State Space Models

All state space models are written and estimated in the R programming language. The models are available here with instructions and R procedures for manipulating the models here here.

Tuesday, November 26, 2024

US CO2 Forecasts: Will Trump II Make a Difference?



US President Donald Trump recently announced that he was pulling the US out of the Paris Climate Agreement. This is somewhat of a hollow gesture because he has done it before in the first Trump administration, only to be reversed by the next Democratic Administration. But, there are still strong indications that the Trump administration is no friend of environmental regulation and would like to undo regulations that restrain polluters (see Project 2025 below). We cannot know the future and we cannot know whether or not the Trump administration will be successful in their efforts to roll back environmental regulation, but we can make forecasts and then look back four years from now, after the Trump II administration, and see what effect their policies have had.

The time plot above is a forecast of US CO2 emissions out to 2040. The period from 1960 until the present shows how well the model works in tracking actual emissions (the solid back line). The dashed red line is the step-ahead forecast, the dashed blue and green lines are the 98% bootstrap prediction intervals. US CO2 emissions dropped in the 1980's but then continued climbing until a peak after 2000 (during the Bush II Administration). From there forward the model predicts a decline in CO2 emissions out to 2040 and to levels lower than in the 1960s. What causes the forecast decline?


The best model for US CO2 emissions is driven by dynamics in the World System. The best model was determined by comparing seven competitor models. The state of the World System is described by three state variables: (1) An overall state variable describing all the indicators, (2) A second independent state variable describing biodiversity and food production and (3) A third state variable dominated by world oil and agricultural markets. The effect of changes in the state of the World System on US CO2 emission can be seen from the shock decomposition diagram above. Growth of the World System and expansion of food production increase US CO2 production while increase in World oil prices decreases US CO2 production. 

The best way to limit US CO2 production is to allow world oil prices to increase and to stabilize the growth of the World System.

As growth of the World System stabilizes, there will be less demand for US industrial and energy production. Stabilizing world population will limit demand for agricultural products and destruction of biodiversity. Rising oil prices increase the attractiveness of renewable energy and will help build the infrastructure for electric transportation vehicles. The Paris Agreement may or may not have an impact on growth of the World System. However, US natural gas production through fracking has had an impact of keeping gasoline prices down, which is counterproductive.  

Unfortunately, US Oil Prices are controlled by the World Market. Any shocks to prices will increase inflation and create political problems for the administration in power. In a future post I will look at the World Oil Market and the expected future path for prices.




Project 2025

Project 2025 is a political initiative published in April 2022 by the American conservative think tank the Heritage Foundation (you can read the full report here). Somewhat surprisingly, in the 922 page report, CO2 emissions are only mentioned on page 378 under Conservative reforms for the Office of Energy Efficiency and Renewable Energy (EERE). Project 2025 calls for the EERE to focus on energy access and energy security rather than "entirely on the reduction of CO2 emissions." Project 2025 calls for the EERE to shift away from prioritizing "...decarbonization of the electricity sector, the industrial sector, transportation, buildings, and the agricultural sector". If the Trump Administration abolishes the EERE or changes it's mission without picking up the CO2 Emission priority in some other Federal Agency, any changes may or may not affect the trajectory of emissions which are on a strong anticipated downward path.

To put all this in context, the New York Times (here) also predicts that CO2 admissions for the EU, US, and China are reaching, at least around 2100, a steady state.




Thursday, November 21, 2024

US Inequality

ChatGPT a generative artificial intelligence model lists eight causes of increasing income inequality in the US (Wikipedia list twelve--see below). A GINI coefficient measures the level of income inequality as a percentage with 0=(complete equality) and 1=(complete inequality, one person owns all the wealth). Above 50% is considered high and the US (above 40% right before COVID-19) is considered a country in the middle range.

The history of income inequality after 1970 (in the graphic above) shows major deviations from the bootstrap 98% prediction intervals: (1) during the Ford and Carter administrations (low), (2) during the Bush I, Clinton and Bush II administrations (high) and (3) during the Trump I administration, low as a result of COVID-19


One commonly mentioned cause (regressive taxation) that I can test easily is displayed above. Effective taxation, over the period 1980-2024, is mostly steady with shocks during the Bush I, Trump (result of COVID-19) and Biden administrations.



What will happen in the future under the Trump II administration is speculative. Results from my four models (see below) are displayed above: (1) The RW and BAU models predict continuing high inequality and (2) the US and World System models show peaking and declining inequality, the greatest decline being as a result of World System forces. Clearly, the RW and BAU models would be favored by Trump and any future Republican administrations. As a practical matter, World System linkage would be similar to policies favored by the Obama Administration even though the effects were cyclical.

My tentative conclusion is that Income Inequality would stay high until the end of the Trump II administration and certainly not go down. After that, we would need policies driven by something other than Neoliberalism.

Causes of Income Inequality

Typical causes include: (1) Globalization, (2) Technological Change, (3) Decline in Unions, (4) Stagnant Wages, (5) Executive Compensation, (6) Education, (7) Skills mismatch, (8) Regressive Tax Policy, (9) Erosion of Social Safety Nets, (10) Inheritance, (11) Real Estate and Investment gains, (12) Racial and Gender Inequality, (13) The history of Segregation, (14) Rising Healthcare Costs, (15) Affordable Housing Shortage, (16) Lobbying by the wealthy and powerful, (17) Campaign Finance, and (18) Neoliberalism.

I have tested four models using the AIC: (1) A Random Walk (RW) Model, (2) A Business As Usual Model (BAU), (3) a US Economy Model and (4) a World System model. In the short run (year to year), the RW model is best. As an attractor path, the BAU model (no input variables) is best. From the discussion above, my conclusion is that, in addition to historical determinants, generating inequality is simply an output of the US Capitalist System.


 

Monday, November 18, 2024

Hardship and the 2024 US Presidential Election

 


The scramble is on among the Main Stream Media outlets to explain the factors that led to Donald Trump's victory. There seems to be agreement that the COVID-19 unemployment and the COVID-19 Inflation made the US Economy the main issue. However, my work with John Shefner and Aaron Roland on Hardship leads me to think that single-variable explanations are too limited. So I constructed a Hardship Index (similar to Shefner,  Roland and Pasdirtz, 2015) for the US from 1980-2009 (US_HARD1, see the  graphic above, click to enlarge) and made projections (with prediction intervals) for the period during and after the Trump administration.

The HARD1 index (which explains 70% of the variation in the indicators, see below). reaches a low point right during the Trump I administration (2017-2020, dashed red line in the graphic above) and then begins climbing back up during the Biden Administration (2020-2024) and should keep climbing during the TRUMP II Administration (2025-?). 

The incoming administration will have it's work cut out for it and nothing I have heard, so far, suggests that there are any plans to address hardship. I will update the index again in 2029 to see how the Trump II Administration did.

I should also comment that forces in the World system (similar to those found by Shefner,  Roland and Pasdirtz, 2015 for Mexico in the 1990's) were driving the decline in HARD1 for the US. These forces have now reversed direction and will be difficult for the TRUMP administration to do anything about.

In a future post, I will explain why the HARD1 forecast took a turning point in 2020 and what World System forces explain the longterm trend in HARD1. The primary forces in the WL20 model are found in environmental and world market conditions.

Hardship Index

The following indicators (from Shefner,  Roland and Pasdirtz, 2015) were used to construct the HARD (Hardship) Index. All data were taken from the World Development Indicators (WDI).
In addition to HARD1, two other indexes were constructed, HARD2 (dominated by Unemployment) and HARD3 (dominated by Inflation). Each index explained another 10% (0.841% and  0.933%, respectively) in the indicators. So, indeed, Unemployment and Inflation were important components of US Hardship, but the other indicators also played a role. Whether the US Electorate made a good choice in electing the TRUMP II administration to resolve their issues with hardship will have to wait and be seen over the next four years.

Forecasting Model

The model used to create the Hardship forecast (in the graphic at the start of this post) is a generalization of the Atlanta Federal Reserve's GDPNow Model.



Thursday, November 14, 2024

US Taxes

At the 1988 Republican National Convention, presidential candidate George H. W. Bush  famously said "Read my lips: no new taxes." In fact, taxes went down in the first part of the Bush I Administration (see the graphic above, click to enlarge), but started going up again and reached the lower 98% prediction interval right before the Clinton administration. And, taxes continued to go up until hitting the upper 98% prediction interval right before the Bush II Administration. In fact, taxes appear to go up during Democratic administrations and go down during Republican administrations, while bouncing around a cyclical attractor path (dashed red line above). What's going on here and what might we predict for the new Right-Wing Republican Administration scheduled to take office in early 2025?

First, I can easily predict that tax revenue is going to be lowered. I can also predict that tax revenue will return to the attractor path and increase sometime during or after the Trump administration. History clearly makes the prediction. But what is driving the cyclical attractor path and the prediction of future increases in tax revenue after 2030?

Aggregate tax revenue is being driven by growth in the US SocioEconomic system. When the economy does well, the government increases its revenue. This should be no surprise. But that's not all that is going on because actual tax revenue itself cycles around the attractor path.

What is driving the actual taxation cycle (and this should also be no surprise) are political shocks (see the graphic above, click to enlarge). Republican administrations, starting with Bush II and continuing with Trump I, pass massive tax cuts that eventually have to be rolled back (even during the same administration e.g., the expensive War on Terror starting after Sep 11, 2001 during the Bush II administration).

The assumption is that these cyclical tax cuts help the wealthy and fuel the deficit. I will look at the effects of tax cuts in a future post, but for the present I want to emphasize that tax cuts are both a political football and a countercyclical approach to balancing the economy. Tax cuts could probably be a more effective automatic stabilizer if politics was more rational.



 

Wednesday, November 13, 2024

Alternate Futures for the US

 

The graphic above (click to enlarge) shows four alternative futures for US1, an index of overall growth in the US SocioEconomic System (see the measurement matrix below). It is based on computer simulations of alternative estimated state space models. I would argue that the graphic says a lot about why US voters decided the way they did in the 2024 Presidential Election and what we can expect from the new, Right-Wing Republican Administration after 2025.

First, why did the MAGA movement embrace an extension of the American First isolationist movement? The Foreign Policy of the Obama Administration (2009-2017) was directed toward the World System and was not Isolationist. The World System input model (W) in the graphic above put the US on a slower overall growth path (dotted green line). I would argue that voters were well aware that growth of the US Economy and the US standard of living was slowing (see Economist Kathryn Ann Edward's comments here). The realization led to a Right-Wing backlash (the same thing that happened in Germany during the Inter-War (WWI-WWII) Years, see Arno Mayer's The Persistence of the Old Regime).

Second, the President-elect's policy pronouncements and transition plans suggest abandoning the World System and doing everything possible to stimulate unlimited endogenous economic growth in the US: eliminating regulation, ignoring white-collar crime, closing borders to immigrants, abandoning environmental regulation (allowing businesses to exploit free environmental resources), eliminating labor regulation (a restraint on profits), dismantling the Welfare State, slashing business taxes (another restraint on profits), eliminating funding for and control over education (a stimulus to wage growth and a limit on profits), etc.

The red and the blue dashed lines in the graphic above show possible time paths for the US Economy unleashed.  The models predict uncontrolled, unending exponential growth for the US System. They are, I would argue, a business man's fantasy. Nothing can grow forever and eventually limits will be reached (my models suggest sometime after 2050). Most of us living at the present moment, me included, will be dead by then. It will be someone else's problem.

There is another possible time path for the US SocioEconomic System: the Random Walk (RW, the solid line in the graphic above). No one can know the future. Attempts to dismantle failing US Institutions may or may not happen as imagined. The new Administration's cabinet picks, so far, are not reassuring. Essentially, in a Random Walk, today is like yesterday except for random error, actions by people who are making it up as they go along. 

I am always surprised that commentators can seem so confident about what happened in the 2024 Presidential election and what will happen as a result of it. I'm not. My advice for the future is to take defensive positions and not follow Economic Bubbles that might develop in response to crippling of US regulatory institutions. 

We know our current systems are failing and need to be rebuilt. In future posts I will look more carefully at all of these systems.


US Measurement Matrix

The graphic at the beginning of this post applies the weights from row [1,] of the state space measurement matrix to 36 indicators of US development from 1950-2010. After 2010, the results are simulated from four state space models: RW (Random Walk), W (World System input), US (components from rows [2,] and [3,] in the measurement matrix), and BAU (a Business as Usual model with no inputs).


Measurement Matrix 

     L.US.E. L.US.U. GDP.US.  GDP.C.  GDP.I.  GDP.X.

[1,]  0.1955   0.138  0.1978  0.1972  0.1961 -0.1402

[2,]  0.0669   0.200 -0.0462 -0.0554 -0.0355  0.0751

[3,] -0.0312   0.239  0.0284  0.0314 -0.0272  0.2229

      GDP.G. P.US.TBILL. P.CPAPER. P.FED.FUNDS.  P.CPI.

[1,]  0.1976     0.00429  -0.01554       0.0127 0.19773

[2,] -0.0377     0.40583   0.40460       0.3998 0.00996

[3,]  0.0541     0.07630  -0.00165       0.0986 0.02128

     P.GDP. P.SP500. V.NYSE. P.S.P.DPR. P.S.P.EPR.

[1,] 0.1967   0.1868   0.166     -0.146     -0.112

[2,] 0.0337  -0.1076  -0.131      0.144      0.128

[3,] 0.0240  -0.0277   0.165      0.321      0.384

     Q.H.Starts.   K.US.      M1      M2 P.WPI.    Q.A.

[1,]     -0.0202  0.1974  0.1953  0.1979 0.1932  0.1918

[2,]      0.0392 -0.0418 -0.0145 -0.0325 0.0612  0.0766

[3,]     -0.4666  0.0446 -0.0318  0.0472 0.1051 -0.0973

        Q.I.   O.B. P.FUELS. P.W.AG. P.W.MFG. Q.OIL.

[1,]  0.1967 -0.173   0.1834  0.1983   0.1990 -0.113

[2,]  0.0374  0.186   0.0269  0.0131   0.0125  0.312

[3,] -0.0683  0.110   0.2538  0.0489   0.0232 -0.145

       N.US. IMM.US.   U.US.   CAPU     EF Globalization

[1,]  0.1951  0.1440  0.1967 -0.141 0.1867        0.0818

[2,]  0.0746  0.0546  0.0477 -0.153 0.1096       -0.2964

[3,] -0.0642 -0.1371 -0.0589 -0.164 0.0303        0.3905

        CO2 Q.FOSSIL.

[1,]  0.180     0.129

[2,]  0.155     0.283

[3,] -0.110    -0.157


 Fraction of Variance 

 [1] 0.698 0.854 0.900


Atlanta Fed Economy Now

My approach to forecasting is similar to the EconomyNow model used by the Atlanta Federal Reserve. Since the new Republican Administration is signaling that they would like to eliminate the Federal Reserve, the app might well not be available in the future.




Hurricane Forecasting

My vision for SocioEconomic system forecasting is to follow the US National Oceanic and Atmospheric Administration's (NOAA) approach to hurricane (Economic Crisis?) forecasting using Spaghetti Models (see below).


Currently, Economic forecasting does not use Multimodel Inference but it is getting there! The best state space model for the US SocioEconomic System in the graphic at the beginning of this post is the World System (W) model based on the AIC Criterion.

Climate Change

Another comparison for what I have presented above are the IPCC Emission Scenarios. These scenarios are for the World System. Needless to say, the new Right-Wing Republican administration plans on withdrawing the US from all attempts to study or ameliorate Climate Change.




Tuesday, November 5, 2024

Is the US Presidential Election a Random Walk?

 


Today is (finally) the US Presidential Election. Can the results be predicted ahead of time? The Economist Magazine has a prediction model (here)  but most experts seem to be saying that the race is too close to call. I'll make a prediction because I have a little different approach than other forecasters. My results won't be very comforting.

First, I've created an Index of Political Support (see the graphic above, click to enlarge, the solid red line is the toss-up election line). A positive standard score indicates support for the GOP; a negative value indicates support for the Democratic party (this isn't a value judgment, just a way to present data).  Elections happened every four years in the US, but the support index is continuous and nonlinearly interpolated between elections. Support has cycled back and forth between parties over time with the mid-1970s to mid-1990s being GOP high points (except for the Carter years).

Then, I tested four models: (1) A Random Walk (RW), (2) A BAU (Business-As-Usual), (3) a US Economy Index model and (4) A World System Index model. The Index models are similar to the Atlanta Fed GDPNow Model except that the Political Support Index is substituted for GDP. If we accept the results of opinion polling (here), the US Economy should be the major determinant of Political Support with Foreign Policy (the World System Index) being the fourth most important. The BAU model would essentially mean that people cast their votes traditionally, possibly based on family upbringing. But, what would the RW model mean?

A Random Walk model, Support(t) = Support(t-1) + E(t-1), indicates that Political Support today is a function of yesterday's Political Support plus Random Error, E. It is sometimes called the Drunkard's Walk which assumes that randomness rules our lives, "...history being one damned thing after another." It would mean that Political Support is essentially unpredictable.

It turns out the the RW and BAU are the best models using the AIC Criterion. The RW or BAU result feel pretty appropriate for the current Presidential Election in the US. The voters seem to be very conflicted and acting randomly. The GOP nominee, a former single-term president, has an unsavory political and criminal record. The Democratic nominee, the current Vice President, is fighting deep-seated racial and misogynistic undercurrents in American history. The US has had only one Black president and has never had a female president.



What does this mean for our political system? David Easton's model of the Political System (here) might need to be modified. Support might be a poor method of controlling the System if it is simply a Random Walk. We have to wait for some Shock such as the Vietnam War or the 1973 Oil Crisis or the Great Inflation the Dot-Com Bubble or the Subprime Mortgage Crisis or the COVID-19 Pandemic (did I miss anything) to force the country out of random voting. Even then, it seems that the voters have poor memories (the botched handling of the COVID-19 Pandemic was during the Trump Administration and the Inflation Reduction Act did not cause the COVID-19 supply-chain inflation).

There is plenty of blame to go around. Can the Two-Party System be expected to field reasonably competent candidates who will run the country effectively? Can the Legal System hold political elites accountable for criminal activity? Can Congress impeach presidents who are incompetent? Can our Educational Systems produce citizens who can process economic and political information and make decisions?

Maybe my most depressing question is whether Democracy is compatible with Capitalism? Every day we go to work in organizations that are Authoritarian. They are not democracies. Our only experience with Representative Democracy is periodic elections which seem unable to sort out competent from incompetent candidates. Competent candidates seem to want nothing to do with political offices.

One positive trend is the Ballot Initiative (there should be more of these and some States do not allow them). Some of the big questions of the day (here) should be put to the voters. The political parties should make recommendations for each initiative and give their reasons. Only minor issues should be put to incompetent political representatives.


Data and Sources




Monday, November 4, 2024

Are Migrants Stealing Your Jobs?



The US Presidential Election is tomorrow. In a prior post (here) I presented poll results from the Pew Research Center showing that the Economy, Health Care,Violent Crime and Immigration were the major issues, of concern to more the 60% of all voters. What the polling doesn't tell me is whether these issues are interrelated. I expect they are but in no straight forward way (that's why they are political issues).

There is a mountain of research results on each issue; none of it is very easy for the average voter to wade through. Here's what the Artificial Intelligence (AI) System ChatGPT produces after surveying the results:

In summary, while there can be local variations, the data does not support a general link between higher immigration and increases in either unemployment or crime.

Interestingly, AI does not agree with Political Intelligence here. For example, the Heritage Foundation, a Right-Wing Think Tank,  argues that "...2/3 of Federal Arrests involve non-citizens". Of course, both ChatGPT and the Heritage Foundation have biases because they base their conclusions on very limited data and biased mental models. So, the assertions and arguments don't satisfy me.

In the causal diagram above (click to enlarge), I try to make some links that are missing from the literature which concentrates almost entirely on presenting numbers for Net Migration, Employment, Unemployment, Crime and Healthcare. My working hypothesis is that all these variables are being driven by Shocks and internal dynamics within the US Economy and the World-System. Causal links with "?" indicate that the direction of causation is unclear, at least to me.

First, let me summarize my results before getting into details. The models I will use are similar to the Atlanta Federal Reserve's Economy Now model but I've expanded the variables of interest to include Crime, Net Migration and Healthcare (the EconomyNow app, which you can download on your cell phone, covers GDP, Wages, Employment and Consumer Prices). It's the same model I have used in prior post (here and here). To summarize my findings:

  • Crime Crime rates have been declining since 1974 with 1981 being the peak for violent crime rates. In the short run, crime rates are a random walk or driven by US Economic performance. In the long run, crime rates are driven by the World-System, particularly events in Latin America (this result should appeal to the Right-Wing). But, it is not driven by immigration.
  • Government Healthcare Expenditure In the short-run, healthcare expenditure is also a random walk and driven by events in the World-System. It also is not driven by immigration.
  • Unemployment In the short run, and the long run, unemployment is driven by events in the US Economy and not by immigration.
  • Net Migration For net migration (more people entering than leaving the country), I don't get a very clear picture (this is probably why it is a perfect issue for wild political distortions). The best model is a Business-As-Usual (BAU) model, likely the result of US immigration restrictions. However, it does not support the assertion that the US has Open Borders.
If you are still interested, here are the details.

Crime

The graphic above (click to enlarge) shows violent crime over time across administrations (left panel) and violent crime after a NETM shock. Crime rates have been declining since a high in the late 1970s, bottomed out during the Obama Administration, began to climb again during the Trump administration and leveled off during the Biden Administration. A shock to NETM (right panel) decreased crime rates but not by a lot (the y-axis of both graphs is measured in standard score units).

The shocks to CRIME from World-System events are displayed in the graphic above. The two major shocks were the September 11 Attacks in 2001 during the Bush Administration and the COVID-19 shock during the Trump administration.

Government Healthcare Expenditure

The graphic above (click to enlarge) shows Federal Government Healthcare expenditures across administrations (left panel) and shocks to Health Care expenditure from NETM in the right panel. You can clearly see that Obama Care (the Affordable Health Care Act of 2010) "bent the curve" on US Healthcare expenditures. It is interesting that the time path of GHEALHX is best described as being driven by World-System events, specifically the state of the economy in Latin America. To my knowledge, this link has never been demonstrated before and should be considered controversial. However, shocks from NETM, although increasing GHEALHX, only have minor effects. And, the COVID-19 shock during the Trump Administration can be clearly seen. It would be fair to conclude from the graph above that abolishing Obama Care (a plank of the Trump Administration and Project 2025) would "unbend" the GHEALTHX curve.

Unemployment


The graphic above shows Unemployment (LU) by administration (in the left pane, driven by the US Economy) and shocks to LU from NETM in the right pane. You can clearly see the LU shocks at the beginning of the Reagan Administration, the beginning of the Clinton Administration, the beginning of the Obama Administration and the COVID-19 LU shock during the Trump Administration. Shocks to NETM (right pane) actually decrease unemployment but the effects are not very large.

Net Migration


The graphic above (click to enlarge) shows Net Migration (NETM) being driven by the World-System (left pane) and a shock to NETM from LU. NETM was particularly high in the Bush I and Clinton administrations as a result of World-System events. Shocks to LU decreased NETM, meaning that when Unemployment was high in the US, NETM was reduced (no jobs for migrants).

Summary

I have already summarized the results above, but I hope that the displays of Crime rates, Government Healthcare Expenditure, Unemployment, and Migration (GDP and Inflation here and here,  respectively) might help an interested voter make decisions about the upcoming Presidential Election.

One point that I hope is also made by the graphics is that "cherry-picking" isolated numbers and percentages (as seems to be the case in political rhetoric and Media presentations) mean very little without historical analysis. All the data presented above (and a lot more) are available from the World Bank Development indicators (here) and is freely available to you if you have other topics of interest.

If you would like to see some other series presented as an Economy Now model, please contact me. I am happy to post about other topics (otherwise, I'll just keep pursing my interests).