Problem Statement

You are consulting for a real estate company that has a niche in purchasing properties to rent out short-term as part of their business model specifically within New York City. The real estate company has already concluded that two bedroom properties are the most profitable; however, they do not know which zip codes are the best to invest in.

Objective

Find properties based on zipcodes in the New York city which would yield the maximum return on investment (ROI)

Assumptions

1. Occupancy rate has been assumed to be constant throughout the year irrespective of the fact that it can change due to holidays
2. Review score rating has been used to adjudicate the occupancy rate as it is the overall score based on
   • accuracy
   • cleanliness
   • check-in
   • communication
   • location
   • value
3. Cost of property has been predicted using time series forecasting from the Zillow Data set. Extraneous variables like natural calamities, slowdown in economy which can affect the cost of the property have been ignored.
4. Mean and median prices have been taken to calculate the breakeven period.
5. The data was last scraped in the third quarter of the FY 2017 and the analysis is done in 2019. Factors like holiday effects, superhost, location, amenities etc. could effect the price variation for short term rentals and hence prediction of rental price is not done.

Metadata Created

1. predicted_price -This column contains the predicted cost price from the zillow data set. In the data set only the price as of 1 August 2019 is taken and is in dollar amount
2. occupancy_score -Same as review_scores_rating
3. occupancy_rate -Percentage occupancy of the airbnb listing. It is represented as intervals.
4. breakeven_years -Time it takes for the property to return it’s cost price. This is also known as breakeven period and it is taken in the form of years

Exploratory data analysis

Data Integrity and Data Quality

After doing exploratory data analysis and data munging, we can comment on the data quality and data integrity

I concur with the statement “Bad data is worse than no data”.

1. Missing values : There were a few columns in which missing values were present. The variables important for our analysis are
   • review_score_rating : occupancy_rate is directly dependent on the score
   • price : There are 41 properties which do not have their price listed.
2. Data Quality : Some zipcodes had length other than 5 which questions the data credibility
3. Quality Check of data was performed by
   • removing inconsistencies : zipcodes of abnormal length, missing values in price columns, rating columns
   • conversion into suitable data type : There are certain columns which were reformatted for calculation,data type was changed from character to factor, or from factor to numeric based on the charting requirements as well as limitation of the tool at hand.
4. Uniqueness/Variance : Some of the columns had a low count of unique values. Although low count of unique values are a distinguished feature of binary variables(0 or 1, TRUE or FALSE, this or that) but there should not be a single value i.e no change in variable throughout the dataset. Such columns have been removed or not considered for analysis
5. Timeliness: The data was last scraped in the third quarter of the FY 2017 and the analysis is done in 2019. Prediction could have been done (is done but not used) for Zillow properties but we don’t have 2019 Airbnb data our analysis is restricted to FY 2017

Visualizations of the properties versus price and other parameters

There are few zipcodes which have over 70 properties. The red dotted line indicates the arbitrary y-axis intercept at 70

The boxplot shows

1. Three phases can be seen in the cost price

• Zipcodes with very high price (10013,10014,10011)
• Zipcodes with medium price (10023,10028…100217)
• Zipcodes with low price (100215,….100304)

Average price is decreasing from 320 to 70

Median price is decreasing from 300 to 70

Averages are bit higher than median price suggesting averages are being pulled over because of outliers.

The distribution of airbnb price is right skewed. The red dotted line indicates the arbitrary x-axis intercept at 750 which is relatively higher compared to other Airbnb prices and corroborated by the outlier plot.

It is skewed towards the left which means that majority of the properties have received a score greater than 50

Which properties to invest in ?

Conclusion

The ROI is less for some zipcodes but there are not many properties. We atleast need a few properties to be able to make a firm decision.

Therefore the zipcodes to invest in are

• 11231
• 11217
• 11215
• 10036
• 10025

Future Steps

1. There are over 35,000 observations besides 2 bedrooms hence further analysis can be done.
2. Factors such as transportation, proximity to work location can be used to map the area for long-term rentals
3. NY being a financial capital has a scope for long term rentals. Hence, the properties can be rented out on a long term basis as well.
4. There are 618 rows which have improper length zipcodes. These can be imputed using the latitude and longitudes
5. Crime score is an important factor in determining the price. Crime score is freely available online and can be used for deeper analysis.
6. Majority of the tourists book short term rentals in NY city hence holidays (variable pricing can be done ) and closeness to tourist locations should be accounted for.