Blog Post < Previous | Next >
![](https://s3-us-west-1.amazonaws.com/themarinpost/img/22112/_list/Example_b83.png)
Guy
Marin County Crime Rate Trends
Introduction
Using the Marin County Sheriff Reported Crimes data, I reviewed annual crime rate trends from 2013 to 2023 for 9 different crimes:
- Vandalism
- Theft
- Sexual offense
- Auto theft
- Fraud
- Drugs (drug and narcotics infractions)
- Burglary - auto (or motor vehicle0
- Burglary - residence
- Assault
The above are among the most prevalent and specified crimes within the mentioned data outside of traffic violations (of no interest to me).
I focused on 11 different cities with readily available population data. And, after computing the relevant crime rates, I split them into four tiers as shown in the table below.
The three cities in Tier 0 had so few crime occurrences that there was not enough data to analyze long-term trends. So, I left them out of this analysis.
All the other cities needed to be grouped in separate tiers because their respective crime rate levels were too divergent to observe within a single group. This is a scale issue.
The scale issue
If City A has a crime rate of 1,000 per 100,000 individuals (the denominator for crime rate), and Cities B, C, D have crime rates ranging from 10 to 50, you can't visualize their crime trends within the same graph. City A's crime rate will look really high. Meanwhile, Cities B, C, D will all have crime rates appearing very close to Zero. However, some of those cities may have a crime rate that is 5 times greater than the others.
To resolve this scale issue, you visualize City A's data separately from Cities B, C, D. By comparing cities that fall within a similar scale you can better visualize their respective long-term trends.
The different crime rate scales
The table below shows the median annual crime rate for the eight cities for nine different crime types. The data is color-coded ranging from green = very low to red = very high. And, the color coding is focused on a single row of crime types at a time. This facilitates the comparison of the eight cities' crime rates focusing on a single specific crime type at a time (example: Vandalism).
The table below looks at the maximum annual crime rate over the 11 years (2013 - 2023). And, it stresses the need for grouping cities in different tiers, with each tier having a similar scale.
When you look at both tables above, there is a consistency in the color coding.
The four cities in Tier 1 (Fairfax, Novato, San Anselmo, Tiburon) are mainly green denoting very low crime rates.
The two cities in Tier 2 (San Rafael and Sausalito) are yellow denoting relatively low crime rates.
Mill Valley in Tier 3 has relatively high crime rates in orange.
Marin City in Tier 4 has very high crime rates shown in red.
The graph matrix depicting 9 different crime rate trends
We will be looking at the graph matrix depicted below for each city tier. The embedded graphs are pretty small so I disclose what specific crime each of the graph depicts.
Once you recall what specific crime each graph depicts, it is easy to interpret the disclosed crime rate trend for each crime type.
Tier 4: Marin City
When looking at the graphs above, it is key to focus on the scale of the different Y-axes. They each have their own scale. And, they can be very different. For instance, the Y-axis for "Marin City: Theft" peaks at above 2,500 per 100,000. Meanwhile, the one for "Marin City: Sexual Offense" peaks at just above 150 per 100,000.
As shown below, Marin City's crime rate has dramatically declined for four different crime types: Fraud, Drugs, Burglary Residence, and Assault.
All four trends are very favorable. The trends for the other five crime types are more volatile. However, overall they typically also show a declining trend. These trends indicate that Marin City, even though it has a high crime rate vs. the others, is definitely becoming safer over time.
Tier 3: Mill Valley
When looking at Mill Valley's crime data, I checked for a pre-pandemic pattern. The hypothesis would entail that crime rates may have gone down after 2019 once more people stayed home. You would think this would be associated with fewer burglaries, theft, vandalism, etc. after 2019. As we will see this pattern does not often play out.
Below I focus on four crimes associated with declining trends over the past several years. A couple of them did peak in 2019, and declined thereafter. But, most other crimes in Mill Valley did not follow the pre-pandemic pattern I hypothesized. This was true for the other cities as well.
Mill Valley's relatively high crime rate was surprising. I did not expect its crime rate to be higher than San Rafael and Novato.
Keep in mind that crime occurrences are very sensitive to location at the street level. I remember a couple of years ago, I reviewed this data at the street level, and I could not find a single burglary event over the 2013 - 2021 period for numerous streets in Mill Valley. Yet, just half a mile away there were numerous burglaries frequently reported at Nextdoor.com.
From just looking at a neighborhood, you can't always guess what is its crime rate. Two neighborhoods can be equally nice. But, one neighborhood provides burglars easy escape to 101 or Shoreline Highway, and the other neighborhood does not. And, that will make all the difference in the respective crime rates.
Tier 2: San Rafael and Sausalito
San Rafael and Sausalito's crime rates are on a similar scale allowing us to compare both cities' trends together. within the graphs shown below.
Sausalito's trends (green line) are more volatile than San Rafael's (red line). This is a function of population size. Sausalito's population at around 7,000 is only about 1/9th the population of San Rafael at about 60,000. The smaller the population or sample size the more volatile the data.
Focusing on Auto theft (shown in the graph below) is interesting. You can readily observe a strong negative correlation in Auto theft between San Rafael and Sausalito over the 2013 - 2023 period. Over that period, San Rafael's Auto theft rate rose from 21 to 48. Meanwhile, Sausalito's rate declined from 54 to 0. Sausalito experienced a rapid decline in 2015. And, it kept this rate low ever since.
Within the graph, the line reveals the underlying trend in the data, and the LOESS 50% confidence Interval captures the uncertainty associated with the underlying trend line. As defined about 50% of the data points should fall within the depicted confidence interval. It is not quite the case because of small sample errors (just 11 annual data points from 2013 to 2023).
Notice how the confidence interval for San Rafael (red) is so much narrower than for Sausalito (green). That's because San Rafael's population is so much larger. As a result, the crime rate from one year to the next literally "trends" much more within a narrower band of uncertainty.
Tier 1. Fairfax, Novato, San Anselmo, Tiburon
The crime rate data for these four cities is fairly trendless. Novato (dark green line) being by far the largest city within this tier (51,000) has much less volatile crime data than the other smaller cities. As mentioned earlier, larger population = lower data volatility and vice versa.
Keep in mind, when focusing on Tier 1 all crime rates as visualized are pretty low. You can confirm that by observing the range of the respective Y-axes. None of them exceed crime rates much above 100 per 100,000. These are very low maxima compared to the other tiers.
Reference: how I coded that stuff
If you are interested in that aspect of this project, you can check my Medium article.
THE END