Should The Random, Drunkards Walk Policy Model be Taken More Seriously?

It's hard to look at the current policy environment in the US or in Argentina and seriously think about Rational Policy Models. In the current Trump II Administration, events seem to happen randomly: Tariffs being imposed and withdrawn; Immigrants being expelled and then brought back by the courts; Universities having research funding withdrawn for political reasons and then given back after a shake-down; National Guard Troops being deployed to the streets of states and cities with democratic Governors and Mayors and then being withdrawn; the Supreme Court supporting some radical policy measures and not others, etc. etc. What Rational Policy Models can explain all of this?

Q(t) = Q(t-1) + E(t-1)

The Random Walk model is really simple (equation above) and easy to estimate from a statistical standpoint (there is only one parameter in the model and its value is given as 1.0): Tomorrow is like today except for Random Error or History is One Damned Thing After Another or The Drunkards Walk: How Randomness Rules Our Lives.

When using Multimodel Inference (MMI) and selecting models based on the Akaike Information Criterion (AIC), the Random Walk (RW) model should always be one of the competitors. Any model I present will always be tested against the RW. For many countries and time periods, especially using year-to-year data, the RW is often the best short-term model. Yet, we continue to hold on to the belief that there is (rational?) causal structure in macro-social systems.

In Systems Models, there are many kinds of RW models: RW with drift, RW with Feedback, Partial RW (some diagonal elements of the System Matrix are 1.0), in addition to the pure RW model:

The pure RW model above is from Argentina. The only distinguishing feature of the model (from any other country) is the initial state:

If the on-diagonal elements of F take on values other than 1.0 or the off-diagonal elements of F take on values other than 0.0 or if there is a fourth column to F (a constant vector capturing drift) we have the other RW models. To decide which one is most appropriate in a given historical period, I use the AIC criterion.

To supplement Universal Growth Theory in Evolutionary Economics (Random Walk -> Malthusian Growth Model -> Neoclassical Growth Model), I would add an initial RW stage (above) that can be revised at any time in history when the Environment changes and a random search for new approaches is needed.

Or, in Marxist Economics, as the initial stage prior to Feudalism.

Or in Post-Modernist variants (above) with a feedback loop returning to RW when the Environment Changes.

Below in the Notes are some concrete examples. 

Notes

Examples:

• Is the US Presidential Election a Random Walk?
• Trump Takes the US On a Random Walk
• Mexico Takes a Random Walk
• Argentina Takes a Random Walk
• What We've Learned about Economic Bubbles
• A Random Walk Among the Undead
• The Drunkards Walk and the Moment of Truth
• What Caused the Financial Crisis
• Policy Models, Randomness and Free Will
• The Deficit and the 2012 Federal Budget Overview
• The Edge of Chaos in Switzerland

World-System (1960-2010) Controlling the Argentine Economy.

 

Economists* should probably admit that they don't know how to control the Economy. When an economist and politician such as Javier Milei gets elected as president of Argentina in 2023 and starts waving a chainsaw around as a symbol of cutting government, critics start to get nervous.

In a prior post (here) I found that Latin American Integration could stabilize the economy of Argentina. Unfortunately, Latin American Integration has been tried before and mostly failed, probably because Latin American has it's own problems with instability. The problem leaves me searching for other Geopolitical Alignments. In this post, I'll look more carefully at the ARL20 BAU model, that is, turning inward and concentrating on Business as Usual.

Why all the hand-wringing over Argentina? Millie has become the poster boy for the US Right Wing after giving a speech (with Elon Musk) at CPAC in 2025. The current Trump Administration and it's Department of Government Efficiency (DOGE), originally chaired by Elon Musk, seems intent on copying Milei's shock therapy. Unfortunately, or predictably, it seems that Milei's shock therapy has failed and will require a Bailout from the IMF and the US. So, it seems important to ask the general question about how (if at all) an unstable economy such as Argentina can be controlled?

The argument of Shock Therapy is that if we get the Government out of the economy, the Free-Market will take over and ensure prosperity. In other words, the free market will control the economy. If you have problems, it is because the market is not free of government interference. The "free market" assertion can be proven wrong (here).

From the standpoint of Systems Theory (where we have the best understanding of how to control systems), the first step is to establish an attractor path** among the competing Geopolitical models.

The attractor path (AP) for the ARL20 LAC Input model is presented above (dashed line) with the actual historical data (solid line). There are few serious deviations from the attractor path, except for AR3 (the definitions for the state variables are given in the Measurement Matrix below in the Notes) around 1975 and after 2000. However, AR2 and AR3 are Environmental-Unemployment-Globalization controllers and should be relatively stable over time.

The attractor path (AP for AR2 and AR3) for the ARL20 BAU model is presented above. Notice that it differs from the ARL20 LAC Input model AP. The period from 1980 onwards shows departures for both historical feedback controllers. For AR2=(LU+EF+KOF-CO2), unemployment, Ecological Footprint (EF) and Globalization (KOF) departures were very large relative to Emissions (CO2). For AR3=(EF+HDI+CO2-KOF-LU), departures for Globalization (KOF) and Unemployment (LU) dominated. 

The difference between the two time plots above shows that Latin American forces caused the departures and that the two historical feedback controllers (AR2 and AR3) were unable to correct the system with a period of decades (see the Eigen Modes.

Also, in the DCM model (see the Notes below), these two historical feedback controllers interact: shocking AR2 increases AR3=(EF+HDI+CO2-KOF-LU);  shocking AR3 decreases AR2=(LU+EF+KOF-CO2). In other words, Globalization, Unemployment and Environmental degradation are used as historical feedback mechanisms to control the Economy. 

The effects take decades to work out. The historical feedback controller coefficients are weak (the off-diagonal elements in the System matrix below). The feedback effects in the full ARL20 BAU model (including growth components) are also weak.

Exercise 1: Strengthen the feedback coefficients in the ARL20 BAU model and see if you can better control the system.

Controlling how the system responds to Unemployment, Globalization, Environmental degradation will be a great deal more challenging than cutting Government spending, but better system control is needed in Argentina and a free market will not accomplish everything that is needed (here) while Latin American Integration is a long way off in the future.

Notes

* Part of my Interdisciplinary degree at the University of Wisconsin--Madison (1981) was in Economics, so I should probably include myself in this criticism!

** The attractor path of a Dynamics Components State Space Model (DCM) can be computed with a free simulation starting from historical initial conditions (see Pasdirtz 2007). The free simulation that minimizes the AIC among competing Geopolitical models is considered the "best" attractor path, using some historical judgment when competing attractor paths are not well separated.

ARL20 model AIC summary:

ARL20 model Measurement Matrix AR1=(Overall Growth), AR2=(LU+EF+KOF-CO2), AR3=(EF+HDI+CO2-KOF-LU):

ARL20 model State Space Time Plot:

ARL20 BAU model Historical Feedback Controllers System Matrix:

ARL20 BAU model Historical Feedback Controllers Shock Decomposition: 

ARL20 BAU model modes:

ARL20 LAC Input model modes:

World-System (1960-2100) A Stable Geopolitical Alignment for Argentina?

Argentina just entered what Paul Krugman has called (here) a Classical Monetary-Financial Crisis. The Peso is collapsing, the Central Bank is trying to defend the currency (but is running out of money) and the US is ready to support Argentina with a $20 Billion Swap Line (Bailout). What triggered the crisis? According to Google AI:

After discussing the economics of the crisis, Paul Krugman concludes:

In this post, I'm going to argue that the "alternative strategy" involves Geopolitical Realignment within the World-System. The forecast plot at the beginning of this post shows the ARL20 model being driven by three alternative input systems: (1) the Random Walk (RW, no input, muddle through), (2) the USL20 model (Hegemonic Dominance) and (3) Latin American Integration from the LAC20 model. US attempted dominance of Argentina has been going on since after World War II and it is about to fail (ARL20 model US Input [96.74 < AIC = 106.6 < 117.3]). 

The best Geopolitical model for AR1 (Growth in the Argentinian Economy, see the Measurement Model in the Notes) involves integration with other Latin American countries [63.72 < AIC = 123.9 < 154.1].

Not only is the future forecast for Argentinian growth better under Latin American Integration, but also the system response to shocks is better than the ARL20 BAU model and the integrated system is stable. Under the BAU model (an unstable system), positive shocks to the system reduce growth; under the LA Integration input model (see the Notes), positive shocks increase growth as would be expected.

However, we can't give up on the BAU model (here). Latin American Integration has been tried before and has a history of failure. After World War II  the US tried to drive the movement but simply ended up as the dominant Hegemon. A question I will investigate in a future post is whether Latin American Integration would benefit other countries in the region. Until stable Integration does benefit enough countries (a long time in the future?), the Economy of Argentina will likely continue lurching from one crisis to another (unless changes are made in the BAU model).

You can experiment with the LA20 BAU model here. Suggestions are given in the code for how to stabilize the model.

Ex. 1.0 Can you find a way to eliminate cycles once the model has been stabilized? 

The solution to this Exercise can be found in the LA_TECHP model which I will describe in a future post. 

Descriptions of how the Dynamic Component State Space models are constructed are given in the Boiler Plate.

Notes

Reading:

• Why is Trump Bailing Out Argentina -- Paul Krugman (2025)
• The Bubbly Economy of Argentina
• Seven Futures for Argentina
• Austerity and Anti-Systemic Protest (Argentina)
• Unstable Systemic Cycles in Latin America

Please see the Boiler Plate for a description of the DCM State Space Models.

 ARL20 model  Measurement Model:

The ARL20 State Space includes three Historical Feedback Controllers: AR1=(Growth-EF) Overall Growth balanced against the Ecological Footprint (EF). AR2=(LU+EF+KOF-Q) an historical feedback controller balancing Unemployment (LU), Ecological Footprint (EF) and Globalization (KOF) against Output (Q). AR3=(EF+CO2-KOF-LU) an historical controller balancing Ecological Footprint (EF) and CO2 Emissions against Globalization (KOF) and Unemployment (LU). Together, the components explain 98.8% of the variation in the indicators.

 ARL20 model  LA Input Model:

Compare the LA Integration System matrix (above) with the  ARL20 BAU model. Notice that (1) all the coefficients in the System Matrix (F)  are reduced in size and (2) the largest effects in the Input matrix (G) are from the LAC Unemployment (LU) controller (LU-Q-EG=1.57) and the LAC Labor Force Controller LA3=(N+L-CO2-Q=1.6)  on AR2 (Argentina's unemployment Controller).

 ARL20 model  LA Shock Input:

The shocks presented above are from the LACL20 Model (see below): (1) A shock to growth of the Latin American Region increases growth in Argentina. (2) A shock to the Unemployment (LU) controller (LU-Q-EG) reduces growth. (3) A Shock to the Populaton-Labor Force Controller (N+L-CO2-Q) increases growth. An explanation of Historical Feedback Controllers can be found in the Boiler Plate.

LACL20 Model Measurement Model:

The LAC State Space contains one Overall growth component (LA1, all indicators weighted positive and approximately equal) and two Historical Feedback Controllers (see the Boiler Plate): LA2=(LU-Q-EG) an Historical Unemployment controller balancing Output (Q) and Energy Use (EG) against Unemployment (LU), explaining 2% of the variation in the indicators. LA3=(N+L-CO2-Q), an Historical Labor Force controller balancing Population Growth (N) and Labor (L) Force with CO2 Emissions and Output.

World-System (1960-2010) The Globalization Controller and the UK Labor Market

 

UK3 = (LU+L+N+EF - KOF-HDI)

Notes

UKL20 Measurement Matrix:

World-System (1960-2010) What Explains Labor Force Growth in the UK?

Notes

UKL20 Measurement Matrix:

EUL20 Measurement Matrix:

AIC (Akaike Information Criterion) statistics:

• EU [1720 < AIC = 1725 < 1729]
• RW [1441 < AIC = 1455 < 1471]
• BAU [1789 < AIC = 1798 < 1807] 

 

World-System (1960-2015) How is the French Labor Market Controlled?

 

In an earlier post describing how the Economy of France works (here) I showed that the Historical Labor Market Feedback Controller was described by:

FR3=(LU-L-N)

Where LU=Unemployment, L=Labor Force and N=Population. What the controller says is that Unemployment is monitored relative to Population and the overall Labor Force. The historical FR3 controller which explains 3% of the variation in the FR State Space (see the Measurement Matrix in the Notes Below) peaked between 1985 and 2000 and then declined after that. Starting in the 1970s, this was a very high period for French Unemployment which took almost 30 years to correct. This historical bulge in Unemployment gave the Economy of France a reputation of having persistent high Unemployment which is not supported by the data. But what caused the Unemployment Bulge?

The Unemployment Bulge from 1975 to 2010 in France was caused by the World System, specifically by shocks to Agricultural and Oil Markets.

In this post I will show results from the FRL20 Model to explain the conclusion. The graphic above also shows the attractor path for FR3 (dashed red line) driven by the World System (WL20 Model). The attractor path suggests that this source of World System shock is over for the immediate future. However, the system is sensitive to World Agricultural and Oil Markets. It seems unreasonable to forecast that these markets will adjust less radically in the future.

Keep in mind that the FRL20 BAU model is approaching a steady state around 2035 (graphic of the FR1 Growth Component above). In a steady state economy, there is a constant stock of capital and people. The only variability in the FR3=(LU-L-N) historical feedback controller is essentially Unemployment (LU) and possibly Labor Force participation, L. Unemployment might also stabilize, but that is not a certainty (Goldman-Sachs predicts that AI could replace 300 million full-time jobs, here, between 2025 and 2030). On the other hand, current AI Energy Demands might not be sustainable (here). And, needless to say, we are dealing with the predictions of a model--we will have to wait for the Future to know what actually happens.

You can experiment with FR3 in the FRL20 Model here.

Notes

The Measurement Matrix for the Economy of France is presented above. All indicators (columns of the matrix) are taken from the World Development Indicators. FR1=(Overall Growth), FR2=(Historical Environmental Controller) and FR3=(Labor Market Controller). For information about how the models were constructed, see the Boiler Plate.

World-System (1950-2010) Has Austerity Helped or Hurt the French Economy?

 

The French have gone to the streets protest the Austerity measures of Emanuel Lacron's government (here). What's going on and what does it mean for the French Economy

The history of Austerity in France is captured by the FR_AUST index (see the Notes below). There are three components: (1) Education and Military expenditure compared to others (68% of the variation), (2) Health expenditure compared to others (10% of the variation) and (3) Overall Growth (4% of the variation). All the indicators reached a low point between 2010 and 2020 but then began accelerating (Emanuel Macron has been in and out of office since 2012 and has become associated with the return to Austerity .  

I can explore the impacts of Austerity on the FRL20 model using a Shock Decomposition Impulse Repspone (above). Let''s just look at growth (FR1) for the moment. Shocks to Austerity, after a few years, increase growth (the effects are not large) but the effects of Overall Growth in Austerity (column three) are all negative. So, the effects are complicated.

Given that the French Government is about to fall again (here), it might be prudent for politicians to find other ways of conrollingt the economy and leave Austerity at the low levels that were achieved between 2010 and 2020.

Notes

The AUST -> FR Systems Model:

The Austerity Indicators and Index (see Shefner et. al. , 2005):

The data for the AUST index is taken from the World Development Indicators (WDI). The indicators and definitions are listed in the table above. NOTE: AUST is entirely measured by budgetary categories as percentages; the cyclical nature of the index is a result of percentages hitting up against limits [0%,100%].

The AUST index contains three components that explain 94% of the variation in the indicators. 

AUST1 = (0.433 GED + 0.4571 MIL - 0.4477 G - 0.393 GE - 0.4701 GH)  

AUST2 = (0.822 GHE - 0.357 GED - 0.377 GE) 

AUST3 = (Overall Growth) 

AUST1 and AUST2 are historical feedback controllers for the budgetary categories defining Austerity. AUST1 focuses on controlling Education, Military expenditure, Overall Government Expenditure and Health Expenditure. AUST2 focuses on controlling Health and Education Expenditure.

In the Economy of France, Austerity, Debt and Globalization (KOF) are closely related. The relationship can be seen from the Measurement Matrix above when DEBT and WorldGlobal (KOF) are added to the model. In future posts, I will investigate all the indicators of Neoliberalism in France.

The state space of the French Economy is dominated by three components explaining 98% of the variation in the underlying indicators: 

FR1=(Overall Growth)

FR2= (CO2+EF-KOF)

FR3=(LU-L-N) 

FR2 and FR3 are Historical Feedback Controllers regulating Environmental Impacts of Globalization and Unemployment, respectively. EF is the Ecological Footprint and KOF is the Index of Globalization.