Table 2.
Changes to particle and ozone concentrations associated with variation in the gasoline-ethanol fuel mix
| Column number: | (1) | (2) | (3) | (4) | (5) |
| Dependent variable: | BC | PM2.5 | PM 100–800 nm | UFP 7–100 nm | Ozone |
| Unit: | µg m−3 | µg m−3 | cm−3 | cm−3 | µg m−3 |
| Mean over hour window: | 08:00 | 24-h | 08:00 | 08:00 | 12:00–16:00 |
| Sample period: | Oct/2010 to Apr/2011 | Nov/2008 to | Jan/2011 to | Jan/2011 to | Nov/2008 to |
| & Oct to Nov/2012 | May/2013 | May/2011 | May/2011 | May/2013 | |
| Number of sampling sites: | 1 | 3 | 1 | 1 | 12 |
| Source: | Own | CETESB | Own | Own | CETESB |
| Share of Gasoline E20/E25 in the flex fleet rises from 30 to 80% | −0.3 ± 1.9 | 0.2 ± 3.9 | −1,249 ± 1,669 | 8,713 ± 4,559 | −8.3 ± 5.0 |
| Equivalently, share of Ethanol E100 in the flex fleet falls from 70 to 20% | |||||
| Control variables (to correct for the influence of other determinants of particles) | |||||
| Site-specific linear trend | Yes | Yes | Yes | Yes | Yes |
| Week-of-year fixed effects | No | Yes | No | No | Yes |
| Day-of-week fixed effects | Yes | Yes | Yes | Yes | Yes |
| Radiation (+100 W m−2) | 0.5 ± 0.7 | −0.4 ± 2.2 | 4 ± 825 | 235 ± 1,798 | 4.2 ± 0.7 |
| Temperature (+1 oC) | 0.0 ± 0.2 | 1.2 ± 0.5 | 236 ± 235 | −847 ± 836 | 3.1 ± 0.4 |
| Humidity (+10%) | 0.1 ± 0.7 | −1.0 ± 1.6 | 349 ± 721 | −1,020 ± 1,712 | −4.9 ± 1.3 |
| Wind speed (+1 m s−1) | −3.2 ± 1.2 | −6.5 ± 2.8 | −2,102 ± 1,410 | −4,217 ± 3,489 | −13.2 ± 2.1 |
| Other meteorological and road traffic conditions (see notes) | Yes | Yes | Yes | Yes | Yes |
| R 2 | 62.0% | 73.4% | 76.0% | 69.8% | 70.7% |
| Number of observations | 129 | 511 | 80 | 80 | 13,203 |
| Number of regressors | 18 | 74 | 19 | 19 | 96 |
| Mean value of dependent variable | 6.0 | 13.8 | 3,577 | 18,659 | 72.2 |
Coefficients and 95% confidence intervals, i.e., point estimate ± 2 standard errors. An observation is a date (columns 1, 3, 4) or a date-site pair (columns 2, 5). Samples exclude the colder months of June to September, and include all days of the week (columns 2, 5) or non-holiday weekdays only (columns 1, 3, 4). Radiation, temperature, humidity, and wind speed in the recorded unit. All columns additionally include several precipitation, thermal inversion and road traffic congestion indicators. Columns 1 to 4 further control for wind direction and column 5 follows Supplementary Table 4. Since the longer samples encompass 2010, columns 2, 5 include site-specific intercepts indicating the opening of the Greater São Paulo beltway’s southern section on March 31, 2010. The effect of raising the gasoline share in the flex fleet is scaled for in-sample variation from 30 to 80%. The corresponding variation in the ethanol share is one minus variation in the gasoline share. Ordinary Least Squares (OLS) estimates, with standard errors calculated by bootstrapping (200 samples each): (i) the consumer-level fuel choice data, to account for sampling variation in the predicted gasoline share in a first-step consumer demand model, and (ii) the pollutant-meterology-traffic data in the second-step particle regression, clustering by date