Data Analysis of Location Factors Affecting Neighborhood Property Prices of Miami.

A. Introduction

A.1. Background and Problem

?Miami is a beautiful coastal city in the USA.?Its growing metropolitan area is blessed with a favorable climate and a low tax environment. Inspired by the famous real estate axiom "Location, location, and location," I will try to find optimal neighborhoods to live in in the Miami area, which is made out of Downtown Miami and Miami Beach.

The final decision on optimal property location is a complex process that entails many choices. I'm interested to see how location factors such as the density of schools, beaches, golf courses, restaurants, or places one could go to exercise, do shopping, or do cultural activities and crime rates influence property prices. I'll use data science tools of visualization, correlation, and k-means clustering algorithm to group the neighborhoods into clusters.??

The conclusions of differences and similarities between Miami neighborhoods will be useful for not only prospective buyers out of town but also local real estate brokers. It will help quickly to obtain information about location characteristics, different venue densities, property prices, and crime rates in each neighborhood.

A.2. Data Description????

?I based my analysis on the following databases, APIs, libraries:

·?????I found the list of all Miami and Miami Beach neighborhoods and their typical home prices from Zillow Database. Zillow website provides the Zillow Home Value Index (ZHVI): A smoothed, seasonally adjusted measure of the typical home value and market changes across a given region and housing type. It reflects the typical value for homes in the 35th to 65th percentile range. I used the latest available dataset from the end of May 2021. Source:?https://www.zillow.com/research/data/

·?????Google API?helped me to get the latitude and longitude coordinates for Miami neighborhoods.?

·?????Folium library was used to visualize the map of Miami.

·?????I used Foursquare API to get venue names, locations, and venue categories

·?????List of all public schools and their GPS coordinates (latitude and longitude) was found in?Miami-Dade County’s open data hub. Source:?https://gis-mdc.opendata.arcgis.com/datasets/private-school?geometry=-81.005%2C25.493%2C-79.921%2C25.926

·?????Finally, for crime data, I used the data of the Area Vibes website. This website compiles the number of violent crimes (murder, rape, robbery, or assault) per 100,000 people living in each neighborhood from the FBI and local law enforcement agencies and when not available, also includes estimates based on demographic data.?https://www.areavibes.com/miami-fl/most-dangerous-neighborhoods/.

B. Methodology

I used Zillow and Google API to build my master data frame “mia_neigh”, which has the main components?Neighborhood, City, Typical Property Value, Latitude,?and?Longitude?information. (I only include the first 5 out of 35 rows to save space.)

Then, I used the python?folium?library to visualize the map of Miami and then superimpose neighborhoods using GPS coordinates from the table above:

Using describe function (), I discovered the statistical distribution of property prices across different neighborhoods:

?Looking only at the information from my master data, I can see that there are 35 neighborhoods in Miami. The average property price is around $ 2 million and there is a vast price difference between different neighborhoods. For example, a typical home in Star Island ($28 mln) and Little Havanna ($200k) have almost 150 times the price delta!

To understand the reason for this vast price delta, I put together detailed descriptions of each neighborhood: 1) map visualization, 2) Density of the essential categories such as restaurants, schools, gyms, cultural activities, shopping, and availability of golf courses or beach in the vicinity and 3) top 10 most frequent venues table.

To calculate the venue density table, I used Foursquare API. I computed the density of the essential categories such as restaurants (Restaurant Freq), shopping (Store Freq), gyms (Sport Freq), cultural activities (Art Freq), and the availability of golf course, park (Golf|Park Freq), or beach (Waterfront) in the vicinity. To do so, I searched through key names across the Foursquare name categories and then added up their frequencies. For example, for Sport Freq, I was searching for "Gym, Yoga, Tennis, Stadium, Studio" venue names. The Waterfront factor was calculated differently. Here, I used a boolean value of 0 or 1 if either Foursquare API has the name "Beach" or?the neighborhood name has words "Island, Isle, Pointe, Point." See below the merged table (I only include the first 5 out of 35 rows to save space) of these location factors with the master data table:

?Also, using Foursquare API I put together the Top 10 most Frequent Venues for each neighborhood. Looking at this table helps to get a feel for each neighborhood (only the first 5 out of 35 rows included to save space):

Finally, I felt that these factors were not enough, so I've added Crime per 100k and School Count factors. The crime rate was easy. I found the website AreaVibes that already provided the number of violent crimes per 100,000 people for each neighborhood in Miami.

The school count was a bit more complicated. I had a dataset of all public schools with names and GPS coordinates from the Miami-Dade county government data hub, but no number of schools in each neighborhood. To resolve this, I've calculated the distance from each school to each neighborhood using Geopy library. Then, I added up only those schools, which are less than 1 mile away. Finally, I merged all twelve factors together:

C. Results and Discussion

C.1 Correlation Matrix

Once I got all the necessary information, I run a correlation to discover the relationship between these different location factors.

From the correlation matrix, we can see that Waterfront, Golf |Park Freq, and a lot of places to go for exercise ("Sport Freq") in the vicinity have a positive correlation with a property price. On the other hand, I had a surprising finding that the frequency of restaurants/stores/art/schools negatively correlates with property prices in Miami. But I could see how it makes sense. First, we don't know the ranking of those venues. To have density doesn't imply quality. Second, Miami's nature is tourism. Tourist influx brings a lot of foot traffic and a high probability of crime to specific neighborhoods. Indeed, crime rates negatively correlate with property prices and positively correlate with restaurants/stores/art factors.?Finally, the density of public schools doesn't affect the high-end real estate market. This is probably because of the demographics of high-net-worth individuals. They are usually not full-time residents of Miami. That's why most of these buyers are after the convenience of the location and fantastic water views.

C.2 K-means Algorithm

Neighborhoods have common locations factors, so I used an unsupervised learning K-means algorithm to cluster the neighborhoods into 5 clusters. K-Means algorithm is one of the most common cluster methods of unsupervised learning. Here are my merged table with cluster labels (again only the first 5 rows presented to save space) and the map with neighborhood clusters:

?This clustering revealed the following: the most expensive neighborhoods in the Miami area have a central location -- conveniently located btw touristy South Beach and arty/foodie Downtown Miami. They all have a waterfront (which I define as either having a water view or beach) and a golf course or park in the vicinity. They also have a lot of places to go to exercise. In that sense, the famous real estate axiom "Location, location, and location" proved to be true.

C.3 Results and Description of Each Neighborhood Cluster

?"Upscale Islands": Star, Hibiscus, Palm, and other Venetian Islands.

?Looking at the map, we can see that all of these neighborhoods are small islands, which are right next to each other. They are conveniently located at the center between touristy Miami Beach (aka South Beach) and arty/foodie Miami (aka Downtown Miami). Looking at the top 10 venues in these islands, we can see that these upscale areas are filled with water views, spas, places to exercise, and nearby hotels. They all have a golf course or park in the vicinity.?There are no schools or stores. The violent crime rate is 50% less than the average of Miami. These locations obviously come with the price. The k-cluster algorithm rightly put Star Island in its own cluster 1 or purple color as it is especially expensive (typical property value here is $ 28 million !).?The houses on the Star Islands are much bigger than on other islands, and this island has some celebrity owners (hence the name of the neighborhood).?Hibiscus Island is cluster 0 or red color with typical property value here is $ 4 million). Palm and other Venetian islands are cluster 4 or orange color with property values around $ 5 million. I took the liberty to combine them into one cluster because of the high prices.

?"High-End Residential Neighborhoods": Belle Isle, South Pointe, Nautilus, Biscayne Point, and La Gorce in Miami Beach; NE and SW Coconut Grove, and Fair Isle in Coconut Grove.

These neighborhoods are cluster 3 or green colors. Miami Beach neighborhoods have condos and a higher crime rate, given it is at the heart of tourism. Coconut Grove mainly houses, which is quieter with less crime and more schools nearby. Both areas are by the water and next to golf courses, but much more livable than upscale island clusters. They have higher restaurants density. The typical property prices are here $ 1 million, except La Gorce, which is $ 2 million (because it has its own golf club).

"Mid and Lower Tier Residential Neighborhoods"

Cluster 2 (blue color) is very large and represents 19 out of 35 neighborhoods in Miami. The k-means clustering algorithm rightly picked up that property prices here are much lower, below USD 500k. Here the majority of full-time residents live and work. Let's review and break them down in more detail.

"Mid-Tier Neighborhoods ": North Shore, Isle of Normandy, Bayshore, Oceanfront, City Center, and Flamingo Lummus in Miami Beach. Shenandoah, Coral Way, Alameda-West Flagler, Flagami, Upper Eastside, Little Havana in Miami.

These neighborhoods' typical property prices of USD 300-500k. Miami Beach neighborhoods are in the heart of touristy areas, they have so much to offer: many restaurants, bars, hotels in the vicinity. Beach or water view nearby. But this comes with a price of double the average Miami crime rate and smaller condo apartments. Miami neighborhoods have no beach, but some golf or park in the vicinity. They are primarily residential because they have a high density of stores and restaurants. A lot of public schools in the area. Good density of cultural activities as well. The crime rate is lower than Miami Beach but higher than the Upscale islands. These are excellent areas to live in for families.

?"Trendy Miami Neighborhoods": Downtown, Brickell, and Wynwood-Edgewater.

?Historically, these neighborhoods have been dangerous, and still, the violent crime rate is almost 3 of Miami's average, but things are changing. Miami Police headquarters are between Overtown and Downtown. Despite the crime, these areas are considered trendy because each neighborhood has its unique purpose. Downtown has business/government/museums, Brickell is a business district, and Wynwood is an art district. These neighborhoods have high-rise condo residential developments along the water along Brickell Ave, Biscayne Blvd, and Edgewater. They have the highest density of restaurants and cultural activities in Miami. Yet the neighborhoods don't end there but go much more inland, where unfortunately still poverty, homeless people, and crime. There is no beach but some parks in the vicinity.

"Dangerous Lower-Tier Neighborhoods": Liberty City, Allapattah, Little Haiti, and Overtown.

?These neighborhoods are poor and don't have any waterfront. And have more than double the average crime rate of Miami. Liberty City is an exceptional case; with 1919 violent crimes per 100k people, it is considered one of the most dangerous neighborhoods in the US. Typical property prices here are USD 200-300k. Not surprisingly, Miami policy headquarter is located in Overtown.

F. Conclusion

The final decision on optimal property location is a complex process that entails many choices. I only cover some factors; others include noise levels, proximity to major roads, real estate availability, prices per square foot, area of the property, year build, social and economic dynamics, etc.???????????

My project focused on location factors such as availability and density of restaurants, schools, stores, art, places one could go to exercise, and crime rates. I identified the following neighborhood clusters in Miami: "Upscale Islands," "High-End Residential Neighborhoods," "Mid-Tier Neighborhoods" in Miami and Miami Beach, "Trendy Neighborhoods," and "Dangerous Lower-Tier Neighborhoods." By analyzing these using data science tools, I gave a general quantitative description for each neighborhood.

My data analysis showed that the waterfront, the availability of a park, golf course, places one could go to exercise, and the low crime rate significantly affects property prices. However, the density of schools, restaurants, stores, and art negatively correlates with property prices. This is surprising, but I could see how it makes sense as a resident of Miami myself. First, we don't know the rankings of schools/art/stores nearby. To have density doesn't imply quality. Second, Miami's nature is tourism. Tourist influx brings a lot of foot traffic and a high probability of crime to specific neighborhoods (South Beach, Downtown, Wynwood, and Brickell). Finally, the density of public schools doesn't affect the high-end real estate market. This is probably because of the demographics of high-net-worth individuals. They are usually not full-time residents of Miami. That's why most of these buyers are after the convenience of the location and fantastic water views. (see Upscale Islands or High-end residential clusters)

Another surprise could be that the violent crime rate is higher than it is perceived in specific neighborhoods. For example, the trendy neighborhood cluster (Downtown, Wynwood, and Brickell) has double Miami's average crime rate. This makes sense if you look on the map to see how Google API defines a particular neighborhood area. For example, according to Google Maps Brickell, Downtown, Wynwood/Edgewater areas are much bigger than the perceived narrow strip along the water. The neighborhoods don't end with high-rise luxury residential condominiums along Brickell Ave, Biscayne Blvd, and Edgewater, but go further inland. And here, unfortunately, we still have a lot of poverty, homeless people, and crime.

In conclusion, waterfront, a healthy lifestyle, and tourism define the Miami area real estate. These factors might not work in other cities, but one could use my data science methodology to discover what works in the city of their choice.

I hope my findings will aid stakeholders in narrowing down their search for an optimal property location in Miami.

Kind regards,

Liliya?

?