R Notebook: Artemis MKTCAP

Stablecoins mktcap data

On choisit les stablecoins suivant : USDT, USDC, DAI, PYUSD, FDUSD, EURC. Ces tokens structurent l’essentiel des volumes/market cap USD sur les classements “stablecoins” et catégories stablecoins, et parce qu’elles ont suffisamment d’observations pour cette étude.

On récolte ensuite la market capitalization de ces stablecoins sur Artemis.xyz, une plateforme d’analyse des données de crypto.

Pourquoi la market cap ?

• La market capitalization est définit par : Supply x Price

• un choc (speech positif envers CBDC) va impacter la demande de stablecoin pour différentes raisons,

• demande < offre -> pression à la baisse sur les prix,

• donc baisse des prix des cryptos -> baisse market cap qui ne serait pas captée par la supply uniquement,

• si pressions trop fortes, l’émetteur intervient en réduisant la supply (burn);

• donc baisse de la supply -> baisse market cap;

• En utilisant la supply, on ne capte qu’une partie des effets. A la rigueur, la supply peut être représentative dans le cas des algorithmic stablecoins car l’ajustement de l’offre est fait de manière automatique.

Data import

We also collect price and supply for those stablecoins from Artemis. We fill NAs with linear interpolation.

In this article https://www.artemisanalytics.com/resources/outstanding-supply by Artemis.xyz, they suggest to use the Outstanding Supply of stablecoin to be closer to stock supply (Outstanding Shares in stocks) and have a good representation of the true circulating supply. We gather data on Outstanding Supply from TokenTerminal (https://tokenterminal.com/explorer/metrics/outstanding-supply) : (The data are available on this link : https://tokenterminal.com/explorer/studio/charts/new?v=ZGRlNjhlMWJkNDYzZDM0ZDdjNGMxZmUy).

Final Database

We drop dates of stablecoins supply after July 2025 because the last observation is 2025-06-16 for speeches about CBDC, and adding too much after this dates will just add zeros.

Note : je n’ai pas les données de prix et de circulating supply avant 2020 parce que j’ai épuisé mon quotas d’extraction de données, je pourrai en récupérer plus à partir du 1er février.

Models

We estimate the following local projection model :

\[ Y_{i,t+h}-Y_{i,t-1}= \alpha_i + \beta^h X_t + \gamma Z_t + \mu_{i,t} \] Where \(Y\) is the log-difference of stablecoin supply, \(\alpha_i\) is an individual FE, and \(X_t\) is our index of speech CBDC-related sentiment/stance. \(Z_t\) are control variables representing the VIX and the GPRD and their lags, and dummies representing Terra-Luna and FTX crashes. We include individual (\(\alpha_i\)) fixed-effects.
Note : we take the 90% interval confidences for all estimations. Standard errors are heteroskedasticity robusts (clusterized by stablecoin).

Daily

Sentiment / Stance

Sentiment, as measured by Auer & al. (2021) tend not to capture the effect of CBDC’s related communications on stablecoin supplies. However, the Machine Learning framework does capture the negative effect of an increasing sentiment/stance from Central Banks, suggesting a substitution between stablecoins and CBDC. This effect is relatively small yet significant, and may be due to the fact that an important number of speeches are not particularly focused on CBDC, but may treat several subjects that Central Banks take care of and mention marginally CBDC eventually just to talk about it without really using a forward guidance design. We thus focus on important speeches, i.e. communications that explicitly focus on CBDC in their title (Note : another approach could be to weight the speeches by the number of sentences related to CBDC, or the share of these sentences in the total number of sentences).

Immediate effect following important CBDC-related speeches, i.e. speeches that explicitly mention CBDC or related keywords in their title. The effect is almost immediate, and the communications tend to decrease stablecoins supply by 0.3% after about a month.

In addition, we can decompose the market capitalization effect on two effects : the price effect, and the supply effect. We retrieve both price and supply from Artemis.xyz.

Overall, we can see that an increase in sentiment/stance of CB towards CBDC tend to decrease the market capitalization of stablecoins, mainly through a decrease in supply to defend the parity. Indeed, the price do not tend to move while the supply does, implying a decrease in demand for stablecoins to ensure that demand = supply so that price is at par.

We can also test with the Outstanding Supply :

Decomposition by type

Une hausse du sentiment/stance à propos des MNBC general ou non-spécifié ne tend pas à créer une baisse de la demande pour les stablecoins

Decomposition by Process

Decomposition by type and process

Features

Process

Risk-Benefits

Decomposition by Markets/Central Banks

DevEme

ECB vs other

What we observe overall : - An increase in sentiment/stance tend to decrease the demand for stablecoins - This results hides heterogeneity - An increase in sentiment/stance for retail and wholesale CBDC decrease the demand for stablecoins, while an increase in sentiment/stance for General/Unspecified CBDC tend to increase the demand or have no effect depending on the discourse associated. - An interpretation is that when Central Bank talk about General/Unspecified CBDC, it contains information about the project of other central banks “central banks of major countries are carrying out in-depth research recently on digital currencies” or be interpreted - in some cases - as a sign that central banks will adopt an already-existing stablecoin. - However, when the speech clearly states that it aims to issue a retail/wholesale stablecoins, it is perceived as a substitute for existing stablecoins. - Notably, when central banks announce progress on Wholesale (Features particularly) or Retail (Process particularly) CBDC, it dampens the demand for stablecoins. - The effect is particularly visible when Central Banks from Developed Markets communicate about CBDC - and especially ECB.

Important speeches only

Decomposition by type

Decomposition by Discourse in important speeches

Decomposition by type and process in important speeches

Features

Process

Risk-Benefit

Note : Je pense qu’on peut utiliser uniquement les speeches importants, ils donnent des résultats plus clairs et c’est assez facile de justifier leur utilisation + les résultats sont robustes mais moins marquants quand on utilise tous les speeches.