Can you model social change as linear line graphs? The Cincinnati City Council presentation

CATEGORY: GOVERNMENT PROJECTS: CITY

Daniel Chen, DrPH, George Washington University

Source: Cincinnati City Council
Read time: 2.4 minutes plus a short video
 
This Happened
A professor of public health demonstrated a “system dynamics modeling” software program as one component of a larger plan to persuade Cincinnati city government to implement systemic social changes. The professor was one of six presenters in the group who came before a committee of the Cincinnati City Council on April 2, 2024.
 
Who Did This?
Daniel Chen, DrPH, from the Global Health Department of George Washington University’s school of public health, presented the software. He has been the first author on one peer-reviewed paper and secondary author on five others focused mostly on trauma-informed care.
 
The Presentation
Chen’s software demonstration was one component of a larger strategy claiming that adverse childhood experiences (ACE) cause permanent brain and body damage that leads to a wide variety of mental problems, physical diseases, and social dysfunctions. Hence, the group argued, government investments are needed to relieve these stressors.
The purpose of the software modeling is to provide a tool for policy makers to make informed decisions. Chen said his system dynamics modeling was based on 300 variables and about 500 equations.
For the first simulation, Chen input a “policy lever” on the model’s dashboard—a program to provide financial assistance for renters to become homeowners. The model then spit out a line graph to show how the percentage of homeowners increased every year.
A bit later, Chen input another policy lever to provide financial assistance to prevent foreclosures, and again, this produced a line graph, this time showing, predictably, fewer foreclosures every year.
Lastly, Chen ran both policy levers simultaneously to show how they would impact “population health” by increasing the number of individuals with Good or Excellent Health.
 
 
Analysis
One concern about modeling of complex public health problems, in general, has been unreliability. Take, for instance, the infamous Imperial College London model at the beginning of the COVID-19 pandemic that predicted peak mortalities above 215 deaths per million in Great Britain. This announcement played a large role world-wide to drive harsh distancing measures and lockdowns. In reality, Great Britain flattened the curve at 13.9 deaths per million [1].
The COVID model failed because the humans using it made pessimistic and unchanging guesses about the infection rate, death rate, time to recovery, and the rate of passing the virus between persons. In addition, it did not account for motivations to misattribute deaths to COVID, how lockdowns would prolong the epidemic, or how therapeutics would shorten it. Put simply, almost nothing in nature or human society changes in a straight line.
 
The video (below) shows the first simulation. As noted above, Chen inputs a “policy lever” on the model’s dashboard—a program, called ADDI, to provide financial assistance for renters to become homeowners. He set the lever at 1% in the neighborhood of Avondale starting in 2025, meaning that of 4,085 renters, 40 would become homeowners the next year, and 1% more would be added in each successive year. The percentage of homeowners increased from about 25% in 2025 to a nearly miraculous 50% in 2055. As you watch the line graph in the upper right, note how it is a perfectly straight upward-trending line for thirty years.

Another concern, and perhaps the inherent fatal flaw of models, is that they almost never can predict how individual behaviors vary over time. Consider Daniel Partick Moynihan’s famous scissors graph in his 1965 report (below). For decades, as unemployment for nonwhite males rose or fell, applications for welfare logically rose or fell in tandem. But in 1960, something changed; unemployment dropped but applications for welfare increased. Moynihan attributed this to a shift in Black family structure towards single-parent households. Put simply, almost nothing changes in the real world as policy makers intend.

In the graph above, note how, starting in 1960, unemployment and applications for Aid to Families With Dependent Children (AFDC) welfare assistance suddenly went in opposite directions [2].

​When showing models in this manner to legislators, it can easily be misleading:

• The legislators were not told that this model assumes it is just as easy to add the second 1% as it is to add the first 1%.
• Legislators were not told that desirable housing stock to own might not be available.
• There was no consideration of how real estate markets react to an influx of more buyers (i.e., prices go up) and how that would impact the program.
• Legislators were not told that the model seems to assume that renters with insufficient funds to buy a house often also may have insufficient funds to sustain homeownership that requires property taxes, major repairs, and insurance.
• Population health was never defined.
• It was not explained how home ownership leads to improved health.
• Incredible gains were demonstrated in line graphs but the software does not track the amount of funds invested or future cost savings.
• There was no consideration about whether families versus non-families or if young versus old renters behave differently, or dozens of other variables that could impact rental and housing markets.

 
Chen and his group pitched the model as being based on empirical evidence, when actually it’s based on a utopian idea that everything goes according to their plan.
 

Like Trauma Dispatch? You can subscribe to our email notices of new posts on this page.
 
References
[1] Boretti A. After Less Than 2 Months, the Simulations That Drove the World to Strict Lockdown Appear to be Wrong, the Same of the Policies They Generated. Health Serv Res Manag Epidemiol. 2020 Jun 17;7:2333392820932324. doi: 10.1177/2333392820932324. PMID: 32596417; PMCID: PMC7301657.
 
[2] Daniel Geary, The Moynihan Report: An annotated edition, The Atlantic, September 14, 2015. Accessed May 4, 2024.