Testing Ideas (Ben)

ideas/Data400Idea1Updated.md

@@ -0,0 +1,29 @@
+## Idea #1:
+
+**Research question:** Do historically racist housing policies have an impact on modern-day home values in Carlisle, Pennsylvania?
+
+**Data source(s):**
+
+1.  Cumberland County GIS provides parcel-level data on properties in the county, including land value, home value, property taxes collected, year built, and name of owner.
+
+2.  During the research process for my senior thesis in History, I have brought together several historical data sets linked to racial housing discrimination in Carlisle.
+
+    a)  The first is a list of households headed by Black individuals from the 1900 census, including number of people in the household, ward number, and address.
+
+    b)  The second is a collection of a dozen property development documents in the Carlisle area that include racial restrictions for ownership.
+
+I plan to bring these disparate data sources together, using covenant status and Black ownership in 1900 as predictive variables for present-day home value.
+
+**Models:**
+
+1.  Least-squares regression - a least-squares regression model will be useful to estimate the impacts of Black ownership and racial covenants on present-day values.
+
+    a)  In the case of home values, there are many complex relationships between confounding variables that might be related to home values. One way to address this issue is to use geographic or socioeconomic controls within a least-squares model. For example, the ward in which the home was built might be used as a control.
+
+2.  Random forest regression - if the relationship between covenants/Black ownership and home values is non-linear or very complex, a random forest regression model will help unravel those nuances. It also provides feature importance scores, which will allow me to understand the impacts of the variables relative to each other.
+
+**Impact on stakeholders**: This is a truly consequential issue. Colin Gordon, an historian at the University of Iowa called racial covenants the "original sin" of segregation in America, and largely blames them for today's large racial wealth gap. At a time when millions of white Americans were building generational wealth for themselves and their families by purchasing homes in prosperous neighborhoods, African Americans and other racial minorities were deliberately excluded. Over several generations, this opportunity gap widened into hundreds of thousands of dollars of wealth out of the reach of Black families.
+
+There are several possible outcomes of this project. If I find that coverage by racial covenants are correlated with increased modern property values - or that Black ownership in 1900 was related to lower values - it raises major questions about wealth inequality in Carlisle. In the case of Evanston, Illinois, a progressive suburb of Chicago, evidence of discrimination against African Americans living in the town in the early twentieth century led to \$10 million in reparations to the descendants of Black residents.
+
+**Ethical implications:** Studies of racial discrimination, especially quantitative ones, always face certain ethical dilemmas. It is impossible to distill decades of poor treatment, from legal discrimination to outright violence, into a mathematical model. Any attempt to do so is in poor form, and disrespects the legacy of those who fought for equality and civil rights in their communities. It is key, then, to define the limitations of the project and its aims. in this case, the goal is to estimate the impacts of two variables linked to racial housing discrimination to modern-day home values. The limitations are many, but in particular stem from a limited cross-sectional data set and an incomplete record of racial covenants in Carlisle.

ideas/Data400Idea2.md

@@ -0,0 +1,27 @@
+## Idea #2:
+
+**Research question:** How did demographic change impact local housing markets throughout the 20th century?
+
+**Data source(s):**
+
+1.  The Federal Reserve Bank of Philadelphia published a data set with historical home price indices for major U.S. cities across more than a century.
+
+2.  The U.S. Census every ten years contains city-level population information, including total population and demographics.
+
+I plan to combine these two sources into a data set that contains population change rates, home price and rental price index for major U.S. cities every ten years. 
+
+**Models:**
+
+1.  OLS with fixed effects - since I have a longitudinal data set, a fixed effects model controls for city-specific variables that won't change over time. 
+
+2.  Difference-in-Difference - this type of statistical model is useful if I want to isolate specific moments (e.g., the Great Migration) and their impacts on home values in specific cities. 
+
+3. Bayesian techniques - since I have a highly-varied data set (cities differ from each other significantly) and a longitudinal structure, bayesian techniques might be effective. In particular, the ability to incorporate prior domain knowledge in a context like this is highly useful.
+
+  a) One possibility is to use a Markov Chain Monte Carlo technique (yes, I did have to look this up). Since I believe the effects on each city to vary substantially (say, across a normal distribution), and the data is not very large (just every 10 years) MCMC's ability to create uncertainty distributions and generate many samples is a plus.
+
+**Impact on stakeholders**: The stakes of this analysis are similar to Idea #1, but on a much larger scale. The conventional wisdom dictates that increasing population leads to higher demand, thus increasing home values and rental prices. But, in the 20th-century U.S., this was complicated by discriminatory real estate practices against immigrants and Black migrants from the southern states to the urban north. At a time when millions of native-born white Americans were building generational wealth for themselves and their families by purchasing homes in prosperous neighborhoods, African Americans and European, Latin American, and Asian immigrants were often deliberately excluded. 
+
+There are several possible outcomes of this project. If the growth of immigrant and minority populations in U.S. cities leads to higher home values and rental prices than simple population growth, this may indicate that these newcomers were charged disproprortionately high rents and home prices. If it leads to disproportionately low home values and rental prices, it may mean that patterns like 'white flight' and the subsidization of public housing tanked real estate values in these cities.
+
+**Ethical implications:** Since this project looks at these issues at such a large scale, the ethical implications are minimal. Still, it is important to take any conclusions with a grain of salt and place them in historical context.