We had an insightful conversation with David Stroh, Martijn Eskinasi, and Kaveh Dianati at our latest seminar. We learned about how using causal loop diagrams based on simple archetypes, illuminate interdependencies across housing availability, wealth inequality, homelessness, and economic development. We explored the dynamics of the housing crisis in London and how System Dynamics is helping the Dutch Ministry of Interior make better housing polices. 

Learn more about the Seminar Series.

Responses to Housing Dynamics Panel Questions

DAVID STROH:

David’s cautionary tale suggests avoiding early complexity in engagement with the problem owner. Does he have a method for keeping it simple?

I use a four-stage process to engage stakeholders in applying systems thinking to catalyze real change. Systems analysis, the second stage, is just one part of this overall change strategy. The analysis itself primarily draws on systems archetypes to simplify complexity. At the same time, building on individual archetypes and identifying multiple ones provide complexity while retaining the power of illuminating recognizable storylines. For more information, see my book Systems Thinking for Social Change (Chelsea Green, 2015).

How do you collect data for these projects and how you do model assumptions about specific issues you can’t collect data for?

Primary data comes from interviews with a diverse set of stakeholders – both decision makers and people closer to the front lines. Secondary data comes from written documentation provided by the stakeholders. We (often Mike Goodman and I) enlist a modeling team drawn from these stakeholders to test and refine the initial models we build – often by having people first build their own simple models based on what we see to be the relevant archetype(s). We also collect and add mental models to the qualitative causal loop diagrams to help them see how their mental models impact the system dynamics.

Can you name examples of elites fanning ethnic conflicts? Is there really an elite behind this or is this potentially too, a structural issue?

One need look no further than the Republican Party in the U.S., and the elite donors who fund them, for an example of fanning ethnic tensions – between whites and not only blacks but also immigrants. I first came across this dynamic nearly 20 years ago in advising a group of NGOs in Burundi who was looking to rebuild civil society thereafter the war between Hutus and Tutsis had almost destroyed their country. To me, the existence of an elite IS a structural issue. See for example my recent article “Overcoming the Systemic Challenges of Wealth Inequality in the U.S.” in The Foundation Review.

How do you deal with the delay of the long-term solution where the workers are needed within a short time to keep the economic development happening?

In any Shifting the Burden dynamic, the key is to reinforce investment in the long-term solution while either discouraging the quick fix or seeking to mitigate (or even reverse) the negative impacts of the quick fix on the fundamental solution.

David – curious why you leave out the link polarities. As a System Dynamicist, this makes it harder to validate your assumptions.

I leave out polarities in my presentations to stakeholders because they are covered by how I translate loops and archetypes into everyday recognizable language and stories. I occasionally use different colored links for + and – since they are a little easier to make sense of.

In Los Angeles, we have a serious homelessness challenge which is quite visible. Can you comment on how your work can address the challenges here?

The change process is as important as the insights it generates. I don’t know enough about the politics of LA, what has been tried so far, who has been involved and who has been excluded, etc. to comment.

to all speakers: advocators of big data research state that with the amount of data available nowadays it is more promising to look at correlations than causation

There is never enough data nor will there ever be. The problem in influencing how people think is often not insufficient data, but insufficient attention to the non-cognitive aspects of systems change – such as the emotional, behavioral, and spiritual issues involved. See for example the last chapter of Systems Thinking for Social Change and the blog post “Systems Thinking: It’s Not (Just) What You Think”.

KAVEH DIANATI

How do you collect data for these projects and how you do model assumptions about specific issues you can’t collect data for?

In the case of the work on London’s housing crisis, most of the data were collated from publicly available government statistical sources. Often, data is incomplete and/or contradictory between different sources and you have to address these issues on a case by case basis. In the case where data for a particular key parameter is not available, a reasonable expert estimate can be used to let the model run, and sensitivity analysis must be carried out to understand the impact of the uncertainty and results and policy recommendations.

Kaveh: to what degree do you expect that your model/findings would also apply to other cities (with different parameterization)?

To a large extent, the dynamics described in this work apply generally to attractive metropolitan areas in liberal advanced economies, where housing is regarded as an attractive investment asset and where housing is financialised, and where unregulated mortgage lending by commercial banks is driving house price inflation.

What would be the policy leverage points to keep the system “in balance” at a reasonable affordability level for housing, avoiding the exponential growth and boom and bust cycles?

As explained during the presentation, key recommended policy levers are macroprudential policies such as imposing restrictions on loan-to-value and loan-to-income ratios, as well as requiring banks to base their property valuations on a moving average of past prices when issuing loans. In parallel, affordable housebuilding programs must take off again to ensure access to housing to the lowest decile households.

What role do zoning rules play in the reduction of housing construction in London?

There is a great recent paper published on that by Gallent et al (2021) titled “Is Zoning the Solution to the UK Housing Crisis?”. Zoning has been suggested by various authors as a policy to accelerate supply, and my modeling shows that easing planning restrictions can have a significant impact on housing supply and therefore slow down the worsening of affordability. So, zoning is very important. However, note that, as I explained during the presentation, this would have no impact on mitigating the boom and bust cycles, because these are not related to the supply side but rather to the demand side. Furthermore, housing supply is necessarily finite while demand that is backed by newly created mortgages and international footloose capital is virtually infinite, and therefore without regulating the demand side, supply-side policies are not enough to address the housing crisis.

Can you comment again on the large dip in the near future projection around 2045 – 2050?

This has to do with the ‘bust’ period of the recurring boom-and-bust cycles, which happens when the burden of housing-related housing debt becomes unsustainable, putting pressure on household consumption, savings, economic investment, productivity growth, and the economy as a whole over the long term, feeding back to restrict new mortgage lending and demand for housing, triggering all the reinforcing mechanisms (Investment Loop and Housing-Finance Loops) in the opposite direction of growth and leading to a precipitous fall in house prices and mortgage lending.

A general question. When starting out on building a model, is it the speaker’s typical first step to hunt for the available data? What was the first thing done when tackling such a wicked problem as housing?

The first step in the System Dynamics method is problem definition via formulating a number of reference modes, i.e., past behavior of key variables over time. So yes, you could say that at the outset, especially in a quantitative modeling study, I would look for historical data on key variables of interest and look at how they have been developing over the past. The next step is then to try and understand these developments from a systemic feedback-centered perspective.

To all speakers: advocators of big data research state that with the amount of data available nowadays it is more promising to look at correlations than causation (‘let the data speak for us’); this would somehow remove the inherent bias of our models; what is your opinion on this criticism?

First, correlation methods are unable to explain the “why”s, i.e., explaining why things behave the way they do. In other words, they are unable to tell interesting ‘stories’ that capture our collective imaginations and can mobilize society’s forces towards our common goals. The feedback-oriented explanations offered by System Dynamics offer such relatable narratives that are powerful communication media. Secondly, since correlation methods base their forecasts solely on past behavior, and do not model the structure of the system, more often than not they involve extrapolation of current trends and are unable to foresee or warn about potential reversals in trends.

How do you validate these housing models? Do you use publicly available data/records?

Yes, for behavioral validation. For structural validation, you need to follow the best practice suite of tests available (e.g., sensitivity tests, extreme condition tests, etc.) as explained clearly by Barlas (1996) “Formal aspects of model validity and validation in system dynamics”.

The London CLD seems to focus on financing the DEMAND side. What about the SUPPLY side? More construction would reduce prices. What is inhibiting construction?

There is extensive discussion about that, in particular around issues such as increasing concentration and consolidation in the housebuilding industry, as well as issues around land banking. These aspects were fully addressed in my thesis but excluded from my model, due to reasons discussed extensively in the thesis, to be available hopefully soon online.

Kaveh, you indicated things can’t grow forever so when in the future does the model crash?

Rather than focusing on “when”, I would like to emphasize that the tendency for boom-and-bust cycles is embedded into the structure of the system, but no one can tell you when exactly. However, if ‘your doctor tells you that you will have a heart attack if you do not stop smoking, this advice is helpful, even if it does not tell you exactly when a heart attack will occur or how bad it will be’ (Meadows, Richardson, & Bruckmann, 1982, p. 279).

MARTIJN ESKINASI

David’s cautionary tale suggests avoiding early complexity in engagement with the problem owner. Does he have a method for keeping it simple?

In general, it is good practice to stick to small comprehensible models. I’ve got bad experiences with large models, which can become unmanageable quickly. If possible, implicit System Dynamics models in your field of interest can be translated to stocks and flows. This also helps comprehensibility in contact with stakeholders

How do you collect data for these projects and how you do model assumptions about specific issues you can’t collect data for?

In the Netherlands, there is also lots of statistical data available. Expert guesses are another useful source. But is it not only about data, it is also about system structures.

What would be the policy leverage points to keep the system “in balance” at a reasonable affordability level for housing, avoiding the exponential growth and boom and bust cycles?

I’d agree with Kaveh here too. Lots of other works point in that direction. In NL, plenty of feasible scenarios were elaborated for reducing procyclical housing finance. System dynamics will add up here, but there are no breakthrough insights to be found there anymore. But it might be interesting to model political processes why is it hard to reduce or even abolish abundant finance schemes

I would be interested to know more about the link between housing affordability and car-oriented transport policies leading to car dependence and urban sprawl.

Some work was done on that. See Chapter 3 in my Ph.D. thesis. 

How would you ensure that your own bias as a researcher does not lead to unintended consequences? Would you conduct this research in parallel to another researcher, for instance?

That is exactly why I am a proponent of close cooperation with other researchers. Repenniing made a strong case already in 2003.

A general question. When starting out on building a model, is it the speaker’s typical first step to hunt for the available data? What was the first thing done when tackling such a wicked problem as housing?

Not necessarily. I was mostly looking for template models and institutional structures. Data comes later, though in NL we are relatively spoilt with data.

Are you always using your “goal” as one of the elements in your system? ie: The ability to pay for quality housing

Not always. What about finding structures capable of generating reference modes of behavior?

To all speakers: advocators of big data research state that with the amount of data available nowadays it is more promising to look at correlations than causation (‘let the data speak for us’); this would somehow remove the inherent bias of our models; what is your opinion on this criticism?

In public policy, there is nearly always a need to improve system performance. There will be a gap between observed and preferred outcomes and a theory of how measures propagate through the system and reduce the gap. Data won’t tell you that. And data can be interpreted in order to fit any frame, especially if there is a lot of.

How do you validate these housing models? Do you use publicly available data/records?

I agree with Kaveh here

Watch the recording below