The lack of pass-through of the LCFS subsidy can be easily visualized

Comparing the East Coast results here – which are for Trenton, NJ – to their results from Newark, NJ for unbranded E10, I find lower rates of pass through, especially when dropping the RIN Price Shock period. Like mine, their sample includes a period with a significant upward shock to RIN prices – the first eight months of 2013 – that has a significant impact on estimates of pass through. In that period, D6 RIN prices rose from under 10 cents/gal to over $1/gal and was the first time RINs represented a substantial portion of gasoline margins. They argue that much of the findings of incomplete pass through are driven by the market learning how D6 RIN prices affected rack margins. Figure 12 suggest a different story. The soybean boom that began in 2020 put substantial upward pressure on bio-diesel prices, and therefore, D4 RIN prices. RIN subsidies for bio-diesel surpassed $3/gal in 2021, more than tripling since 2020, and nearly double its all-time high . Blenders adjusted quickly to this, fully passing through the subsidy as it grew exponentially. This is visually clear in Figure 16, and suggests the RIN subsidy is salient to market participants. In fact, only when excluding the RIN Price Shock period can I reject complete pass through in the Gulf and East Coast. Although reaching record highs, the RIN subsidy was still meaningful prior to the RIN Price Shock period it represented around half the price of B100 for much of the time. Pass through of the RIN subsidy was lower prior to the RIN Price Shock period, so the salience argument doesn’t explain the finding of incomplete pass through of the RIN subsidy in this paper.In this section, pass-through of the LCFS subsidy to rack margins for bio-diesel are considered. An important difference relative to RIN subsidies, LCFS subsidies exhibits less daily variation,grow lights for cannabis which introduces a lack of statistical power, especially in the short run. Generally, however, the results presented in this section will suggest that LCFS subsidies is not fully passed through in the California rack markets in my sample.

Table 5 presents short- and long-run estimates from the unrestricted and restricted models in and , respectively. Columns 1 and 2 show that all short-run coefficients are statistically insignificant when the LCFS and RIN subsidies are included separately in the model. The coefficients in column 3, however, are statistically significant and suggest 70 cents/gal of the combined subsidy is passed through in the long-run, and it takes over a week to reach the long-run rate of pass through. The restricted model is more parsimonious, but the combined subsidy may better explain variation in rack margins since they are stacked, and blenders are total policy revenue is what’s salient to the blender.Long-run pass through of the individual and combined subsidies exhibits heterogeneity within in California. Table 6 shows that pass through is lower in larger cities in the sample and higher in the smaller cities . The RIN subsidy is fully passed through in the smaller cities on average, and 60 cents/gal of the LCFS subsidy is passed through on average, but the 95 percent confidence interval includes both zero and one. The 95 percent confidence interval for the combined subsidy in the smaller cities is. The majority of operational bio-diesel production capacity is in Los Angeles, San Diego, and San Francisco . Los Angeles and San Francisco also both have major spot market hubs for ULSD. Therefore, the remainder of this section will focus on those larger cities. Only half the RIN subsidy is passed through in the urban markets on average and pass through of the LCFS subsidy is not statistically different from zero. Therefore, complete pass through is rejected for all three variables.The results of the LCFS subsidy analysis discussed thus far have been averages over the full sample. So, for example, it could be that those estimates reflect an average of incomplete pass through in earlier years due to a lack of salience but pass through becomes complete later in the sample as the market better understands how LCFS credits affect margins . Also, the LCFS subsidies for bio-diesel have risen from an average of 9 cents/gal in 2015 to over $1/gal since 2018, so it may have been easier to hide changes in the subsidy when levels were so low.

To explore the aforementioned questions, I estimate and , keeping observations in Los Angeles and San Francisco, for each sub-sample defined in Table 2. I present those results in Figure 14. The first observation is the downward shift in the rates of pass through from the Low LCFS Price period to the Mid LCFS Price period. Pass through of the RIN subsidy, on average, is close to one in the Low LCFS Price period and falls to nearly a quarter in the next period. There isn’t much signal in the LCFS subsidy during the Low LCFS Price period, but the point estimate is also close to one. In the Mid LCFS Price period, pass through of the LCFS subsidy falls to 0.06 but the 95 percent confidence interval includes up to half. Since the LCFS subsidy estimates are so imprecise in the Low LCFS Price period, I’m not able to confidently rule out the salience argument, but the near-zero point estimates in later periods, and confidence intervals with upper bounds around a half, suggest that a change in salience is unlikely to be the driver of the finding of incomplete pass through of the LCFS subsidy when utilizing the full sample. Results from the restricted model resemble RIN pass through from the unrestricted model, which is to be expected as there is more variation in RIN subsidies than LCFS subsidies over most of the sample. These RIN subsidy pass-through estimates are put into context with other cities in the sample in Figure A-4, which shows that RIN subsidy pass through, apart from the East Coast, was complete or near complete in all sub-samples.Figure 13 showed some RIN subsidy pass-through heterogeneity across blends in California. Figure 15 paints a similar picture for the LCFS subsidy; B20 is the only blend with a confidence interval including complete pass through and point estimates decrease with the share of ULSD in the blend. This result, combined with Figure 13, is consistent with blenders having greater market power in the higher blends, especially above B20. The bottom panel of Figure 16 plots rack margins for B100 in Los Angeles, along with the LCFS subsidy, the RIN subsidy, and the combined subsidy. It is important to reemphasize that west coast margins are calculated using the B100 spot price in NY Harbor Barge and therefore neglect transportation and other costs, which will inflate the observed margins above true levels by an unknown amount. I assume these costs are constant over time but don’t speculate as to their level. With that assumption, I plot the same thing but shifting margins down to better visualize comovement with the subsidies in the top panel of Figure 16. At first glance, it appears that the rack margin follows the combined subsidy well , especially in early 2016 when the LCFS credit price rose. That’s not the case, however, because the BTC was put back in place in 2016, and, under the assumptions laid out earlier in this section, blenders received an additional 50 cents/gal of B100 from the BTC. This, coincidentally,indoor cannabis grow system coincided with the LCFS subsidy increasing by about 50 cents/gal of B100 due to the modeling change by CARB. In subsequent years, rack margins generally followed movements in the RIN subsidy, but not the LCFS or combined subsidies. I use Los Angeles B100 as an illustrative example in Figure 16; however, the picture looks similar in San Francisco and other blends.Pass through of the LCFS subsidy to blended bio-diesel prices is not statistically different from zero, and in the preferred specification, pass through is 0.02 on average. There are a few important limitations that could significantly impact the results presented above. First is, again, the lack of a clear understanding how the BTC affected the relationship between RIN subsidies and rack margins, and whether the effect is different for California relative to other U.S. regions.

Second is the absence of a California-specific spot price of SME B100. If the assumption that the spot price, or the blender’s marginal cost, of SME B100 is equal to a NY Harbor Barge basis plus a constant transportation cost is violated, the estimates of LCFS subsidy pass through will be biased if California-specific costs are correlated with the LCFS subsidy. Specifically, if transportation costs exhibit a positive relationship with the LCFS subsidy, I will systematically underestimate the rate it is passed through. With the data available at the time of writing, the prevalence of that relationship is untestable. Lastly, margins for diesel blended with SME bio-diesel are volatile and data availability limit the analysis to SME bio-diesel only, which is rarely used in California. The advantage of focusing on SME bio-diesel, however, is that its CI has been relatively constant over time . Yet, if the CI of SME bio-diesel used in California changes significantly within any given year, it will have two direct effects on the models used to estimate subsidy pass through, and . The first is that the true spot price of SME B100 in California will likely move inversely with the CI, which will violate the assumption that the California spot price equals the NY Harbor spot price plus a constant. The second is that the implicit subsidy calculation in will be inaccurate. Take a simplified example. Suppose blenders start purchasing bio-diesel with a lower CI and a higher price mid-year and nothing else changes. The observed margin in will not change because I don’t observe the true California spot price. Since the number of credits per gallon would not change. The subsidy calculated in would only change if the decreasing CI is associated with an increase in the LCFS credit price, which would attenuate the estimates of LCFS subsidy pass through. However, it’s unlikely CI scores of SME bio-diesel are changing meaningfully within years in my sample. Accurate, high-frequency bio-diesel, as well as renewable diesel, pricing data for feed stocks and localities would allow researchers to study pass through using prices for fuels that reflect a much more significant market share of the biomass-based diesel consumed in the state. Additional data on the CI of fuels in the rack pricing data would allow for more accurate calculations of the implicit LCFS subsidies they receive, instead of assuming the volume weighted average CI score for the listed feed stock. Quantity data would also be particularly useful in the LCFS analysis given how thin the market for SME bio-diesel is in California. Many of the rack prices in my sample for California may come from transactions with relatively small quantities of fuel and may not be fully representative of the entire market.Credit market data for the CFP is only available since 2017 and there hasn’t been much variation in credit prices. Additionally, spot prices in the PNW and Oregon rack margins are very volatile. This makes identifying pass through of the CFP subsidy difficult and results presented here are imprecise. Table 7 presents the short- and long-run estimates from and , using the Oregon cities in the sample. All estimates of CFP pass through are statistically insignificant. Like California, short-run estimates of the combined subsidy are measured more precisely than either of the individual subsidies. An important contribution of this paper is utilizing institutional details and context to build an understanding of how the stacked costs and incentives from multiple market-based environmental policies propagate through the supply chain of fuels. This paper provides a framework to evaluate tax and subsidy pass through in the diesel sector, which is nontrivial, especially relative to the gasoline sector, due to the intermittent nature of the Blender’s Tax Credit . Although a similar tax-subsidy scheme is used, pass through of the LCFS and CFP taxes and subsidies can’t be evaluated in the same way as the RFS, due to nuanced policy differences. Section 1.4 showed that applying the same framework to the LCFS leads to the wrong conclusion about tax pass through.