

While both Baker v. Carr and Roe v. Wade sparked national debates that remain part of today's national political discourse, not all US states have played an equal role in the setting and interpretation of case law on these issues since these landmark rulings. At the heart of the American common law system is the notion of stare decisis, or the doctrine of precedent, in which courts create law that becomes binding on future decisions. Caldeira (1985) notes that while political scientists have not traditionally exploited the role of proximity on politics, legal precedent has a geographic component; neighboring states are likely to share social, political, and cultural values, and justices are less likely to invoke precedents from other states with increasing physical (or psychological) distance from the home jurisdiction. Caldeira (1985) demonstrates that judicial communication between SSCs (i.e. ‘who cites whom’) is influenced by a number of characteristics including an inverse relationship with the distance in miles between state capitols, while Harris (1985) found that the strongest predictor of judicial communication was interstate migration. The role of precedent is powerful, as SSC decisions can be binding on federal courts as well, as long as there is no conflict with the U.S. Constitution. Geographic patterns of case origin were also identified in three categories of landmark U.S. Supreme Court cases: redistricting, abortion, and environment (Brunn, Shelley, Webster, & Ahmed, 2000).

We used a geographic information system to manage and map published SSC cases in order to visualize states’ relative influence on the issues of redistricting and abortion. Our approach is rooted in the emerging tradition of geovisualization, which concerns the visual exploration, analysis, integration, and presentation of geographic data (Dykes, MacEachren, & Kraak, 2005), and which has been the subject of a number of new volumes over the last decade, particularly as related to the spatial sciences (Kraak & MacEachren, 2005; Moore & Drecki, 2012). We theorize that the geovisualization of court decisions may offer a preliminary understanding of the extent to which states disproportionately set legal precedent and subsequently influence the national discourse on a given issue. We are interested in two hypotheses: that the distribution of published SSC cases by state will be disproportionate to state populations, and that these discrepancies will exhibit geographic patterning.

2. Methods

Our approach involves filtering appropriate court cases by state from a commercial database, calculating a standardized Z-score for each state scaled by an appropriate state population benchmark, and then mapping the Z-scores in a style typically used to map Electoral College delegate counts during the national election season.

We used the LexisNexis^® Academic online research database as the source of eligible published court cases. We narrowed the search by selecting ‘Federal & State Cases’ from the US Legal tab on the main search page, and performed a search on the keyword ‘abortion’ within ‘legal topics’. We used the date parameter to constrain the search to decisions occurring after 22 January 1973 (the date of the Roe v. Wade decision), and set the jurisdiction to ‘All States Highest Courts’ in order to see results from SSCs. This search, conducted in November 2013, yielded a total of 649 eligible cases, but these initial search results contained a high number of cases that incidentally included the word ‘abortion’ in the case summary and were not about abortion issues. We proceeded to manually filter out actual abortion cases, which we defined as those relating to parental consent, public funding for abortion procedures, or that questioned the legality of abortions, thus yielding a final count of 135 cases. We replicated this process for redistricting cases by searching on the keyword ‘redistricting’ and constraining the date to after 26 March 1962 (the date of the Baker v. Carr decision). This search (also conducted in November 2013) initially yielded 306 eligible cases, which we subsequently filtered down to 228 cases that specifically concerned elections. Cases dealing with the reapportionment of school districts were beyond the scope of this project and excluded.

An important caveat of the LexisNexis^® data is that not all cases are required by law to be published. States have different legal criteria for case publication, and the individual desires of the case parties also influence whether or not a particular case is published. Generally speaking, published cases are those that set some legal precedent or new interpretation of the law and thus are worthy of publication, so while published cases do represent a sample of all cases heard on a particular subject, they likely represent a majority of the important cases.

To visualize the proportion of SSC cases (redistricting or abortion) in each state relative to all SSC cases across the nation, we compute a standardized score similar to a location quotient (LQ) for each state. The LQ is a ratio that compares an attribute of interest (i.e. cases) for an area to a larger reference region based on some baseline characteristic, and is a classic measure of spatial concentration (Haggett, 1965; Isard, 1960). Because the LQ has no fixed upper bound, its distribution is often right-skewed, and despite efforts to construct confidence intervals and objective cut-off values for LQ values (Moineddin, Beyene, & Boyle, 2003; Tian, 2013), it ignores absolute differences between observed and expected counts, does not yield traditional measures of statistical significance, and can be problematic for further spatial analysis.

First, we calculate expected cases by state using total population as the baseline characteristic, just as we would for an LQ, with the explicit assumption that a state with a higher population will have proportionately more people engaged in behavior that could lead to a case reaching that state's Supreme Court, so that ultimately the number of SSC cases for a given issue should be in proportion to that state's population relative to all other states. But instead of calculating an LQ, we recognize that SSC cases are discrete, rare events which follow a Poisson distribution, and by treating the case counts as a Poisson variable, the mean and variance for each state is equal to the expected case count. Hence, we calculate, as a descriptive measure of each state's proportion of observed vs. expected SSC cases, a standardized Z-score for each state using the formula in Equation (1):(1)

where x_i is the number of cases in state i, and μ_i is the number of expected cases in state i given that state's population. If Z > 0, the number of cases is higher than expected for the given state, and if Z < 0, the number of cases is lower than expected given the state's population. We compute state Z-scores and P-values for the number of abortion and redistricting cases separately, using the average decennial state populations over the study time frame as the baseline for the expected counts. For the abortion analysis, we computed the average state populations as the average of the values from each of the 1970–2010 US censuses, and for the redistricting analysis, we computed an average of the 1960–2010 US censuses.

Next, we evaluate whether there is a difference between the observed and expected number of cases for at least one state. We compute the P-values for the number of abortion and redistricting cases separately, using the observed and expected counts from the Z-score calculations. Here, we perform separate hypothesis tests for the one-sided alternative hypotheses. To test the hypothesis that the actual number of cases is more than the expected number of cases for at least one state at a significance level of 5% for the family of 50 hypothesis tests, we first calculate for each state the P-value using the cumulative Poisson probability formula (Equation (2)):(2)

where k is the number of cases. Since we have 50 multiple comparisons, and we want an overall significance level of 5%, we then assess significance for each state at α = 0.05/50 = 0.001. This adjustment to α, known as the Bonferroni correction, will give a familywise Type I error rate, and hence a significance level for the family of 50 hypothesis tests, of just under 5%. Similarly, to test the hypothesis that the actual number of cases is less than the expected number of cases for at least one state at the 5% significance level, we use Equation (3) for each state,(3)

to find the P-value, and then assess significance for each state at α = 0.05/50 = 0.001.

Finally, the Z-scores were mapped using ArcGIS 10.2 (ESRI, Redlands, CA), and we assess spatial autocorrelation of the Z-scores for both abortion and redistricting using the Moran's I and LISA statistics (Anselin, 1995) for the conterminous 48 states. The authors declare no conflicts of interest that would influence this research.

3. Results

The computed Z-scores and Poisson P-values for abortion and redistricting are presented by state in Table 1 along with the observed and expected case counts, and average state populations. The Z-scores for each state are also presented in the Main Map.

Table 1. Published SSC cases related to abortion and redistricting, with mean population during the study period, expected and actual cases, Z-score, and Poisson P-value for each alternative hypothesis.

		Abortion, 1973–2013						Redistricting, 1962–2013
						P-value						P-value
State		Mean pop.	Expected cases	Actual cases	Z-score	Actual < expected	Actual > expected	Mean pop.	Expected cases	Actual cases	Z-score	Actual < expected	Actual > expected
Alabama	AL	4,121,084	2.20	20	12.01		<.001	3,978,736	3.77	5	0.63		.326
Alaska	AK	518,212	0.28	5	8.98		<.001	469,510	0.44	11	15.83		<.001
Arizona	AZ	3,935,541	2.10	1	−0.76	.380		3,496,618	3.31	1	−1.27	0.157
Arkansas	AR	2,429,874	1.30	4	2.37		.043	2,322,562	2.20	2	−0.14	0.623
California	CA	28,901,306	15.42	3	−3.16	<.001		26,703,922	25.30	13	−2.45	0.006
Colorado	CO	3,544,528	1.89	3	0.81		.294	3,246,107	3.08	9	3.38		.004
Connecticut	CT	3,281,288	1.75	2	0.19		.522	3,156,907	2.99	2	−0.57	0.425
Delaware	DE	698,000	0.37	0	−0.61	.691		656,000	0.62	3	3.02		.025
Florida	FL	12,851,426	6.86	7	0.06		.529	11,534,855	10.93	14	0.93		.212
Georgia	GA	6,881,125	3.67	1	−1.39	.119		6,391,438	6.06	1	−2.05	0.016
Hawaii	HI	1,082,944	0.58	0	−0.76	.560		1,007,953	0.95	2	1.07		.246
Idaho	ID	1,105,045	0.59	0	−0.77	.554		1,032,037	0.98	4	3.06		.018
Illinois	IL	11,844,336	6.32	2	−1.72	.049		11,550,447	10.94	15	1.23		.142
Indiana	IN	5,758,602	3.07	7	2.24		.037	5,575,835	5.28	4	−0.56	0.393
Iowa	IA	2,897,691	1.55	0	−1.24	.212		2,874,409	2.72	1	−1.04	0.245
Kansas	KS	2,526,459	1.35	0	−1.16	.259		2,468,549	2.34	5	1.74		.088
Kentucky	KY	3,789,327	2.02	1	−0.72	.401		3,664,106	3.47	5	0.82		.269
Louisiana	LA	4,214,256	2.25	1	−0.83	.343		4,054,713	3.84	1	−1.45	0.104
Maine	ME	1,189,946	0.63	0	−0.80	.533		1,153,121	1.09	1	−0.09	0.703
Maryland	MD	4,798,128	2.56	4	0.90		.255	4,515,274	4.28	7	1.32		.142
Massachusetts	MA	6,067,891	3.24	6	1.54		.110	5,914,743	5.60	6	0.17		.488
Michigan	MI	9,450,968	5.04	1	−1.80	.039		9,179,640	8.70	3	−1.93	0.026
Minnesota	MN	4,495,925	2.40	3	0.39		.430	4,315,604	4.09	11	3.42		.003
Mississippi	MS	2,624,581	1.40	4	2.20		.054	2,550,151	2.42	4	1.02		.225
Missouri	MO	5,259,250	2.81	3	0.12		.533	5,102,708	4.83	7	0.99		.213
Montana	MT	834,274	0.45	1	0.83		.362	807,728	0.77	3	2.55		.043
Nebraska	NE	1,633,956	0.87	5	4.42		.002	1,596,797	1.51	8	5.27		<.001
Nevada	NV	1,438,025	0.77	0	−0.88	.463		1,245,855	1.18	1	−0.17	0.670
New Hampshire	NH	1,064,028	0.57	0	−0.75	.566		987,857	0.94	7	6.27		<.001
New Jersey	NJ	7,893,934	4.21	6	0.87		.248	7,589,445	7.19	7	−0.07	0.570
New Mexico	NM	1,542,432	0.82	1	0.20		.560	1,443,860	1.37	1	−0.31	0.602
New York	NY	18,418,931	9.82	4	−1.86	.033		18,146,109	17.19	7	−2.46	0.005
North Carolina	NC	7,034,846	3.75	2	−0.90	.277		6,621,705	6.27	6	−0.11	0.563
North Dakota	ND	644,869	0.34	0	−0.59	.712		642,724	0.61	3	3.06		.024
Ohio	OH	11,037,317	5.89	5	−0.37	.464		10,815,431	10.25	2	−2.58	0.002
Oklahoma	OK	3,186,335	1.70	2	0.23		.507	3,043,280	2.88	7	2.42		.028
Oregon	OR	2,963,805	1.58	3	1.13		.211	2,764,671	2.62	0	−1.62	0.073
Pennsylvania	PA	12,104,493	6.46	4	−0.97	.228		11,973,578	11.34	7	−1.29	0.123
Rhode Island	RI	1,000,289	0.53	2	2.01		.099	976,741	0.93	5	4.24		.003
South Carolina	SC	3,567,382	1.90	1	−0.65	.434		3,369,985	3.19	0	−1.79	0.041
South Dakota	SD	724,362	0.39	0	−0.62	.677		717,135	0.68	4	4.03		.005
Tennessee	TN	5,085,567	2.71	2	−0.43	.491		4,832,473	4.58	1	−1.67	0.057
Texas	TX	17,681,858	9.43	9	−0.14	.531		16,331,548	15.47	9	−1.65	0.056
Utah	UT	1,847,991	0.99	1	0.01		.628	1,688,493	1.60	0	−1.26	0.202
Vermont	VT	550,714	0.29	0	−0.54	.748		523,928	0.50	3	3.55		.014
Virginia	VA	6,252,451	3.34	1	−1.28	.154		5,871,543	5.56	3	−1.09	0.195
Washington	WA	5,005,334	2.67	2	−0.41	.501		4,646,612	4.40	2	−1.14	0.185
West Virginia	WV	1,829,662	0.98	2	1.04		.257	1,834,718	1.74	4	1.72		.099
Wisconsin	WI	5,013,256	2.67	1	−1.02	.254		4,836,380	4.58	1	−1.67	0.057
Wyoming	WY	462,511	0.25	3	5.54		.002	440,426	0.42	0	−0.65	0.657

States with statistically significant differences in actual and expected case counts are highlighted separately in the Main Map. For the abortion case analysis, Alabama (20 actual cases, 2.2 expected cases, P < .001) and Alaska (5 actual, 0.28 expected, P < .001) had a statistically significantly higher number of cases than expected, while only California (3 actual, 15.42 expected, P < .001) had a statistically significantly lower number of cases than expected. For redistricting, we observe statistically significantly high case counts in Alaska (11 actual, 0.44 expected, P < .001), Nebraska (8 actual, 1.51 expected, P < .001), and New Hampshire (7 actual, 0.94 expected, P < .001), while no states had statistically significantly low case counts.

We clearly observe broad variation in the Z-scores of published cases that reached the SSC level relative to a state's base population. Given the small number of cases in many states, we interpret states with Z-scores between −0.50 and 0.50 as having a number of published cases that is similar to what would be expected relative to these states’ average populations. Due to the right-skewed nature of the Poisson distribution, we consider states with Z-scores > 0.50 as being nationally overrepresented in published cases relative to their base population, with Z-scores > 5 being particularly high, and we consider states with Z-scores < −0.50 to be nationally underrepresented, with Z-scores < −2 being particularly low.

In the top panel of the Main Map (abortion cases published since 1973), Alaska, Alabama, and Wyoming have the highest Z-scores, which is attributable to modest numbers of cases in states with the lowest two population averages in the nation. For example, there were as many published SSC cases in Wyoming (3) over the study period as in California, though Wyoming contained less than 2% of California's average decennial population. The eight highest Z-scores ( > 2) are located in two small clusters of contiguous states in the Great Plains (WY, NE) and Southeast (AR, MS, AL), plus Indiana, Rhode Island, and Alaska. There is no overwhelming geographic pattern for the 24 states with Z-scores < −0.50, though there are potential clusters in the southwest, northeast, and within a snaking chain of contiguous states from North Dakota to Georgia. Eleven states had no abortion-related cases reach their respective SSC during the study period, and these states are geographically spread out around the nation. Highly populated New York and California had fewer cases than would be expected, while the published caseloads for Florida and Texas were in proportion to their populations.

The national variation in Z-scores is much higher for redistricting cases than is observed for abortion cases, and this reflects the higher overall number of published cases during the study period. In the bottom panel of the Main Map (redistricting cases published since 1962), Alaska again has the highest Z-score, and we observe 23 states with a Z-score > 0.50, primarily concentrated in the Midwest and Great Plains regions. Only seven states fall into the ‘proportionate’ category, and another four had zero cases during the study period (OR, WY, UT, SC). Many of the 20 states with Z-scores < −0.50 are located in the Sun Belt and Great Lakes regions.

We evaluated the Z-scores for the abortion and redistricting case distributions for global spatial autocorrelation using Moran's I statistic, and we restricted our analysis to the contiguous 48 states to remove spatial biases associated with Alaska and Hawaii. The Z-scores for abortion cases did not reveal any statistically significant pattern after testing a variety of contiguity- and distance-based spatial weights matrices. The Z-scores for redistricting cases did exhibit positive spatial autocorrelation using both a queen's case contiguity matrix (I = 0.190, P = .030) and an inverse-distance weights matrix (I = 0.226, P = .045) to compute Moran's I. We then explored local spatial autocorrelation by computing the LISA statistic using a variety of spatial weights matrices and a correction for the false discovery rate (to minimize bias due to multiple comparisons, consistent with our original interpretation of P-values for the Z-scores in the Main Map). The LISA statistic revealed no statistically significant clusters or outliers for the abortion Z-scores, but did identify Nebraska and South Dakota as a statistically significant cluster of high Z-scores for the redistricting case data using an inverse-distance weights matrix. If the false discovery rate correction is removed from the redistricting spatial analysis, thus relaxing the threshold for statistically significant clusters, North Dakota joins the Great Plains cluster of Nebraska and South Dakota, Vermont and New Hampshire emerge as a second cluster in New England, and New York is identified as a spatial ‘low–high’ outlier (i.e. an unusually low value surrounded by high values). If we do the same for the abortion spatial analysis, only Mississippi emerges as a lone high-Z-score cluster.

4. Conclusion

This paper presents innovative maps that help us visualize the relative state-by-state distribution of published SSC cases on abortion and redistricting. By mapping a Z-score for each US state, we can visually compare the differences between the actual and expected number of published SSC cases for each state relative to its population, and proceed to search for spatial clusters of high or low values. We only observed one local cluster of high Z-scores in the redistricting map, as statistical corrections for multiple tests constrain the identification of statistically significant spatial clusters of state scores, despite the qualitative appearance of spatial clustering. More importantly, because the mean number of published cases by state during the study period is small (2.7 for abortion and 4.6 for redistricting), the Z-score for this type of Poisson variable may be sensitive to the addition or subtraction of a single case in some parts of the map, particularly for low-population states. This sensitivity, in addition to the underlying variation in state law and motives regarding the publication of SSC cases, leaves us openly wary of overinterpreting state-to-state comparisons, and so we focus instead on some of the potential hypotheses generated by exploring the broader geographic and cultural trends that may influence the variation in Z-scores.

There are a number of hypotheses that we might pose upon visual exploration of the Z-score maps for abortion and redistricting, and we might begin by exploring any legal explanations or unusual state requirements about SSC case publication that could account for extreme Z-scores. Does politics play a role in the number and publication of SSC cases? For example, all of the highest-Z-score states in the abortion case map are traditionally politically conservative states, with the exception of MA. Do states such as Alaska or Wyoming have stricter state regulations on abortion, poorly written laws, or both? Abortion was illegal in six of the states with high Z-scores for abortion cases (IN, MT, NE, RI, WV, WY) prior to Roe v. Wade, and several others had various restrictions on access. Today's abortion restrictions manifest in a complex menu of parental notification, mandatory waiting period, abortion counseling, and mandatory ultrasound laws that vary by state (Guttmacher Institute, 2014a, 2014b, 2014c). We might use these data to generate hypotheses about the relative influence of specific abortion restrictions on litigation activity.

With respect to redistricting, is more severe gerrymandering occurring in northern states, and is this actually securing partisan victory? Perhaps it is occurring at the local (intra-state) level, but this seems unlikely at the state level, given that the states with the highest Z-scores have low populations and do not generally swing national elections. Do states with high Z-scores correspond to those traditionally affected by Section 5 of the Voting Rights Act of 1965 (which was recently invalidated by 2013's Shelby County v. Holder)? Surprisingly, of the nine states requiring statewide coverage under Section 5, only Alaska, Mississippi, and Alabama have positive Z-scores. This may inspire various hypotheses about the relative influence of Section 5 on election disputes in these southern states. As a sidebar to Section 5's core mission of upholding voting rights, perhaps the preclearance process succeeded in keeping some potential disputes out of the judicial system.

Because we did not map the cases from the perspective of a plaintiff/defendant ruling, or for a particular political ideology, one cannot reasonably argue for any patterns related to case outcomes in these exploratory maps. But these finer grains of detail are also inherently mappable, and SSC rulings might still harbor geographic patterns when additional case attributes are taken into account. This methodology, which additional refinement, may also have implications for the analysis of forum shopping and other geographically linked legal phenomena. For example, future analyses might explore spatial interaction effects between states, such as litigation in one state being influenced by laws or legal precedent set in contiguous states. Forum shopping of patent cases has been analyzed geographically (Moore, 2001), but these types of spatial effects have otherwise received little social science attention. Spatial weights matrices could also be adapted to capture alternative geographies such as correspondence between courts, cultural similarities, or learning across states.

Despite the limitations of this data set, published SSC cases generally represent challenges to state laws, and these decisions ultimately not only set precedent for future cases, but often guide legislation in other states and create cultural ripples that can influence federal court decisions. Cases reaching an SSC do not necessarily pertain to federal law, and only a small subset of state cases are eligible for review by the U.S. Supreme Court; yet, these cases are often an important part of broader state, regional, and national dialogues. Small states with high Z-scores may have a particularly disproportionate influence on case precedent and these social dialogues even though the laws being challenged may apply to a very tiny minority of the national populace. Conversely, states that have published zero SSC cases may have played little, if any, part in setting precedent, even though cases on the topics of abortion and redistricting may have been heard in these states. It could be argued that these states are being left out of the national discourse on a given issue. The number of expected cases might also be scaled by other factors besides mean population, such as the overall number of lawsuits or lobbying dollars spent, or the proportion of cases on a particularly subject relative to all SSC cases heard in the state. Such factors are likely to be specific to the legal issue of interest, and may present new challenges for the standardization and comparability of data across all 50 states, but do present another opportunity for future research.

By geographically visualizing the distribution of published SSC cases, we can hypothesize about the spatial patterns associated with certain types of court decisions. The geographic information system – or any geovisualization tool – thus helps us ask questions that can then be addressed in richer detail using complementary methods for court data as previously demonstrated with network analysis (Caldeira, 1988) or case history visualization (C. Harris et al., 1999). This approach allows us to see where issues of national importance are being addressed across the US court system while exploring key drivers that may influence legal decisions. We have just scratched the surface of the potential for geovisualization of legal information, and we hope that the increasing availability of public records and computational methods will encourage further geovisualization and spatial analysis of our judiciary system.

Acknowledgements

We thank Robin Schard at the University of Miami School of Law for her guidance and suggestions, as well as four reviewers whose suggestions greatly improved this paper.