Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. = # Critical Minerals, Geopolitics, and the Green Transition Authors: Tomas Dominguez-Iino, Jonathan T. Elliott, Allan Hsiao Discussant: Laura Alfaro Video: https://www.youtube.com/watch?v=U_ssYRB6uPs&t=25392s ## Talk (07:03:12 – 07:31:57) [07:03:24] alright, we are very excited to have [07:03:26] the last paper of a great day [07:03:28] so Tomas, floor is yours [07:03:30] you have 30 minutes [07:03:32] thank you very much for the opportunity to present this paper [07:03:36] thank you [07:03:38] this is a joint work with my fantastic [07:03:40] co-authors, Jonathan Elliott and Alan Xiaow [07:03:42] we are seeing right there [07:03:44] I will be doing the presentation then Jonathan will be handling the Q&A [07:03:48] so this paper is on the energy transition [07:03:50] which as you are probably aware [07:03:52] will require batteries to store renewable energy [07:03:54] and in turn those batteries will require middles [07:03:56] so the focus of this paper [07:03:58] is going to be on the upstream mining of these minerals [07:04:00] taking place at locations like this [07:04:02] or lithium mine [07:04:06] concretely we are thinking of minerals such as [07:04:08] lithium, nickel and cobalt [07:04:10] these are critical for producing advanced batteries [07:04:12] and they are concentrated geographically in a few countries [07:04:14] this resource concentration [07:04:16] is what gives these countries their market power [07:04:18] and importantly they actually [07:04:20] consider exercising it [07:04:22] whether it is Indonesia as the top nickel producer [07:04:24] the DRC as the top cobalt producer [07:04:26] or Australia as the top lithium producer [07:04:28] all of them actively [07:04:30] intervene in the market [07:04:32] often with supply restrictions that have the explicit goal of boosting prices [07:04:36] so the research question in this paper is [07:04:38] what is the impact of this upstream [07:04:40] market power [07:04:42] on first of all aggregate battery adoption [07:04:44] we are going to call that green transition [07:04:46] and second the distributional impacts across the supply chain [07:04:48] and we are going to link that to geopolitical outcomes [07:04:52] so to answer that we are going to estimate a dynamic [07:04:54] equilibrium model of global mineral markets [07:04:58] now in a way we are asking a very classic question [07:05:00] in resource economics [07:05:02] but in this next generation of energy commodities [07:05:04] and there are some distinct features of this setting [07:05:06] that have to do with the fact that we are now [07:05:08] in a world with multiple natural resources [07:05:10] and that is the minerals [07:05:12] these distinct minerals and they are complementary [07:05:14] inputs in battery production [07:05:16] so what this means is that we are going to have [07:05:18] complementarities across resource markets [07:05:20] that is going to generate distinct economic mechanisms [07:05:22] for what we are typically used to seeing in single resource settings [07:05:24] with market power so for example [07:05:26] to work on oil and OPEC [07:05:28] so to expand a bit on those mechanisms [07:05:30] at a very high level the core idea [07:05:32] is that market power in a world with compliments [07:05:34] is fundamentally different than in a world with substitutes [07:05:36] the first way to see this is that complementarity [07:05:38] flips the incidence of market power [07:05:40] so for example in a classic OPEC setting [07:05:42] when Saudi Arabia does an oil supply [07:05:44] cup US oil producers win [07:05:46] because they produce a substitute to Saudi Arabian oil [07:05:48] in our minerals context [07:05:50] if we have for example Indonesia do a nickel supply cup [07:05:52] the other nickel producers win [07:05:54] because they produce a substitute to Indonesia's nickel [07:05:56] but Australian producers lose [07:05:58] because they produce a complement [07:06:00] that is when nickel gets expensive [07:06:02] it has to be paired with lithium [07:06:04] so lithium demand falls [07:06:06] and that makes lithium prices fall [07:06:08] the second way in which complementarity matters [07:06:11] is that it's going to flip the role of market power [07:06:13] for the green challenges in terms of whether it [07:06:15] accelerates it or slows it down [07:06:17] so to see this think of the foreign thought experiment [07:06:19] suppose you have a bunch of nickel producers [07:06:21] and they decide to merge [07:06:23] into a nickel cartel [07:06:25] we're going to call that within mineral consolidation [07:06:27] because that's consolidation happening [07:06:29] within the nickel market [07:06:31] this nickel is now being [07:06:33] this nickel is now substitutable across all these cartel members [07:06:35] so that consolidation leads to the standard result [07:06:37] that you have higher markups [07:06:39] that's going to raise battery prices [07:06:41] and slow down the green energy transition [07:06:43] by contrast [07:06:45] think of a different setting [07:06:47] where you have let's say a lithium and a nickel cartel [07:06:49] and now they decide to merge [07:06:51] into what we're going to call a multi-metal cartel [07:06:53] the joint decides on lithium and nickel output [07:06:57] in that case what's going to happen [07:06:59] well you need to understand that prior to this merger [07:07:01] if you take say the lithium producers [07:07:03] they were setting lithium prices too high [07:07:05] because they were not internalizing that they were destroying nickel demand [07:07:07] through complementarity [07:07:09] and the nickel guys were doing the same thing [07:07:11] so effectively there was this double-battery [07:07:13] so effectively there was this double-marginalization [07:07:15] that were posing on each other [07:07:17] so when they merged they internalized that [07:07:19] they lowered the mineral prices [07:07:21] and that's good for the energy transition [07:07:23] that accelerates it, so that's the key idea [07:07:25] so in short complementarity [07:07:27] has distinct economic mechanisms [07:07:29] in this multi-metal setting [07:07:31] that doesn't show up in single-resource settings [07:07:33] where there's only resubstituted forces at play [07:07:35] so that's the core idea [07:07:39] in this talk I'm going to go over first [07:07:41] on the background of the critical mineral sector [07:07:43] then I go into the model and finally the counterfactuals [07:07:47] so we could probably spend an entire day [07:07:49] on just features of the critical minerals [07:07:51] sector [07:07:53] so really I'm going to just boil it down to the bare essentials [07:07:55] that are necessary for the mechanisms in this paper [07:07:59] so first of all we need to define critical minerals [07:08:01] the term comes from lists of raw materials [07:08:03] that governments typically maintain [07:08:05] and they do these lists [07:08:07] with both economic and national security [07:08:09] gold's in mind [07:08:11] the exact definitions of value across country [07:08:13] but all of them typically capture some notion of low substitutability [07:08:15] and high supply risk [07:08:19] so these days about half of the periodic table [07:08:21] is on these lists so we're going to need to [07:08:23] boil it down [07:08:25] to say something meaningful [07:08:27] and given our focus on the energy transition [07:08:29] we're really going to focus on the subset [07:08:31] that are critical in terms of their energy [07:08:33] storage properties [07:08:35] so these are what we call battery minerals like lithium, cobalt and nickel [07:08:37] we're not going to be focusing on this paper on rare earths [07:08:39] that's a separate subset of critical minerals [07:08:41] that aren't used in batteries [07:08:43] because they don't have energy storage properties [07:08:45] they have other properties, magnetic properties [07:08:47] they're used in lasers [07:08:49] and they're often associated as well with national security purposes [07:08:51] which are not really going to be in this paper [07:08:55] so we're really going to focus on battery minerals here [07:08:57] in terms of demand [07:08:59] we're going to be focusing on demand from the electric vehicle sector [07:09:01] and that's because it's the key demand source [07:09:03] for batteries around the world [07:09:05] and also a major demand source for most of our minerals [07:09:07] in terms of data [07:09:09] we're going to be using micro data on both the supply and the demand side [07:09:12] on the supply side [07:09:14] we're going to construct a mine level [07:09:16] panel data set [07:09:18] using a combination of proprietary data sources [07:09:20] that we validated against official [07:09:22] government statistics [07:09:24] that are aggregated [07:09:26] and this is going to cover basically the near universe [07:09:28] of world production for our three minerals [07:09:30] on the demand side [07:09:32] we're going to have [07:09:34] battery consumption [07:09:36] these are like battery installations by EV manufacturers [07:09:38] so in this model it's the EV manufacturers [07:09:40] which are going to be consuming [07:09:42] buying the batteries [07:09:44] to then plug into the vehicles [07:09:46] and then finally we're going to be using [07:09:48] what we're going to call [07:09:50] battery mineral recipes [07:09:52] these are basically chemical recipes for each battery [07:09:54] that allow us to map battery demand [07:09:56] into mineral demand [07:09:58] so I'm going to go over just three stylized facts [07:10:02] that motivate the ingredients in our model [07:10:04] so the first one is that mineral supply [07:10:06] is geographically concentrated [07:10:08] so I'm showing you here is a share of world production [07:10:10] by the top three producers [07:10:12] for various commodity markets [07:10:14] and the main takeaway here is that [07:10:16] middle commodities are substantially more concentrated [07:10:18] than our traditional fossil fuel markets [07:10:20] that we typically worry about [07:10:22] because of resource concentration [07:10:24] and supply risk [07:10:26] this for production you can do the same plot [07:10:28] with reserves and the main takeaway still holds [07:10:30] okay [07:10:33] so one of the [07:10:35] reasons why you have this concentration [07:10:37] re-boils down to geological fundamentals [07:10:39] and just differential endowment across space [07:10:41] and that brings me to the second fact [07:10:43] which is that there's heterogeneous costs [07:10:45] across mines and countries [07:10:47] on the left what I'm showing you here [07:10:49] is a density plot [07:10:51] for the individual mines [07:10:53] of the average production costs for cobalt [07:10:55] and you can see for instance here [07:10:57] that cobalt mines in the DRC [07:10:59] have substantially lower costs than the rest of the world [07:11:01] and a large part of this is driven [07:11:03] by geological fundamentals [07:11:05] which are shown on the right [07:11:07] and that is that the ore grade of mines in the DRC [07:11:09] is about five times higher than in mines in the rest of the world [07:11:11] so we're going to be incorporating [07:11:13] this [07:11:15] mine level ore grade data [07:11:17] to get at this heterogeneity at a micro level [07:11:19] finally on the demand side [07:11:23] what I'm showing you here is a time series plot [07:11:25] of battery consumption over time [07:11:27] so this is the flow [07:11:29] of battery installations [07:11:31] that's added to the battery stock each year [07:11:33] so these are the new batteries being produced and consumed [07:11:35] and I'm showing you here [07:11:37] for different technologies [07:11:39] and basically what you can see is that it boils down [07:11:41] to two technologies [07:11:43] LFP, that's the blue line [07:11:45] those are lithium iron phosphate batteries [07:11:47] and NMC, those are nickel manganese [07:11:49] cobalt batteries [07:11:51] and they vary in their mineral recipes [07:11:53] and that's what I'm showing you on the right [07:11:55] this is a recipe matrix so it tells you [07:11:57] for each of these batteries [07:11:59] the actual intensity per kilowatt hour of capacity [07:12:01] just a clarification [07:12:03] are the batteries equally powerful or are they too much? [07:12:05] no no there's going to be some differential [07:12:07] in terms of their driving range [07:12:09] and I'll get into that in the estimates [07:12:12] so what I'm also going to do is [07:12:14] you can see that basically from this time series plot [07:12:16] that the market shares of these batteries [07:12:18] has changed over time [07:12:20] so I'm also going to capture that substitution [07:12:22] between battery technologies [07:12:24] which is responding to these mineral price changes [07:12:28] so let's go into the model now [07:12:31] so based on those facts we're going to have [07:12:33] basically these core ingredients to the model [07:12:35] and then we'll jump into the details [07:12:37] first of all on the supply side we're going to have [07:12:39] minerals indexed M, there's a set of them [07:12:41] capital M, there's some supply [07:12:43] S superscript M and we're going to use superscripts [07:12:45] for all upstream mining objects [07:12:47] this is going to come from a dynamic mine level extraction problem [07:12:49] on the demand side [07:12:51] we're going to have some batteries indexed J [07:12:53] these are different batteries technologies [07:12:55] demand is D sub J [07:12:57] we're going to use sub indices for all downstream battery objects [07:13:01] on the demand side we have the EV manufacturers [07:13:03] remember they're the ones consuming the batteries [07:13:05] they're the ones demanding them [07:13:07] so they're going to switch across these battery types [07:13:09] when the battery price has changed that's a substitution force [07:13:11] but then each battery type [07:13:13] is going to use minerals in fixed proportions [07:13:15] according to a recipe like the ones I showed you in that matrix [07:13:17] and those recipes tells you [07:13:19] how much mineral content M each battery J contains [07:13:21] and that's a complementarity force [07:13:23] so we have mineral complementarity within a battery [07:13:25] and then substitution across the battery technologies [07:13:27] then finally [07:13:29] we're going to have an equilibrium in the global mineral market [07:13:31] so we're going to use those recipes [07:13:33] to map the battery demand to the mineral demand [07:13:35] which is going to have to equate mineral supply to the clear markets [07:13:39] so let's jump into the demand model now [07:13:41] so this is the battery demand [07:13:45] yeah so there's heterogeneity in that [07:13:53] and that was the plot I showed in one of the first slides [07:13:55] but I'll discuss that [07:13:57] for lithium and nickel [07:13:59] for lithium and cobalt yes it's all in the most legal kind of batteries yeah [07:14:01] okay so [07:14:03] what's the point of this demand side [07:14:06] we're now flexible [07:14:08] the point of this is to have flexible substitution patterns [07:14:10] across these battery technologies [07:14:12] so for that we're going to use an almost ideal [07:14:14] demand system [07:14:16] in which you can think of an EV manufacturer [07:14:18] living in a region K [07:14:20] time T and it has a budget XKT [07:14:22] that it has to allocate across different battery technologies [07:14:24] so what this equation here says [07:14:26] is W is the expenditure share [07:14:28] on battery J [07:14:30] and that's the standard [07:14:32] expenditure share on battery J [07:14:34] and it's going to depend on some [07:14:36] this alpha JT term which is [07:14:38] unobservable battery J [07:14:40] time T varying characteristics [07:14:42] beta J which is an income effect [07:14:44] and then most importantly these gamma parameters [07:14:46] which are the price coefficients [07:14:48] and those are fully flexible [07:14:50] in the sense that it's indexed by these JJ prime [07:14:52] so that's the fully flexible substitution parameter [07:14:54] as with any demand system [07:14:56] we have an endogenity issue because of unobserved demand shocks [07:14:58] and that's these epsilons that we have here [07:15:00] so what we need here [07:15:02] is a supply shifter to trace out that demand curve [07:15:04] so what we're going to need here [07:15:06] is a cost shifter [07:15:08] for the batteries and that's what we're going to use [07:15:10] as an instrument [07:15:12] to construct that cost shifter what we're going to do [07:15:14] is use what we call non-EV mineral prices [07:15:16] so these are inputs for example [07:15:18] they go into the battery [07:15:20] like say phosphoric acid [07:15:22] and the rationale for that is that [07:15:24] these inputs they're relevant for the battery price [07:15:26] because they're in the battery recipe [07:15:28] but they're excluded because only a tiny [07:15:30] share of the demand is driven by the EV sector [07:15:32] going to Steve's comment [07:15:34] so for example for phosphoric acid concretely [07:15:36] it's relevant because it enters the LFP [07:15:38] battery recipe [07:15:40] but it's excluded because [07:15:42] most phosphoric acid goes into fertilizer [07:15:44] so it's really agriculture demand shocks [07:15:46] that are moved around these prices not EV demand shocks [07:15:48] so that's the idea [07:15:50] so these are the Martialian demand elasticities [07:15:52] that we get that we estimate [07:15:54] from this model [07:15:56] and we're looking at the IB results [07:15:58] which are the black triangles [07:16:00] so the first two rows of this plot [07:16:02] are the own price elasticities of the LFP [07:16:04] battery and the nickel based varieties [07:16:06] so they're negative [07:16:08] and the two bottom ones of the cross price elasticies [07:16:10] which are positive [07:16:12] first thing just conometrically [07:16:14] observation is that the IB results are larger in magnitude [07:16:16] than the least square ones which is what you would expect [07:16:18] with Simone Thanaidi bias [07:16:20] second [07:16:22] the own price elasticities of the LFP batteries [07:16:24] are substantially larger than [07:16:26] nickel based varieties [07:16:28] and this primarily reflects the fact that these LFP batteries [07:16:30] are lower end than nickel based varieties [07:16:32] because they have shorter driving range [07:16:34] and you can think of them as catering to more price sensitive [07:16:36] more price elastic consumers [07:16:38] finally the cross price elasticities [07:16:40] with the positive ones [07:16:42] are going to capture substitution forces [07:16:44] and that's going to weigh against the complementarities implied by the recipes [07:16:46] so that concludes the demand side [07:16:49] and I'm now going to move on to the supply side [07:16:51] so we have this [07:17:00] micro data at the individual mine level [07:17:02] for example we can see the average cost of production [07:17:04] what I'm showing you here is [07:17:06] mine level data for [07:17:08] producing lithium, US dollar per ton of lithium [07:17:10] and we could in principle just [07:17:12] construct these supply curves directly from average cost data [07:17:14] and that would amount to reproducing this figure [07:17:16] we would just take all these mines [07:17:18] we would order them from lowest to highest average cost [07:17:20] and then [07:17:22] the width of these bars is just [07:17:24] proportional to their capacity [07:17:26] but of course to call this a supply curve [07:17:28] we need to make some very strong assumptions [07:17:30] the first one being that these mines have to be acting myopically [07:17:32] they're just comparing [07:17:34] contemporaneous cost to contemporaneous marginal revenue [07:17:36] and second you need to assume [07:17:38] linear cost so that marginal and average cost are the same thing [07:17:40] right so we don't want to do that [07:17:42] so we're going to relax those assumptions [07:17:44] and we're going to have a model that does that [07:17:47] so this model is going to allow for dynamics and cost convexity [07:17:49] this here is the [07:17:51] is the problem of an individual mine I [07:17:53] which is going to extract some row or X [07:17:55] given it's reserves constraint [07:17:57] it's extraction cost and it's expectations over the future [07:18:01] so the first thing is that dynamics are going to come in [07:18:05] because we have finite reserves [07:18:07] and that's going to imply a shadow cost of extraction [07:18:09] or a hotel in condition [07:18:11] and what that means is that if you extract one ton of middle today [07:18:13] you're not going to extract tomorrow because of this finance [07:18:17] so an optimizing mine has to be indifferent between extracting today and tomorrow [07:18:19] and that's what shows up in this order of question [07:18:21] that the payoff for the mine [07:18:23] marginal revenue minus marginal cost of extraction [07:18:25] has to be equal across the two periods [07:18:27] subject to the discounting and the expectation [07:18:29] the convexity is going to come through the extraction cost function [07:18:33] which we're going to call here C of X [07:18:35] here I'm just going to use a simple quadratic [07:18:37] in extraction level X [07:18:39] but in the paper we generalize this [07:18:41] so that cost function has [07:18:43] two things that I want to point out [07:18:45] first of all it's going to have a convexity parameter [07:18:47] which is this capital term [07:18:49] which is going to hit the square term [07:18:51] and then there's going to be an adjacent critical shock [07:18:53] that's going to be the same as the initial term [07:18:55] so with cost convexities [07:18:57] the average and marginal costs now are going to be different [07:18:59] and that's the bottom two right equations [07:19:01] and importantly these idysyncratic shocks [07:19:03] are going to enter both averages and margins [07:19:05] and they're going to generate an endogenity issue [07:19:07] when estimating this capital convexity parameter [07:19:09] so what we're going to do here [07:19:11] is we're going to use that average cost data [07:19:13] I showed you in the previous slide [07:19:15] to in a sense purge for these idysyncratic shocks [07:19:17] the idea being that these average cost data [07:19:19] these are cost realizations [07:19:21] so they do contain in a sense [07:19:23] the unobserved cost shock in them [07:19:25] so I'm going to show you how we do that next [07:19:30] so what I'm going to show you here is just the Euler equation [07:19:32] from the previous slide and I've just applied differences to it [07:19:34] and I've defined an expectation error data [07:19:36] which is going to absorb that expectation term [07:19:38] from the previous slide [07:19:40] and this essentially just tells us [07:19:42] that changes in marginal revenue [07:19:44] have to equal changes in marginal cost over time [07:19:46] where those marginal costs have two components [07:19:48] one is the deterministic component [07:19:50] which is given by the extraction decision [07:19:52] and then the idysyncratic cost which is the epsilon [07:19:54] and the endogenity issue here [07:19:56] is that when you get a high epsilon shock [07:19:58] you're going to extract less [07:20:00] so the right hand side of this equation [07:20:02] the x and the epsilon's are negatively correlated [07:20:04] so rather than using the instrument [07:20:06] we're going to leverage this average cost data [07:20:08] and the idea is the following [07:20:10] this cost data is going to allow us to [07:20:12] leverage an additional moment of the model [07:20:14] which is the average cost definition [07:20:16] which is here in equation 2 [07:20:18] and the idea is going to be that [07:20:20] this is the x here is observed [07:20:22] this delta ac term [07:20:24] delta x is observed so we can just invert these epsilon shocks [07:20:26] as a function of those observables [07:20:28] and this parameter kappa [07:20:30] so we do that in equation 2 [07:20:32] we plug that orange term in equation 1 [07:20:34] and we've brought in rid of these cost shocks [07:20:36] and the final estimate equation [07:20:38] we have at the bottom [07:20:40] which is the old data [07:20:42] and the parameter kappa [07:20:44] and then we have this expectation error data [07:20:46] which we can deal with standard methods such as [07:20:48] x [07:20:51] so in our baseline specification [07:20:53] what we're going to do is we're going to allow [07:20:55] each of these individual minds to vary in their [07:20:57] or grade quality [07:20:59] that's this g term [07:21:01] so for every ton of extraction [07:21:03] or you extract x you get [07:21:05] g times x of middle quantity denoted q [07:21:07] we're also going to allow [07:21:09] the margin of revenue of each individual mind [07:21:11] to potentially deviate from the [07:21:13] perfectly competitive return [07:21:15] so margin of revenue could be equal to [07:21:17] that's the perfectly competitive return [07:21:19] but it can deviate by this amount [07:21:21] and this tau is going to capture things [07:21:23] like for example [07:21:25] export restrictions by the country where the mind sits [07:21:27] like the ones I showed you at the beginning of the slides [07:21:29] and we do several extensions [07:21:31] to this in the paper [07:21:33] these are the results [07:21:35] for the mineral supply elasticities [07:21:37] on the left we have the cost convexity parameter [07:21:39] for different minerals and these basically reflect [07:21:41] technological differences across minerals [07:21:43] so for example in the case of cobalt [07:21:45] we have this mineral [07:21:47] and you can think of that primarily coming from the fact that [07:21:49] a lot of this is mined in the DRC [07:21:51] artisanally almost kind of by hand [07:21:53] so you can think of that as a technology [07:21:55] that's basically linear in labor [07:21:57] whereas the other minerals are much more capital intensive [07:21:59] subject to capacity constraints [07:22:01] in the short run which you can think [07:22:03] that leads to stronger convexities [07:22:05] these cost convexity parameter [07:22:07] intuition is going to [07:22:09] translate to the supply [07:22:11] elasticities which are on the right [07:22:13] these are long run supply elasticities [07:22:15] that we get from the model [07:22:17] for different minerals [07:22:19] and for each mineral we're showing it for the top producer [07:22:21] and also the rest of the producers [07:22:23] so for example if you look at the DRC [07:22:25] cobalt versus the rest of the world cobalt [07:22:27] DRC is substantially more elastic [07:22:29] again consistent with the fact that there's this low tech fringe [07:22:31] sector of mining that can come online [07:22:33] quickly when prices spike [07:22:37] so that's the model in the estimation [07:22:39] I'm now going to go into the counterfactuals [07:22:41] so we're going to take this model [07:22:43] and look at the different economic equilibrium paths [07:22:45] under different policy scenarios [07:22:53] so the first thing we're going to do is what we're going to call a supply chain vulnerability exercise [07:22:57] this amounts to removing the top producer [07:22:59] of each mineral market and then solving [07:23:01] for a full path of equilibrium price changes [07:23:03] and what I'm showing you here in this table [07:23:05] is these average price effects [07:23:07] of these interventions [07:23:09] and we're going to do this for each mineral market [07:23:11] so the first thing to note is that we have [07:23:13] significantly large price increases [07:23:15] for the minerals being disrupted [07:23:17] and that's shown by the diagonal terms in black [07:23:19] in the first three columns [07:23:21] and those range between 76 to over 800% [07:23:25] the second result is that [07:23:27] if we look at the off diagonals [07:23:29] these orange terms, those are all negative [07:23:31] and that's the price effects on the other mills [07:23:33] that are not disrupted [07:23:35] so what this tells us is that minerals are gross compliments [07:23:37] if we would have been in a substitute world [07:23:39] those numbers would have been positive [07:23:41] because as for example lithium gets disrupted [07:23:43] you would have had substitution towards nickel and cobalt [07:23:45] so we see the opposite here [07:23:47] and finally in terms of [07:23:49] battery price pass through, those are the two columns [07:23:51] on the far right [07:23:53] we'll set up some heterogeneity depending on the mineral being cut [07:23:55] so for example if lithium is cut [07:23:57] lithium is a common input to all batteries [07:23:59] so that's going to raise all battery prices [07:24:01] but that's not the case for nickel or cobalt [07:24:05] so next we're going to move to counterfactuals [07:24:07] that explore the role of market power within a single mineral market [07:24:11] so here where I'm showing you is the impact [07:24:13] of Australia [07:24:15] unilaterally doing an optimal lithium supply cut [07:24:19] and these are the changes in surplus [07:24:21] across the different producers [07:24:23] so on the left you can see the lithium producers [07:24:25] Australia obviously benefits from this [07:24:27] because they're the ones doing the supply cut [07:24:29] but then Chile and Argentina also benefit from this [07:24:31] because they free ride on this lithium supply cut [07:24:33] they're producing the substitute to the Australian lithium [07:24:37] the other producers, the nickel and cobalt producers [07:24:39] they all lose from this because they're producing the complement [07:24:41] to the Australian lithium that just got expensive [07:24:43] beyond the spillovers across the upstream producers [07:24:47] we can also look at the effects on downstream green adoption [07:24:51] and this left panel shows you the effects on batteries [07:24:53] of this Australian policy [07:24:55] so all battery quantities drop there because lithium is a common input [07:24:57] but instead if we would have say Indonesia [07:25:00] doing a nickel supply cut on the right [07:25:02] we've gone some substitution [07:25:04] from battery end to battery out [07:25:06] because in that case what's going to happen is that [07:25:08] as nickel gets expensive [07:25:10] within the battery end bundle [07:25:12] there's going to be a decline in demand for lithium because of complementarity [07:25:14] that's going to lower lithium prices [07:25:16] and that affects the access of subsidy to battery out [07:25:18] and that substitution is going to soften the drop [07:25:20] in total adoption from this intervention [07:25:24] finally moving beyond this unilateral supply cut [07:25:26] these unilateral supply cuts [07:25:28] we're going to move to cartel policy [07:25:30] and what I'm showing you here [07:25:32] is the impact of a lithium cartel [07:25:34] as it gradually gets bigger [07:25:36] as it gradually increases membership [07:25:38] we're going to start off with Australia alone [07:25:40] being the only member of this cartel [07:25:42] then we're going to have Archile and then Argentina [07:25:46] and the takeaway here is that this consolidation [07:25:48] within the middle market in this case lithium [07:25:50] is anti competitive in the sense that prices [07:25:52] progress with increase [07:25:54] and anti adoption in the sense that quantities [07:25:56] progressively decline [07:25:58] so this is the standard intuition that you would get [07:26:00] in a cartel with producers [07:26:02] that produce substitutes among each other [07:26:04] consolidation is anti competitive [07:26:08] finally I'm going to move on to the counterfactual [07:26:10] to explore the world of market power [07:26:12] and what this means [07:26:14] so here the idea is the following [07:26:16] suppose you have a bunch of independent [07:26:18] mineral cartels let's say a lithium and a nickel cartel [07:26:20] and now we're going to have the merge [07:26:22] into a single multi-mil cartel [07:26:24] that jointly decides on lithium and nickel supply [07:26:28] so on the one hand we have this anti competitive force [07:26:30] because we have bigger cartels [07:26:32] so we have higher from higher concentration [07:26:34] but on the other this cartel is now internalizing [07:26:36] cross-middle complementarities [07:26:38] and that's a pro competitive force [07:26:40] and again the idea here is that prior to the merger [07:26:42] the lithium cartel was setting prices too high [07:26:44] because it didn't internalize it was destroying the [07:26:46] nickel demand the nickel guys were doing the same thing [07:26:48] so we have some double marginalization [07:26:50] before the merger [07:26:52] but after the merger gets internalized [07:26:55] so we do that exact simulation [07:26:57] in our model [07:26:59] and we have the results here on the right in this figure [07:27:01] and what we find is that both of these forces [07:27:03] the anti and the pro competitive one [07:27:05] basically nearly offset each other [07:27:07] so even though we have this cartel that's bigger [07:27:09] as a result of this merger [07:27:11] the quantities barely change [07:27:13] and importantly if anything the results are pro competitive [07:27:15] and you can see that by comparing [07:27:17] the light and dark blue bars [07:27:19] so the main takeaway here is that consolidation [07:27:21] can be pro adoption [07:27:23] when you have these complementarities at play [07:27:27] so one way to think about this [07:27:29] is that this exercise is really a horizontal merger [07:27:31] of compliments [07:27:33] and that's analogous to a vertical merger [07:27:35] and that both eliminate double marginalization [07:27:37] and that's a classic result [07:27:39] I mean really a classic going back to [07:27:41] Kurno 1838 [07:27:43] yeah [07:27:45] so I think [07:27:47] so Keynes was right we're all really slaves [07:27:49] to a defunct economists [07:27:53] so this multi mineral idea [07:27:55] you might think well maybe this multi mineral cartel [07:27:57] is not necessarily cohesive [07:27:59] and indeed in the paper we show that there are [07:28:01] defection incentives for each of the individual members [07:28:03] from this cartel [07:28:05] but I think the main value of this multi mineral exercise [07:28:07] is that it's more of a thought experiment [07:28:09] or an analogy for something broader [07:28:11] which is coordination across mineral markets [07:28:13] and even if that doesn't happen [07:28:15] because let's say the DRC sits down with Indonesia [07:28:17] it can still happen if through another channel [07:28:19] which is through multi nationals [07:28:21] that are on portfolios of mines [07:28:23] across different mineral markets [07:28:25] and we can see this in the data because we see the ownership data [07:28:27] for each of these individual mines [07:28:29] so I'm showing you what I'm showing on the left [07:28:31] is concentration of world production [07:28:33] but not by the country where the mine physically sits [07:28:35] like I showed you the beginning [07:28:37] but by the country of the multi national that owns the mine [07:28:41] and what you can see here is [07:28:43] there's essentially just one country [07:28:45] that has ownership stakes across all three mineral markets [07:28:47] and that's China [07:28:49] so that's Chinese ownership of mines in the DRC [07:28:51] in South America etc [07:28:53] so what we're going to do with the model [07:28:57] is do some counterfactuals that explore this role [07:28:59] of Chinese market power [07:29:01] as it coordinates across [07:29:03] these mines that it owns [07:29:05] and that's what the figure on the right shows [07:29:07] and the main takeaway here is that this [07:29:09] horizontal coordination [07:29:11] by the Chinese multi nationals [07:29:13] across their mines, across the new markets [07:29:15] is basically going to mimic [07:29:17] the pro competitive mechanisms of a multi mineral cartel [07:29:19] and that's because they also internalize [07:29:21] complementarities [07:29:23] we also do a further counterfactual [07:29:25] which is [07:29:27] basically motivated by the fact that China also [07:29:29] has a large domestic EV sector [07:29:31] which is a vertical coordination between [07:29:33] the Chinese down to new manufacturers [07:29:35] and the Chinese upstream mines [07:29:37] and that's going to layer on further [07:29:39] pro competitive effects [07:29:41] although for a different reason which is just [07:29:43] the standard elimination of double marginalization [07:29:45] in a vertical input output [07:29:47] structure so finally [07:29:51] let me conclude here [07:29:53] with a final counterfactual [07:29:55] which is a bit more tied to geopolitics [07:29:57] so basically [07:29:59] the kind of core idea from the previous slides [07:30:01] is that market power can be held by sovereign states [07:30:03] and this typically exercised [07:30:05] by state run cartels within a single [07:30:07] single, within a single mineral market [07:30:09] or it can be exercised by multi nationals [07:30:11] which are diversified across mineral markets [07:30:13] and we know from the history of natural resources [07:30:15] that geopolitics shifts the balance between the two [07:30:17] for example [07:30:19] if you go back to the mid 20th century [07:30:21] when you had the big oil nationalizations [07:30:23] in the Middle East [07:30:25] you can basically think of that as a transfer [07:30:27] of control of those mines [07:30:29] in the Middle East [07:30:31] from the western four multinationals that were operating there [07:30:33] to a state run cartel [07:30:35] OPIC [07:30:37] so motivated by that we're going to do a similar exercise [07:30:39] with a model which we're going to call a [07:30:41] resource nationalization exercise [07:30:43] which is essentially comparing [07:30:45] the market power of these Chinese multinationals [07:30:47] across these mineral markets [07:30:49] to the state run cartels [07:30:51] which exercised the market power within a mineral market [07:30:53] and the mechanism here is that [07:30:55] what this nationalization does [07:30:57] is that it breaks up coordination across mineral markets [07:30:59] by severing those multi national ownership links [07:31:01] and it reinforces it within [07:31:03] mineral markets by setting up the state run cartels [07:31:07] and we know from the previous slides that that's an anti-adoption force [07:31:09] or an anti-impaired force [07:31:11] so what we show here on the right [07:31:13] is the result of that counter fractional [07:31:15] and the main takeaway here is that [07:31:17] this market power [07:31:19] by these state run cartels is significantly worse [07:31:21] for green adoption than the market power [07:31:23] of these Chinese multinationals [07:31:25] so to conclude [07:31:27] Chinese multinationals presence [07:31:29] upstream [07:31:31] can be a positive force for green adoption [07:31:33] so to conclude [07:31:37] critical minerals are going to power the energy transition [07:31:39] these mineral resources are concentrated [07:31:41] much like other natural resources we've studied in the past [07:31:43] like oil [07:31:45] but they introduce these novel interdependencies [07:31:47] that have distinct implications for geopolitics [07:31:49] and green adoption [07:31:51] thank you ## Discussion (07:31:57 – 07:50:30) [07:31:59] Lord Alfaro is the discussant [07:32:09] so [07:32:11] what is the paper [07:32:15] so thank you very much [07:32:19] for inviting me [07:32:21] to discuss this paper [07:32:23] I am now at the IDB [07:32:25] and so I should have put a disclaimer [07:32:27] although it's a discussion so [07:32:29] not sure how that works [07:32:31] so [07:32:33] what does the paper do [07:32:35] so the central thesis [07:32:37] and I think this is the key contribution of the paper [07:32:39] is that [07:32:41] a [07:32:43] critical mineral markets [07:32:45] have interdependencies [07:32:47] from the joint use [07:32:49] which is what they call these recipes [07:32:51] so it is not about the minerals themselves [07:32:53] but [07:32:55] they do create a network [07:32:57] that could be a complement or a substitute [07:32:59] and that is an empirical question [07:33:01] also [07:33:03] that these recipes can create important asymmetries [07:33:05] and they have this example [07:33:07] so lithium is common [07:33:09] in every battery chemistry [07:33:11] so you cannot substitute it away [07:33:13] so restrict the policy [07:33:15] what it does is make [07:33:17] batteries more expensive [07:33:19] and EV adoption [07:33:21] which is just one tiny form [07:33:23] of battling climate change [07:33:25] let me just say that [07:33:27] because there are countries like [07:33:29] in the Americas that we have hydro [07:33:31] and that's a different thing and it's actually a better energy [07:33:33] but so in this particular case [07:33:35] of EV adoption [07:33:37] it does reduce EV adoption [07:33:39] a nickel is battery specific [07:33:41] and so consumers [07:33:43] can substitute it away by choosing [07:33:45] different type of batteries [07:33:47] and so these abilities of substitute [07:33:49] offsets the effects [07:33:51] of restrictive nickel policy [07:33:53] whether at the end [07:33:55] how it affects the lithium price [07:33:57] it will depend on whether the complements and substitutes [07:33:59] and as I was mentioned [07:34:01] by Thomas in the evidence [07:34:03] they do find that there is a gross complementarity [07:34:05] and then at the end [07:34:07] they do these counterfactuals [07:34:09] that they find that a multi-mineral [07:34:11] culture is better [07:34:13] than a single mineral culture [07:34:15] cartel because internalizes [07:34:17] cross-mineral spillovers [07:34:19] but the fact that we're all dominated by China [07:34:21] is the greatest thing that could happen to us [07:34:23] because [07:34:25] again it controls [07:34:27] many of these things [07:34:29] it is a very ambitious paper [07:34:31] I don't think [07:34:33] the presentation did justice of [07:34:35] all the things that are in this paper [07:34:37] all the data they have [07:34:39] all these [07:34:41] universe of [07:34:43] mines in lithium, nickel and cobalt [07:34:45] I actually also have the S&P data set [07:34:47] so I know all the good stuff [07:34:49] that is in that data [07:34:51] they have cost or grades, capacity, reserve [07:34:53] ownership [07:34:55] they also bring these [07:34:57] engineer recipes EV installation by model [07:34:59] brand region like it does have a lot of data [07:35:01] a lot of these data is using the [07:35:03] identification they have very [07:35:05] cool IV [07:35:07] then they have demand [07:35:09] then they have equilibrium [07:35:11] so I think the paper does have a very nice [07:35:13] contribution to the literature [07:35:15] in particular it does come from this estimation [07:35:17] and it makes it a very complete IOP paper [07:35:21] why do we care though? [07:35:23] we do care [07:35:25] because [07:35:27] there is a concern that these [07:35:29] minerals are not just [07:35:31] geologically concentrated [07:35:33] because they have been geologically [07:35:35] concentrated since the Panjera [07:35:37] has started to move [07:35:39] it's because China is dominating [07:35:41] a lot of these processes as well [07:35:43] lithium, nickel, graphite [07:35:45] and don't get me started in brothers [07:35:47] it is all [07:35:49] top there and the instrument that the paper [07:35:51] mentions are real [07:35:53] we do know that there has been these changes in the quota [07:35:55] in Indonesia and it has been an issue [07:35:57] in Congo [07:35:59] in China again all of these things are [07:36:01] real and resource naturalism [07:36:03] is real [07:36:05] so let me go [07:36:07] over some of the comments [07:36:09] to the biggest point of the paper [07:36:11] I agree we need to look at these minerals [07:36:13] as a network we cannot look at them [07:36:15] in insulation [07:36:17] this is the right way to look at these rocks and [07:36:19] primes and I do think this is [07:36:21] important contribution [07:36:23] but perhaps some of the assumptions [07:36:25] are a tad restrictive [07:36:29] for some of the questions they are tackling [07:36:31] and in fact I think the paper [07:36:33] reads at times as two nice papers [07:36:35] and I think this is also the outcome [07:36:37] that I read the first version and then they send [07:36:39] me the second version so the first [07:36:41] version does read as a very nice [07:36:43] IO paper [07:36:45] that models batteries recipes [07:36:47] very detailed [07:36:49] estimation and then the second [07:36:51] soul of this paper is the geopolitics [07:36:55] of the [07:36:59] and so I'm going to stay away [07:37:01] from a lot of the IO [07:37:03] but I am now a minerals aficionado [07:37:05] because [07:37:07] we're doing a critical mineral [07:37:09] initiative [07:37:11] and I have here [07:37:13] all the details [07:37:15] of the lithium [07:37:17] can tell you [07:37:19] that there is [07:37:21] the Atacama Salar [07:37:23] source of [07:37:25] lithium in the world [07:37:27] so now I know a lot of details about [07:37:29] this thing which is kind of ironic [07:37:31] because if there's a country that has [07:37:33] no minerals it's Costa Rica [07:37:35] like tons of insects [07:37:37] but no rocks [07:37:39] so I'm going to go [07:37:41] into the geopolitical implications [07:37:43] because that's the game they play [07:37:45] like if you read the abstract that's what they're doing [07:37:47] they want to find out the geopolitical [07:37:49] implications and after all [07:37:51] they want to find out the [07:37:53] the [07:38:07] so I will use their lens [07:38:09] and so the comments are an extension [07:38:11] of the principle [07:38:13] and I think the counterfactuals [07:38:15] is using a model that is not [07:38:17] per se equipped to changing [07:38:19] the policy [07:38:21] so all the comments [07:38:23] they have clarifications food notes [07:38:25] and appendixes [07:38:27] but I think sometimes they were [07:38:29] just left as a food note and they were [07:38:31] not really internalizing to the paper [07:38:33] maybe a curtail would have been [07:38:35] so for the first [07:38:37] soul that is just the [07:38:39] ION [07:38:41] I think they take very seriously [07:38:43] a lot of the parts of the minerals [07:38:45] but then they forget about some [07:38:47] But I think some of these simplifications for sure affect the quantitative implications [07:38:53] and may affect the qualitative ones. [07:38:56] How complementary and substitute the critical minerals are will depend on some of their [07:39:00] assumptions and some of their, the assumptions on the estimations. [07:39:04] For the second, I do believe the model neutralizes China, in particular in the midstream. [07:39:13] And so this is also likely to overstate the quantitative importance of complementarity [07:39:18] and perhaps also the some of the implications. [07:39:23] So broadly I'm going to discuss mining production and global value chains, then I'm going to [07:39:28] go into some of the demand and then I'm going to talk about the counterfactuals. [07:39:32] So I actually think that Disney and Snow White and the Smurfs have done a disservice [07:39:37] to the mining industry. [07:39:39] Because everyone thinks that mining is you go with your pickaxe and your shovel and then [07:39:43] you walk away. [07:39:44] And mining, unless you're doing gold in San Francisco and you were very lucky, is anything [07:39:50] but that. [07:39:51] Mining is a very complicated process, is a very chemical and physical process that [07:39:58] includes many parts. [07:40:00] And so you have to do all the R&D, the exploration, the mineral extraction that is a complicated [07:40:05] process that we find in the manufacturing and assembly. [07:40:09] And these parts I think oversimplification may affect results. [07:40:14] So mining upstream in the model it feels like arrives like mana and that is not the case. [07:40:20] You need a lot of inputs, reagents, chemicals and all of these inputs basically come from [07:40:26] China. [07:40:28] For battery manufacturing what they carry is not the rock, is battery grade lithium [07:40:34] and nickel, not the concentrate. [07:40:36] And this includes a lot of processing refining and that does go through a lot of chemicals. [07:40:42] The refining in the mistream, as I said, I think China loses the edge by construction. [07:40:48] And in the downstream, the paper is all about the cathode but they neutralize the anode. [07:40:55] And I joke with my array that I never thought I would use cathode so many times in a sentence [07:40:59] but now I'm like deep into cathode because copper is an important cathode. [07:41:05] There's a lot of more common materials and this affects the network. [07:41:10] So let me go into lithium. [07:41:11] Turns out there's not one lithium. [07:41:13] The paper says this. [07:41:14] The paper is aware about this. [07:41:16] They do go into mine level of retorting, yet equilibrium clears one global lithium [07:41:21] market and one price. [07:41:22] It turns out that the lithium in Australia is very different from the lithium in Chile. [07:41:27] In Chile, basically, you have ponds and you wait two years until that thing dries up. [07:41:32] You put a little bit of chemicals and literally that's what you do. [07:41:34] That's why Chile and Australia is so very low cost. [07:41:38] But it is inelastic in the short run. [07:41:40] You're going to do anything. [07:41:41] You have to wait for the thing to dry up and it can take two years. [07:41:45] So whatever you have at the moment is what you have. [07:41:47] The Australian, which is hard rock, is very elastic but it's expensive. [07:41:50] You have to crush it. [07:41:52] The Chilean brine, you import the chemicals but it's basically at the end refined [07:41:57] once it dries up, you have the lithium. [07:42:00] The Aussie rock, they mostly send it to China. [07:42:03] In the model, they enter at the same level but they're interchangeable. [07:42:08] Chile, lithium is ready for the battery. [07:42:12] Australia rock is not. [07:42:17] Lithium is not a commodity for the most part. [07:42:19] It remains heavily contract based. [07:42:23] It's really sold in long term contracts. [07:42:26] The sport market is very small. [07:42:27] We tend to think that all of the commodities are like oil. [07:42:30] It's sold in a commodity state. [07:42:33] Lithium is not rare, not many are not. [07:42:36] It turns out that all grades differ. [07:42:39] So there is actually not one nickel. [07:42:41] So the nickel that comes from Indonesia is actually of a lower class [07:42:46] in terms of all grade and so disproportionately it goes to stainless steel. [07:42:51] The better lithium comes from other countries, [07:42:53] by the way Brazil has also the better lithium, number three reserves, [07:42:57] in case you care about Latin America. [07:42:59] But this is becoming scarier. [07:43:02] So that's why there's effort to try to see how we're going to deal with the other one. [07:43:06] But most of the other one actually goes to steel. [07:43:10] So it matters because the model clears one nickel, one price and one elasticity. [07:43:16] And there's also this, how exactly do you map the lary that is lower class into a battery? [07:43:23] For battery you need high grade nickel. [07:43:28] Then there is the layers of mistream. [07:43:31] So basically in the paper the price of the battery are the recipes times price. [07:43:36] And there is some markup that absorbs many things. [07:43:41] And this markup is held constantly in all the counterfactuals. [07:43:46] So all mineral shocks pass directly to battery prices. [07:43:50] And so again you do have a very passive mistream and a very passive assembly. [07:43:56] This is a limitation, it's very hard to get data. [07:43:59] And I know that, I understand that. [07:44:02] But it does matter because you're looking at strategic reasons, [07:44:07] analysis, cartels, unilateral restrictions. [07:44:10] And then refining chemicals and manufacturing cannot exercise market power by assumption. [07:44:17] Again, then we take it and we do counterfactuals for many years. [07:44:20] And it is still fixed. [07:44:22] So again, this is why I think sometimes I would be fine if these were the IO exercise. [07:44:28] I have more problems when you're changing completely market structure with different and [07:44:31] keeping this mu constant. [07:44:33] The paper acknowledges this, the paper knows this, as I said, this is there. [07:44:38] And the argument is that markups are small. [07:44:41] But markups can be small because that is the strategic choice of China to keep [07:44:45] them small for a reason. [07:44:46] So I think the question is not if they're large or small. [07:44:49] It's if they can move. [07:44:51] And these markups can move. [07:44:53] Then we go to the anode. [07:44:56] So it turns out that it's very picture for a battery, you need an anode and [07:45:01] you need a cathode. [07:45:02] The cathode actually, it could be lithium, it could be nickel, it could be [07:45:05] gobal, it could be magnesium. [07:45:07] But the anode can only be one, graphite. [07:45:10] And it turns out that is the main part of the battery. [07:45:14] It's proportionally, it turns out that copper and aluminum also are more important [07:45:17] than anything else. [07:45:18] But the graphite is very important. [07:45:21] They don't include it because they say now there is a fake graphite. [07:45:26] But that fake graphite also comes from China. [07:45:28] So I don't think a geological or chemical substitution can undo industrial [07:45:35] concentration. [07:45:37] So the main input is outside of the model. [07:45:40] In this simplification, the paper uses two inputs. [07:45:46] But it does matter that there are more inputs. [07:45:48] And so if you're two, then it does become very easy to determine the cross [07:45:54] effect of lithium and nickel. [07:45:56] When you have more that are important, it is a little bit tricky. [07:45:59] And so for sure it affects quantitative results, but it could actually also [07:46:04] affect qualitative results. [07:46:06] So again, the paper argument, and I agree, is that critical minerals differ from OPEC [07:46:11] because there is joint use complementarity. [07:46:13] But then having a better sense of this matrix, because this matrix does define [07:46:18] power, I think it's important. [07:46:20] But I mean market power. [07:46:23] The recipes are fixed by their own paper page. [07:46:28] They do show that they change. [07:46:32] In the time span that they use it, they're going to change. [07:46:34] So again, this will matter in terms of some of the counterfactuals. [07:46:40] The EV demand is a little bit exogenous. [07:46:43] It's taken outside and they do have some forecasts. [07:46:48] But for 10, 15 years, it starts to become an issue. [07:46:53] There's these other assumptions nicked in there that when the price of one [07:46:58] of the inputs goes up, the whole battery demand goes down, that I think [07:47:02] it would be nice to make it more evident if you really need this assumption. [07:47:06] Because it's not for me trivial. [07:47:09] Like the price of Nike goes up and then we all buy less than issues. [07:47:13] So it is there. [07:47:15] It is a little bit heating. [07:47:16] I think it gives you a lot in terms of the results. [07:47:20] Not all demand in EV is, Nikol doesn't go to EVs. [07:47:28] There is growing demand, changing demand in your sample. [07:47:32] And the other one is these markets are segmented. [07:47:35] And some of this is taking consideration, but it does affect some of the results. [07:47:39] If I'm already lithium, I cannot substitute into more lithium. [07:47:43] So let me just finish with what the cartel is controlling, [07:47:47] which is at the end the main part of the paper. [07:47:51] The model policy instrument is my now put. [07:47:54] But what does supplying the restricting the supply means? [07:47:58] In Chile, what does it mean? [07:47:59] Like I start drying less lithium. [07:48:02] I delay evaporation. [07:48:04] I break my long term contracts. [07:48:07] In Australia, I don't send the rock to China in Indonesia. [07:48:11] I don't do the extraction. [07:48:13] I don't do, again, there is a part of it. [07:48:16] What is the real policy? [07:48:20] And again, this matters because the bottlenecks sits at different places in the value chain. [07:48:25] There's another thing that they don't really control of the production process. [07:48:28] So it really doesn't send like, is China going to be so happy to sell them the inputs? [07:48:34] Again, there is this dependency, dependence downstream, upstream, [07:48:39] that may become an issue. [07:48:42] And again, owning the deposit is not the same thing as controlling effective supply. [07:48:48] It then means that the coalition matters, not just the size. [07:48:53] To be able to determine what is feasible. [07:48:56] To do a collision with a player that requires a lot from China, [07:49:00] either for refining or to the chemicals, [07:49:04] it becomes a little bit complicated to sustain. [07:49:09] So I do think the paper needs a little bit more institutional interpretation [07:49:13] when analyzing this welfare rack. [07:49:14] What is feasible, what is stable, there is a nice food note that [07:49:18] the auditors just want to know if everyone will be happy in and assumes enforcement. [07:49:24] But in cartels, the thing that has always been complicated is the enforcement. [07:49:27] The idea of having a cartel, everyone has it. [07:49:30] But sustaining a cartel is what is complicated. [07:49:33] So I have some suggestions within the machinery of what to improve, [07:49:39] add a little bit more to the input output, add a little bit of non-EV demand. [07:49:43] But I would handle those counterfactuals with a little bit more care [07:49:47] because there's a lot of assumptions in those estimates [07:49:49] that is not clear to me carry over when you're completely changing the market structure. [07:49:54] So to conclude, the paper uses the right set of glasses. [07:49:58] Minerals cannot be understood by themselves. [07:50:01] We need to look at the independence. [07:50:04] I do like a lot your supply chain, [07:50:06] a vulnerability results. [07:50:09] But next steps I would handle with a little bit more care, [07:50:13] especially geopolitical considerations. [07:50:17] And just because I am truly a macro person, let me remind you [07:50:20] that we do know well in Latin America that geology is not destiny. [07:50:24] We know that there is a resource cares [07:50:27] and macro management matters a lot for that. ## Q&A (07:50:30 – 08:02:23) [07:50:30] Thanks. [07:50:36] Fantastic. Let me open it up for questions. [07:50:40] Steve. [07:52:56] That I see is a key message for this. [07:52:58] Maybe I don't know where the recipe language came from, [07:53:00] but I would just take me on to it. [07:53:03] Instead of citing Julia Childs, let's cite it on the air. [07:53:10] It turns into supply, right? [07:53:12] I mean, there's a dynamic part of this too. [07:53:14] If you look at the other price shots, the biggest effect of those [07:53:17] was the drilling boom, the geographic lead that were survived [07:53:21] in the coolant and oil. [07:53:23] That can also happen to some of the minerals, [07:53:27] despite their essentiality. [07:53:30] What's that dynamic response going to be? [07:53:33] That's an interesting question. [07:53:35] But certainly the prices are responding in a way that says [07:53:37] they're critical and in some sense, [07:53:40] we get massive price responses in that shorter run. [07:53:43] And I'd like to see a little discussion of the longer [07:53:45] in terms of dynamic. [07:53:48] John. [07:53:49] I have a related point about the dynamic. [07:53:51] I think it's on the demand side. [07:53:53] There is a great amount of supply side on the demand side, [07:53:55] but we know for a great amount of time. [07:53:56] So I'm dead. [07:54:00] One reason I'm not particularly interested here is that [07:54:04] the elasticity is a different price. [07:54:05] We also get dodging and it's linked to whether I have a stockpile [07:54:08] on some policy choices. [07:54:10] So if there's anything you guys can speak to, [07:54:11] then you have a better quantitative model of this market [07:54:14] that we used to provide. [07:54:18] Why don't you take a few minutes to reply [07:54:20] and then we'll take some more. [07:54:22] Great. [07:54:23] Let me first start by thinking Laura. [07:54:26] Where is Laura? [07:54:28] Thank you. [07:54:28] There you go. [07:54:29] Okay. [07:54:29] Thanks first of all for an excellent discussion. [07:54:32] Thanks also for reading the paper twice. [07:54:36] There's a lot of really great stuff in your discussion. [07:54:40] You know this context really well. [07:54:41] And so it's great to get your reaction. [07:54:44] Let me make just two points and then we can talk offline [07:54:48] about some of the other stuff. [07:54:51] So you noted kind of the different types of these minerals [07:54:57] and that's absolutely true. [07:55:00] I think there are some limits to exactly how seriously [07:55:03] we can take this but certainly on the cost side [07:55:06] we can be quite flexible and we think it's one of the advantages [07:55:09] of the cost approach that we take, [07:55:10] our cost data that we can have very different costs [07:55:13] for different types of mines. [07:55:16] So we do have that in there but point taken about [07:55:20] how that may influence things on the demand side as well. [07:55:24] You also talked a little bit about the midstream sector [07:55:28] which is largely dominated by China. [07:55:34] Due to the data limitations that we have [07:55:37] we're a little bit reduced form about what we can do there. [07:55:41] We feel essentially we've got kind of like a wedge [07:55:44] between what would be implied by the mineral prices [07:55:49] and the battery prices and we feel pretty comfortable [07:55:53] with the way that we approach it in the counterfactuals [07:55:56] where we keep it fixed based on work that has different data [07:56:01] than we do and has looked at the midstream battery manufacturing [07:56:04] and found really limited markups in that setting [07:56:07] but point well taken that that's kind of within the context [07:56:13] that we observed today and we're changing the market structure. [07:56:17] We've explored this a little bit in counterfactuals [07:56:18] that aren't in the paper but we've played around with this [07:56:22] where China responds to some policies [07:56:24] and we actually get a pretty small effect there. [07:56:28] We can think about that more. [07:56:29] Let's see. Okay, so next we've got Steve. [07:56:32] You asked if we could compare this to say like a more macro approach, [07:56:36] a CES approach. [07:56:39] I'm glad you bring that up because we do think that's kind of [07:56:41] one of the advantages of what we're doing on the supply side [07:56:43] is we've got this almost ideal demand system. [07:56:45] The whole point is so we can have these really flexible [07:56:47] substitution patterns which is different from constant elasticity [07:56:52] of supply so we haven't done any kind of comparison [07:56:56] but we definitely do think that that's one of our advantages [07:56:59] and we can think a little bit more about how we can contrast that. [07:57:03] There is also the question about whether we can derive [07:57:05] sufficient statistics or something that is like to a first order. [07:57:10] Like a sufficient condition. [07:57:13] We can think about that a little bit more. [07:57:15] It's a bit complicated but we will think on that. [07:57:22] There was the point about the Leontief production function. [07:57:26] Yeah, we think that's absolutely right in this context [07:57:28] where this is basically coming from chemical reactions. [07:57:31] Regarding the terminology of recipe versus Leontief, [07:57:34] we're definitely talking about the same thing. [07:57:37] We thought that recipe is a little bit more intuitive to people [07:57:40] but yeah, that's the production function that we're thinking of. [07:57:48] And then you made a point and John also made a point on the demand side [07:57:55] about dynamics. Let me first address the supply. [07:57:59] So yeah, when we think about oil prices, prices increase, [07:58:03] we get an expansion in supply. [07:58:05] I think we're much more limited in terms of new entry [07:58:08] in this context because mines take a really long time to enter. [07:58:13] It takes well over a decade to build some of these mines. [07:58:17] We can see that in the data but we do see in our data [07:58:20] kind of like the future pipeline of mines, [07:58:22] what mines are under construction and when they're expected to come online. [07:58:26] We actually do incorporate that. [07:58:28] So that's exogenous but great. [07:58:54] Thank you and point taken. [07:58:57] Then on the demand side, thinking about the dynamics there. [07:59:03] So I think what was brought up was stockpiling. [07:59:07] I think there's a limit to how much you can do this, [07:59:10] just kind of like the chemical stability of some of these minerals. [07:59:16] You might also imagine longer term stuff on the demand side [07:59:23] with these batteries. [07:59:24] We're taking the technologies as they're given today. [07:59:26] It's kind of a forecasting exercise and we do have forecasts [07:59:31] that we use on what expenditures are expected to be [07:59:33] on these different battery technologies into the future [07:59:36] and we use that for disciplining the expenditure on batteries in the future. [07:59:42] So that's what we're doing on the dynamics on the demand side. [07:59:46] Great. So let me ask you the last question and then we'll conclude. [07:59:50] I mean, first I like the paper for the same reason that Steve pointed out, [07:59:54] which is it's a really great attempt to say, [07:59:56] let's take the micro bottlenecks a lot more seriously [08:00:00] and you got pretty different conclusions [08:00:02] from what you would just do at a million mile high of all minerals. [08:00:08] It also seems that you're looking for more geopolitics. [08:00:11] So let me suggest two things and get your thoughts on it. [08:00:13] One is on the offensive side, when you have complementarities, [08:00:18] you might really crank export control and what you control, [08:00:22] not so much because of the direct effect on what you control, [08:00:25] but because of the indirect effect that you might have [08:00:28] somewhere else in the world through the complementarity. [08:00:31] So that might give you a reason why you really go hard [08:00:34] on the few things that you control in the presence of complementarity. [08:00:38] The other way around, you can think of it as [08:00:41] if I want to exert a given amount of power or control, [08:00:46] when you have multiple steps with complementarities, [08:00:49] there might be a way to minimize your cost in acquiring mines [08:00:53] by just picking the right parts of the chain [08:00:56] in some particular combinations to maximize how much you can exert [08:00:59] with potentially very little control. [08:01:02] So those might be applications to geopolitics or geopolitics, [08:01:06] but the paper is almost there and there are more words [08:01:10] and applications rather than the artwork that you already did. [08:01:14] Great, thank you so much for the suggestions. [08:01:17] Yeah, I think you highlight exactly the point that we're really trying [08:01:21] to hone in on with the complementarity. [08:01:23] And I agree with some goal in mind, [08:01:29] if you are producing something that has some compliments, [08:01:32] that complementarity gives you greater power and maybe to the extent [08:01:37] that you can choose what you control, [08:01:39] the complementarity certainly matters there for similar reasons. [08:01:42] So thanks for that suggestion and we'll definitely think about that. [08:01:48] Is that... Great. [08:01:50] Thanks, Sarah. [08:01:56] Thank you so much to everyone for coming here. [08:01:59] I know it's very hard to get people across fields to come together. [08:02:03] And so it's so wonderful to get a room of people working on [08:02:07] the same topics and from different directions and understand it. [08:02:11] So thanks so much, Tel Aviv, for being here all day, [08:02:13] all the presenters and especially all the discussants [08:02:15] for making today's success. [08:02:17] And hope to see you all here next year. [08:02:20] Thank you, Mr. Jeff.