Beginner's Guide to Comparable Sales Analysis: Insights from ChatGPT and Bard

Sid: Using the following sales table comprising all 2023 sales, I have selected five comparable sales (comps) to determine the value of a subject property in the same subdivision. The subject property has heated square footage (HEATED SF) of 1,580. All the comps are not only extracted from the same subdivision, a PUD community, with identical land use and amenities, but they are also very similar in land sizes, year of construction, and construction quality. Additionally, these are all interior lots with very similar surroundings and views. To select the five best comps, I used the method of least price adjustments, ignoring signs, meaning I considered both positive and negative adjustments equal in value. Using this method, I calculated the least price adjustments by subtracting the HEATED SF of each comp from 1,580 and multiplying the result by $183, the median price per HEATED SF. Based on this calculation, I have selected five comps: 4, 9, 13, 16, and 20. The median value of these five least price-adjusted comps is $288,601, which is the estimated value of the subject property. Please review my methodology for selecting the comps and provide me with your analysis and opinion.

ChatGPT: Based on your methodology for selecting the comps, it appears that you have focused on finding properties within the same subdivision that have similar characteristics to the subject property, including land size, year of construction, and construction quality. Additionally, you have used the method of least price adjustments to select comps, considering both positive and negative adjustments as equal in value.

By calculating the least price adjustments based on the difference in heated square footage between each comp and the subject property, and multiplying it by the median price per heated square footage, you have determined the adjusted prices for each comp. Then, by taking the median value of these adjusted prices, you have arrived at the estimated value of the subject property as $288,601.

Overall, your approach seems logical and systematic, taking into account key factors such as the subdivision, property characteristics, and price adjustments. The selected comps provide a range of adjusted prices, allowing for a more accurate estimation of the subject property's value. However, it is important to note that this analysis is based solely on the provided data and methodology.

However, there are a few considerations to keep in mind:

1.    Adjustments: While you have used the median price per heated square foot to calculate adjustments, it's important to note that this method assumes a linear relationship between square footage and price. In reality, the relationship may not be perfectly linear, and adjustments based solely on square footage may not capture all the differences that could affect the value of a property.

2.    Timeframe: The sales data you provided is from 2023, and the subject property's value is estimated based on this data. It's important to consider if market conditions have changed since then. Real estate values are influenced by various factors like location, market trends, and economic conditions, so newer sales data may provide a more accurate representation of the subject property's value.

3.    Sample size: While you have selected five comps, a larger sample size could enhance the accuracy and reliability of the estimated value. Ideally, a greater number of comparable sales within the same subdivision and similar characteristics would provide a more comprehensive analysis.

While your methodology appears to be logical and considers important factors when selecting comps, it's always recommended to consult with a professional appraiser or real estate agent who can provide a more detailed and accurate valuation. They may be able to provide additional insights or adjustments that could further refine the estimate.

Bard: 

Review of Methodology for Selecting Comparable Sales

Your methodology for selecting comparable sales (comps) has several strengths and weaknesses:

Strengths:

·         Focus on subdivision and property type: Limiting your selection to the same subdivision (PUD community) with identical land use and amenities is a good starting point. Additionally, focusing on interior lots with similar surroundings and views further strengthens your selection.

·         Median price per square foot: Using the median price per square foot as the basis for price adjustments is a common and reliable method.

·         Ignoring signs for adjustments: Considering both positive and negative adjustments equal in value is appropriate, as it avoids introducing bias based on the direction of the adjustment.

Weaknesses:

·         Limited number of comps: Selecting only five comps may not be statistically sound, especially given the potential for outliers influencing the median value. Aiming for 6-10 comps is generally recommended.

·         Sole reliance on heated square footage: While heated square footage is an important factor, other property features like the number of bedrooms, bathrooms, upgrades, and condition can also significantly impact value. Considering these additional factors could potentially lead to a different set of comps or adjustments.

·         Mechanical application of adjustments: Simply multiplying the square footage difference by the median price per square foot may not adequately capture the complexity of price adjustments. Some adjustments, such as those for upgrades or condition, may require a more nuanced approach.

Overall:

Your methodology has merit but could be improved by considering a few additional factors:

·         Increase the number of comps: Aim for at least 6-10 comps to ensure a more statistically robust analysis.

·         Factor in additional property features: Don't rely solely on heated square footage. Consider the number of bedrooms, bathrooms, upgrades, condition, etc., and adjust accordingly.

·         Refine price adjustments: Apply adjustments based on the specific characteristics of each comp, rather than simply using a mechanical formula.

By addressing these weaknesses, you can refine your comp selection and obtain a more accurate estimate of the subject property's value.

Additional Points:

·         It's important to document your rationale for selecting each comp and the adjustments made. This transparency will increase the credibility of your valuation.

·         Consider consulting with a professional appraiser for a more thorough and accurate valuation, especially if the property has unique features or the market is complex.

I hope this analysis and feedback are helpful!

Sid: Would it be appropriate to have a larger sample size and make more adjustments - beyond just the heated area - for a community with a more diverse housing stock? For example, should we consider factors such as land size, effective building age, construction quality, exterior, view, etc.? If so, would a regression model be more statistically significant?

ChatGPT: Yes, it would be appropriate to consider a larger sample size and make additional adjustments for a community with a more diverse housing stock. In such cases, factors beyond just heated area, such as land size, effective building age, construction quality, exterior, view, etc., can significantly impact the value of a property.

To account for these factors and make more precise adjustments, a regression model can be a valuable tool. A regression model allows you to analyze the relationships between various property features and their corresponding impacts on price. By collecting data on a larger number of comparable sales and including a wide range of factors in the model, you can determine how each factor contributes to the overall value of a property.

Using regression analysis, you can estimate the coefficients or weights for each factor and apply them to the subject property's characteristics. This will result in a more statistically significant and accurate estimation of its value.

However, it's important to ensure that the data used in the regression analysis is relevant, reliable, and consistent. If the data includes sales that are significantly different from the subject property in terms of location, size, amenities, or condition, the regression model may produce less accurate results.

In summary, for a community with a diverse housing stock, a larger sample size and more comprehensive adjustments, along with a regression model, can help provide a more statistically significant and accurate estimation of the subject property's value. Consulting with a professional appraiser or real estate agent experienced in regression analysis would be beneficial in this case.

Bard: Absolutely! For a community with a more diverse housing stock, adjusting your comparable sales analysis (CSA) to account for various factors beyond just heated area is not only appropriate but essential for accuracy and statistical significance.

Here's why:

Larger Sample Size:

·         With diverse housing options, relying on a small sample of comps can be misleading. A larger sample (ideally 6-10 comps) increases the likelihood of capturing the range of values within the market and mitigating the impact of outliers.

Multiple Adjustments:

·           Just focusing on heated area ignores other crucial factors influencing value, such as:

o        Land size: Larger lots typically command higher prices, especially in areas with limited land availability.

o        Effective building age: While year of construction provides a starting point, consider renovations, upgrades, and overall condition for a more accurate reflection of actual age.

o        Construction quality: Higher-quality materials and construction methods generally translate to higher values.

o        Exterior: Factors like architectural style, siding material, and curb appeal can influence market perception and price.

o        View: Scenic vistas or desirable natural features can significantly increase value.

Regression Model:

·         Regression models can be more statistically significant than simply calculating median values, especially when dealing with multiple adjustments. These models analyze the relationship between sale prices and various property characteristics, providing a more nuanced understanding of market dynamics and potentially revealing influential factors you might have missed.

Benefits of More Adjustments and Regression:

·         Increased accuracy: By considering numerous factors and their impact on value, you generate a more accurate estimate for the subject property.

·         Reduced bias: A larger sample and comprehensive adjustments minimize the influence of individual comps and personal biases.

·         Stronger justification: Documenting the rationale behind each adjustment and using a statistical model strengthens your valuation report and bolsters its credibility.

However, keep in mind that adding too many adjustments can lead to overfitting and reduce the model's generalizability. Aim for a balance between comprehensiveness and practicality.

Final Thoughts:

For a diverse housing community, embracing a larger sample size, multiple adjustments, and potentially a regression model are excellent ways to ensure a robust and accurate CSA. Remember, the goal is to capture the unique characteristics of the market and the subject property to arrive at a reliable value estimate.

Sid: You indicated, "keep in mind that adding too many adjustments can lead to overfitting and reduce the model's generalizability." How does an analyst avoid overfitting a model?

 

Bard: Avoiding overfitting in a regression model used for comparable sales analysis is crucial for ensuring your value estimate's accuracy and generalizability. Here are some strategies to achieve that:

Data pre-processing:

·         Feature scaling: Standardize or normalize your data to prevent features with larger scales from disproportionately influencing the model.

·         Dimensionality reduction: Techniques like Principal Component Analysis (PCA) can be used to identify and eliminate redundant or irrelevant features, reducing the complexity of the model and mitigating overfitting.

Model selection and training:

·         Start with a simple model: Begin with a model with fewer coefficients like Linear Regression and gradually increase complexity, monitoring for overfitting.

·         Regularization: Techniques like L1 or L2 regularization penalize complex models, discouraging overfitting.

·         Cross-validation: Divide your data into training and validation sets. Train the model on the training set and evaluate its performance on the unseen validation set. This helps assess generalizability and prevent overfitting on the training data.

·         Early stopping: Monitor the model's performance as it trains. When the validation error starts increasing, stop training to avoid overfitting.

Evaluation and analysis:

·         Analyze residuals: Plot the residuals (differences between predicted and actual values) to identify patterns and potential outliers.

·         R-squared and adjusted R-squared: While high R-squared might seem ideal, adjusted R-squared penalizes for added features, giving a more accurate picture of model fit while accounting for overfitting.

·         AIC and BIC: Use information criteria like Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) to compare models and penalize complexity, favoring simpler models that avoid overfitting while maintaining good fit.

Remember, there's no one-size-fits-all solution. The optimal approach depends on your specific data and model.

By utilizing these strategies and carefully monitoring your model's performance, you can effectively navigate the trade-off between model complexity and overfitting, ultimately achieving a statistically sound and generalizable valuation through your comparable sales analysis.

Note: All conversations here are unedited. E&OE. ChatGPT and Bard are registered trademarks of OpenAI and Google, respectively.