The MakerDAO community generally looks at gas and the dust parameter from either a risk management or user access perspective. Smaller vaults generally have higher risk of losses during liquidation due to fixed gas expenses; Maker uses the dust parameter to set a minimum required vault debt, helping to mitigate this risk. On the other hand, Maker tries to limit this dust parameter so smaller users can continue to benefit from vaults.
Maker can also consider the dust parameter from a user cost perspective. Although users should have as much freedom as possible to make their own financial decisions, one could argue MakerDAO shouldn’t subsidize risk for services that will necessarily be a bad deal for the end user due to factors outside of Maker’s control (such as the gas costs of managing vault positions). Following from this, if vaults below a certain size are objectively a bad deal for users, we can safely raise the dust limit to this level without considering concerns about user access.
How can we determine what size of vaults are objectively a bad deal? One way to approach this is to calculate the implied APR (annual percentage rate) that gas costs will add to the total borrowing cost of a vault. This depends on several factors, including amount borrowed, prevailing gas prices, ETH price in USD, and the length of time the position remains open. Gas and ETH prices determine the fixed USD costs of using a vault - as they increase, the implied APR increases. These costs are amortized over the amount of funds borrowed and time the loan is maintained - as borrowed amount and term length increase, the implied APR will fall.
You can use this sheet for simulating the impact on borrowing cost as a function of the above inputs (please make a copy of the sheet before editing, and only adjust cells marked in grey). While the cutoff of what is an acceptable borrowing cost is open for debate, lending regulations in the US generally consider rates above 36% APR to be usurious and negative for borrowers and this seems like a reasonable hard upper limit. Once the implied APRs reach above 10%, using a vault likely becomes a much less favorable deal for small users.
The final cost to end users also depends on user collateralization level and price volatility, because the user may need to make additional transactions to top up collateral or repay debt during market downturns to avoid liquidation. Periods of price volatility tend to coincide with high gas prices, which increases the cost impact of additional maintenance transactions.
Below are a few simulations of potential user experience:
Low gas expenses - user has a preexisting DS Proxy and is able to manage their vault with the minimum amount of maintenance transactions, choosing low gas periods to submit transactions.
Moderate gas expenses - user creates proxy and opens vault during periods of low gas, but pays a higher gas price when they want to close the vault.
High gas expenses - user creates proxy and opens vault during low gas period, but is forced to defend their vault during a downturn and also pays higher gas prices when they close the position.
Based on the above simulations, it seems that vault sizes of 1,000 DAI and lower are generally uneconomical for users even under the most favorable assumptions. Under more realistic assumptions, vaults would need to borrow between 5,000-10,000 DAI or more to avoid excessive impact on their borrowing costs.
MakerDAO can consider this while voting on future changes or adjustments to the dust parameter. In general, this analysis questions the viability of subsidizing risk from smaller vault positions, as they offer relatively less benefit to their users due to factors outside Maker’s control. This also shows the importance of serving smaller users in cheaper execution environments such as L2 platforms (Optimism, Arbitrum, etc) or sidechains (xDAI, Polygon, etc) where transaction fees are lower and have less impact on overall vault cost.