Im glad I found this forum because I've had this idea clanging around my head and a paper I've found recently has given me some clearer ideas about how one might go about this.
In short, I was thinking about how it might be possible to utilize input-output analysis to plan precision industrial actions as to maximize capital loss with respect to worker's organizational capacity; a strike in a critical sector can percolate through the economy and be as damaging as a larger strike in a more downstream sector. Additionally, planning is critical to any socialist economy so having a way to embed it in the workers movement beforehand is a major political boon.
The attached paper computes the domar-weighted katz centrality of the input-output matrix of the U.S economy as a measure of the sensitivity of a particular sector to supply shocks. This analysis has some pretty interesting results; one is that car parts manufacturers are amongst the most sensitive whilst car manufacturers are amongst the least sensitive. Additionally the eigendecomposition analysis gives an interesting interpretation of the eigenvectors; the eigenvector corresponding to the largest eigenvalue is the 'direction' of the longest lasting supply shock because successive supply shocks along this vector have the least attenuation (because it corresponds to the largest eigenvalue).
I wanted to bring attention to this paper on this forum because I have a couple of questions; I am an engineering PhD student so I can glean the gist of the paper but I am not generally knowledgable in economics and graph theory, so hopefully some computer science and economics folks on here can help me better understand this work. God willing, we may find a good way to apply it 🙂
Some Questions:
- Why katz centrality? Wouldn't eigenvector centrality work just as good? They already do an eigendecomposition anyway.
- What's the efficacy of the domar-weighting the centrality? How much of the domar-weight is caught up in speculative market stuff? Could weighing sectors by labor content be more useful?
- What's the purpose of their dynamics analysis? I don't get why it's necessary for the computation of the Katz centrality. Their formulation of sensitivity in terms of 'consumer utility' reeks neoclassical; this might just be my lack of general economic knowledge but 'consumer utility' seems a bit of a red flag.
Regardless, it seems that adversarial planning will involve graph theoretical analysis of input-output models as a means to characterize an economy's structure and thereby it's weaknesses. The question then becomes how can workers best exploit these weaknesses?
Pardon the file I attached was too big ill post the URL
Sorry all, here is the link
https://scholar.princeton.edu/sites/default/files/ernestliu/files/liutsyvinski2020_01.pdf
I'm very surprised that transportation isn't in the top 10 sectors with the highest v_i.
@Joe This is why I'm suspicious of parts of their analysis; "housing management" is oddly ranked in the top 10.
The centrality and spectral analysis seems to be the most relevant part of the paper.
Interesting coincidence, I've also been thinking about adversarial planning lately. I was recommended this paper by someone on YouTube, which goes into this stuff from a union perspective. No detailed mathematical analysis however. The Power Resources Approach by Stefan Schmalz and Klaus Dörre
@madredalchemist Interesting paper. I agree that eigencentrality makes more sense than Katz. Strangely they do eigendecomposition of the IO matrix (A) rather than the Leontief matrix (I - A). Or maybe I'm misreading. Either way, it is clear that eigenanalysis tells us what strikes are most likely to make the economy "ring".
One thing I don't see in the paper is sensitivity in terms of number of workers. The best places to strike are those that employ few people while being important to the economy. The ideal case would be one that hinges on a single comrade vital to the entire economy. This doesn't happen of course, but there may be cases that are close. Transportation comes to mind, but also sysadmins at ISPs, given how dependent modern companies are on the internet.
@thardin your mention of the ideal situation reminded me of a news story from a bit ago where French workers at a nuclear plant voted to strike in political protest.
https://www.reuters.com/business/energy/french-nuclear-hydro-output-reduced-by-strikes-2021-10-05/
This is very close to what you are referring to as even if it were just the group of plant operators with enough clearance/know how to stop the plant the same effect could happen.
France might be the best place for action like this since their national rail service is run almost exclusively on electricity.
@joe Oh yeah. Cut power -> no internet, no rail, no telephone (once UPSes and generators run out of juice). In fact the coming energy crisis in Europe this winter is an excellent time for energy workers to strike.
@thardin your right, I didn't notice that before, I wonder what the significance of the difference is; the leontif matrix is basically a supply to demand transform whereas the IO matrix is like a 'loss' transform akin to friction or disputed heat i.e how much of your production gets caught up in production
Also that's a good point, perhaps instead of weighing with respect to to labor content weigh against labor content as it's easier to interrupt production.
Maybe the goal should be the most low labor content sectors that are upstream of high labor content sectors?
@madredalchemist Something like that. Directed action can cause massive value loss for the employing class. However because it does not create any value, it does not create value for the workers themselves as a class. I've been sketching a blog post about this actually, but it's a bit thin so far.
@thardin I've been thinking about this too I had an idea about it that I should start a new topic about. In short, I remember a presentation I saw Cockshott give about LTV and he said that the only other thing that had a positive correlation to value was computation, which he attributed to it's universality. I was curious as to the implications of this on platform cooperatives, such as ride-share cooperatives which by their construction require computation and labor. Maybe co-ops which effectively blend computation and labor can be effective beachheads in the workers movement.
Luca Perrone did some pioneering work along these lines, published posthumously in 1984. I've done some rudimentary numerical experiments like this for the Swedish economy. There are some theoretical and data limitations to this approach from a trade union perspective. You'd need to separat the intensity and scope of industrial action, because it determines the nature of adversarial responses. This in turn may firm-level data on economic variables, such as capital stocks.
@dz1789 This is interesting, I will look into this, got a link?
I remember a presentation I saw Cockshott give about LTV and he said that the only other thing that had a positive correlation to value was computation,
I don't recall such a claim. Do you have a link? I know he touches on this in Computation and Its Limits but I haven't read it yet. Personally I'm skeptical of such claims since it would imply machines create value.