COST EFFECTIVENESS ESTIMATES FOR MOBILE SOURCE EMISSION CONTROL George Miller 10 November 1977 • '..-v Submitted to: Environmental Protection Agency Office of Mobile Source Air Pollution Control Purchase Order CD-7-017T-A REPORT NUMBER VRI EPA-3 FR 77-1 (R) Vector Research, Incorporated Ann Arbor. Michigan ------- EPA-420-R-77-102 CONTENTS Page 1.0 INTRODUCTION 1 2.0 GENERAL APPROACH AND ASSUMPTIONS .... 3 3.0 DETAILS OF THE ANALYSIS 7 3.1 Effectiveness Computations 7 3.1.1 Case 1: Standards Only 8 3.1.2 Case 2: Standards Plus Certification 9 3.1.3 Case 3: Standards Plus Certification with Parameter Adjustment 14 3.1.4 Cases 4 and 5: Standards Plus Selective Enforcement Audit 14 3.1.5 Case 6: Standards Plus Inspection and Maintenance (Diagnostic) 20 3.1.6 Cases 7 and 8: Standards Plus Inspection and Maintenance (Testing) 21 3.1.7 Cases 9 and 10: Standards Plus Recall 25 3.1.8 Case 11: Standards Plus Certification with Parameter Adjustment Plus Selective Enforcement Audit 27 3.1.9 Case 12: Standards Plus Certification with Parameter Adjustment Plus Inspection and Maintenance (Testing) 29 3.1.10 Case 13: Standards Plus Certification with Parameter Adjustment Plus Selective Enforcement Audit Plus Recall 30 3.1.11 Cases 14 and 16: Standards Plus Certifi- cation with Parameter Adjustment Plus Selective Enforcement Audit Plus Inspection and Maintenance (Testing) Plus Recall 30 3.1.12 Case 15: Standards Plus Selective Enforcement Audit Plus Inspection and Maintenance (Testing) 31 3.1.13 Case 17: Standards Plus Selective Enforcement Audit Plus Recall 32 3.1.14 Case 18: Standards Plus Selective Enforcement Audit Plus Inspection and Maintenance (Testing) Plus Recall 33 3.1.15 Case 19: Standards Plus Certification Plus Inspection and Maintenance (Testing) ... 33 ------- i i CONTENTS (Concluded) Page 3.2 Cost Estimates 34 3.2.1 Standards 34 3.2.2 Certification 36 3.2.3 Selective Enforcement Audit 36 3.2.4 Inspection and Maintenance 37 3.2.5 Recall 37 4.0 RESULTS 39 REFERENCES 65 ------- i i i FIGURES Title Determination of Best Estimate and Range of Values for Average Emission Rates for Case 2 (Standards Plus Certification) Alternative Deterioration Assump- tions for Best Estimates and Low Estimates of IM Effectiveness for a Single Vehicle Effect of Recall on Emissions of a Recalled Vehicle ------- iv ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Page 4 10 12 16 18 23 24 35 42 43 44 45 46 47 48 49 v TABLES Title Compliance Assurance Programs Included in the Cost Effectiveness Computations Regression Analysis of In-Use Emissions as a Function of Certification Emissions Regression Analysis of In-Use Emission Rates as a Function of Mileage for Case 2 Regression Analysis of In-Use Emission Rates as a Function of Mileage for Case 3 Regression Analysis of In-Use Emissions as a Function of Assembly Line Emissions Regression Equations Used to Relate Short- Test Results to FTP Levels in the IM Simulation Short Test Cutpoints and Corresponding Percent Savings from Application of IM Simulation Summary of Costs Used in the Analysis Results for Case 1 - Standards Only Results for Case 2 - Standards Plus Certification Results for Case 3 - Standards Plus Certification with Parameter Adjustment Results for Case 4 - Standards Plus Selective Enforcement Audit (100% Rejection Rate) Results for Case 5 - Standards Plus Selective Enforcement Audit (75% Rejection Rate) Results for Case 6 - Standards Plus Inspection and Maintenance (Diagnostic) Results for Case 7a - Standards Plus Idle Inspection and Maintenance (30% Failure Rate) Results for Case 7b - Standards Plus Loaded Inspection and Maintenance (30% Failure Rate) ------- 17 18 19 20 21 22 23 24 25 26 27 28 Page 50 51 52 53 54 55 56 57 58 59 60 61 vi TABLES (Continued) Title Results for Case 8a - Standards Plus Idle Inspection and Maintenance (50% Failure Rate) Results for Case 8b - Standards Plus Loaded Inspection and Maintenance (50% Failure Rate) Results for Case 9 - Standards Plus Recall (40% Response Rate) Results for Case 10 - Standards Plus Recall (70% Response Rate) Results for Case 11 - Standards Plus Certifi- cation with Parameter Adjustment Plus Selective Enforcement Audit Results for Case 12 - Standards Plus Certifi- cation with Parameter Adjustment Plus Inspection and Maintenance Results for Case 13 - Standards Plus Certifi- cation with Parameter Adjustment Plus Selective Enforcement Audit Plus Recall Results for Case 14 - Standards Plus Certifi- cation with Parameter Adjustment Plus Selective Enforcement Audit Plus Inspection and Maintenance (30% Failure Rate) Plus Recall Results for Case 15 - Standards Plus Selective Enforcement Audit Plus Inspection and Maintenance Results for Case 16 - Standards Plus Certifi- cation with Parameter Adjustment Plus Selective Enforcement Audit Plus Inspection and Maintenance (50% Failure Rate) Plus Recall Results for Case 17 - Standards Plus Selective Enforcement Audit Plus Recall Results for Case 18 - Standards Plus Selective Enforcement Audit Plus Inspection and Maintenance Plus Recall ------- vi i TABLES (Concluded) Number Title Page 29 Results for Case 19 - Standards Plus Certifi- cation Plus Inspection and Maintenance 62 30 Cost Effectiveness - Federal Government Dollars Spent Per Ton of Pollutant Removed 63 31 Cost Effectiveness - Total Dollars Spent Per Ton of Pollutant Removed 64 ------- viii ------- 1 1.0 INTRODUCTION This document is a final report of an effort conducted by Vector Research, Incorporated, (VRI) to develop cost and effectiveness estimates for alternative mobile source compliance assurance programs for the Environmental Protection Agency (EPA). The estimates are intended to be used as input to a mobile source strategy paper being developed by EPA. This report includes a general description of the approach and assumptions used in developing the estimates (in section 2.0), a detailed description of the procedures used to develop the estimates for each compliance assurance program studied (in section 3.0), and a presentation of the results of the analysis (in section 4.0). ------- 2 ------- 3 2.0 GENERAL APPROACH AND ASSUMPTIONS The cost effectiveness computations were performed for a set of alternative compliance assurance programs (referred to as cases) selected by EPA and summarized in table 1. Each of these programs assumes the existence of emission standards, and may also include one or more of the following compliance techniques: (1) a certification process as currently performed by EPA; (2) a certification process in which EPA may adjust certain vehicle parameters within manufacturer specifications before testing any vehicle; (3) EPA's assembly line testing program (Selective Enforcement Audit); (4) a state-operated inspection and maintenance program involving diagnostic checks on emission-related equipment; (5) a state-operated inspection and maintenance program involving emissions testing; and (6) a recall program. For each case, effectiveness was computed in terms of reduction in 1975 FTP emissions from pre-1968 levels. Separate effectiveness computations were performed for four pollutants (HC, CO, N0X, and evaporative HC). All analysis was based on 1975 standards in 49 states, and considered low-altitude, non-California light duty vehicles only. Effectiveness is expressed as tons of pollutant over a 100,000-mile vehicle life, while costs are expressed as lifetime costs in 1975 dollars. Both total and government costs are presented for each case. ------- TABLE ^: COMPLIANCE ASSURANCE PROGRAMS INCLUDED IN THE COST EFFECTIVENESS COMPUTATIONS COMPLIANCE TECHNIQUES CASE1 CERTIFICATION SELECTIVE INSPECTION AND INSPECTION AND STANDARDS CERTIFICATION WITH PARAMETER ENFORCEMENT MAINTENANCE MAINTENANCE RECALL" ADJUSTMENT AUDIT2 (DIAGNOSTIC) (TESTING)3 1 2 X X X 3 X X 4 X X (100%) 5 X X ( 75%) 6 X X 7a X X (30%) 7b X X (30%) 8a X X (50%) 8b X X (50%) 9 X X (40%) 10 X X (70%) 11 X X X (100%) 12 X X X (30%) X (70%) 13 X X X (100%) 14 X X X (100%) X (30%) X (70%) 15 X X (100%) X (30%) X (70%) 16 X X X (100%) X (50%) 17 X X (100%) X (70%) 18 X X (100%) X (30%) X (70%) 19 X X X (30%) *Cases 7a and 8a assume the use of an idle test for inspection; cases 7b and 8b assume the use of a loaded test. All other cases involving inspection and maintenance (testing) assume the use of the idle test. 2Parenthetical numbers in this column refer to assumed fraction of test orders which result in rejection of the lot. Parenthetical numbers in this column refer to failure rate which the states are assumed to enforce. Parenthetical numbers in this column refer to assumed response rate to a recall. ------- 5 Because of uncertainty regarding the validity of many of the assumptions of this analysis, and scarcity of data for some of the compliance techniques, a range of estimates has been developed for each case. A best estimate is presented as the cost and effectiveness judged most likely to result from implementation of the case. Low and high estimates are presented as "reason- able" lower and upper limits on effectiveness, with corresponding costs. Because these ranges of estimates represent uncertainty in estimating the effectiveness and costs of the various compliance techniques, the range tends to be smallest for cases involving techniques for which substantial data exist (such as certification). The primary data sources used in this analysis were the 1974 and 1975 Emission Factors Programs. Data from 1975 vehicles from these programs pro- vided a population of in-use vehicles for which to estimate the effectiveness of various combinations of compliance techniques. Other data (such as 1975 certification results, data from the Restorative Maintenance Program, 1975 assembly-line data, and miscellaneous sets of evaporative emission data and cost data) were used where needed;1 all data sources and computational pro- cedures are described in the following section. ^11 data used were current as of July 1977. ------- 6 ------- 7 3.0 DETAILS OF THE ANALYSIS This section describes the analytical procedures used in the cost effectiveness calculations. Section 3.1 discusses the effectiveness assessment methodology for each combination of compliance techniques evaluated, while section 3.2 describes the derivation of the cost estimates used in the analysis. 3.1 Effectiveness Computations Effectiveness results were computed for each case in terms of millions of tons of pollutant saved over a precontrol situation. Precontrol emission rates were assumed to average 7.55 grams/mile for HC, 86.79 grams/mile for CO, 3.40 grams/mile for NO , and 3.33 grams/mile for evaporative HC A during the life of a given vehicle. In all cases the standards assumed to be in effect for exhaust emissions were 1.5 grams/mile of HC, 15 grams/mile of CO, and 3.1 grams/mile of NO . In addition, in-use evaporative HC levels A (measured with the SHED technique) were adjusted to give a modified SHED standard for 1975. To do this it was assumed that prototype vehicles actually built for certification 1n 1975 met the evaporative standard exactly. Cannister test data on twenty 1973 model year vehicles in Los Angeles was examined to develop a relationship between certification and in-use emissions.1 In particular, the ratio of average in-use emissions to the average certification level was computed to equal 2.11. This relationship xThe particular in-use cannister data used correlated fairly well with SHED data, which was also available for these twenty vehicles. ------- 8 was used to predict the standard from the average in-use level of 1975 vehicles. The result was a predicted evaporative standard of 0.834 when expressed in units of grams/mile. Each of the following subsections presents the approach used to compute the effectiveness for a single case or group of similar cases. These presenta- tions generally describe the procedures followed to compute average emission rates in grams per mile under the various strategies investigated. In all cases, these rates were converted to effectiveness values using the following relationship: r^ 3 (uj - c.j) LPk, where r.j = reduction in emissions of pollutant i resulting from implementation of a particular case (in tons) , u.j = uncontrolled average lifetime emission rate of pollutant i (grams per mi le), c.j = controlled average lifetime emission rate of pollutant i resulting from implementation of the case (grams per mile), L = average vehicle life (taken to be 100,000 miles), P = vehicle population to which the benefit applies (taken to be 10 million vehicles), and k = conversion factor to convert grams to tons. 3.1.1 Case 1: Standards Only To estimate the reduction in emissions resulting from promulgation of standards without any additional compliance techniques, it was assumed that ------- 9 manufacturers would respond by designing vehicles which would just meet the standards as prototypes. In order to convert these design emission levels to levels for vehicles in use, regression analysis was performed on data for 570 in-use vehicles from model year 1975 and certification data for corresponding vehicles to develop a relationship between design exhaust emission levels within a given engine family, and 50,000-mile in-use levels experienced by vehicles in that same family. The regression results are given in Table 2. The in-use data were for low-altitude, non-California, light duty vehicles tested in the FY 1974 Emission Factors (EF) Program. Mileage regressions described in section 3.1.2, were used to project these in-use levels to 50,000 mile levels for each vehicle. Because the stan- dard for evaporative emissions was inferred from the 1975 data (as described earlier), the corresponding in-use level is simply the level actually realized in 1975 vehicles of 1.76 grams/mile. No deterioration of evapora- tive emissions with mileage was assumed. A range of values around this best estimate was developed by assuming that, at worst, there would be no reduction in emissions from precontrol levels and, at the other extreme, the reduction would be no better than the most optimistic estimate of the reduction achieved in the presence of certification (I.e., the high estimate for case 2). 3.1.2 Case 2: Standards Plus Certification The effectiveness estimates for the case in which standards and certi- fication are 1n effect were based on linear regressions of EF data on 1975 model year vehicles. These regressions predict mean exhaust emission rates as a function of vehicle mileage for a population of vehicles subject to ------- TABLE 2: REGRESSION ANALYSIS OF IN-USE EMISSIONS AS A FUNCTION OF CERTIFICATION EMISSIONS Relation: + fl.. (1 - 1. 2, 3) where y^j = in-use emission level of pollutant i by a given vehicle j at 50,000 miles; = average certification emission level (including deterioration) for pollutant i for vehicles in the same engine family as vehicle j; and o oij, ^ = regression coefficients. Result8: i ai a i R2 Value Significance Value Significance 1 (HC) 0.37177 0.0268 2.1617 0.0000 0.00861 2 (CO) 1.3517 0.0310 24.535 0.0000 0.00816 3 (N0X) 0.69533 0.0001 1.1748 0.0043 0.02692 ------- n the conditions of this.case. The data used for these regressions were for low-altitude, non-California, light duty vehicles tested in the FY 1974 and FY 1975 EF Programs. To obtain the best possible fit to the data, two separate sets of regressions were performed and combined into a single emissions-versus-mileage regression for each pollutant.1 The first set of regressions used data for the 533 vehicles in the data base for which the measured idle CO level (ICC) was less than 1.5 percent; the second set used the remaining 232 vehicles for which the ICO was greater than or equal to 1.5 percent. The results of the regressions, and the composite estimate of emissions for each pollutant, appear in table 3. The best estimate of emissions was then taken as the value of the composite regression equation at 50,000 miles. To develop an estimate for average evaporative emission rates, 1978 certification data were examined to determine the average percent of the standard at which vehicles were certified. The average evaporative emission rate for vehicles in the field was then assumed to equal the product s • f • fc, where s 3 assumed 1975 standard for evaporative emissions (.83 grams per mile); f » average fraction of 1978 standard experienced on 1978 certification test results (0.30); and i " ratio of in-use evaporative levels to certification levels (2.11, as computed in the derivation of the assumed 1975 standard). Upper and lower bounds on exhaust emissions for this case were compu- ted from upper and lower 95 percent confidence limits on the composite 1See [Williams, 1976] for a description of and rationale for the procedures used to combine these regressions. ------- TABLE 3: REGRESSION ANALYSIS OF IN-USE EMISSION RATES AS A FUNCTION OF MILEAGE FOR CASE 2 i^1 Relation: e^(m) = + n^li m + nli^ + n^ + (v>21 m + n2i^ = m + ni 0 = K 2, 3) where e.(m) = average in-use emissions of pollutant i by a vehicle at mileage m; n. = number of vehicles in sample k (where k=l for idle CO less than 1.5 percent, and k=2 otherwise); ro m = vehicle mileage, in tens of thousands; yki* nki = re9re5si°n coefficients; _")"])+ y2i. and Mi n^ + n2 - = nlqli * n2n2i ni n^ + n2 ------- TABLE 3: REGRESSION ANALYSIS OF IN-USE EMISSION RATES AS A FUNCTION OF MILEAGE FOR CASE 2 (Concluded) Regression Results: k nk i \\ nk1 R2 K Value Significance Value Significance 1 1 (HC) 0.19280 0.0000 0.89979 0.0000 0.04471 533 2 (CO) 1.7824 0.0027 12.872 0.0000 0.01685 (IC0<1.5) 3 (N0x) 0.17265 0.0002 2.2916 0.0 0.02625 2 1 (HC) 0.21144 0.0228 1.7917 0.0000 0.02234 232 2 (CO) 2.7543 0.1793 39.486 0.0000 0.00783 (IC0>1.5) 3 (N0X) 0.0072075 0.9249 2.4507 0.0000 0.00004 LO Composite Results: i "i ni 1 (HC) 0.1984 1.1674 2 (CO) 2.0740 19.9562 3 (N0x) 0.1230 2.3393 ------- 14 emission-versus-mileage regression lines at 50,000 miles (see figure 1). The upper limit on evaporative emissions was taken to be equal to the best estimate (since the best estimate resulted in an in-use value well below the standard), while the lower limit used the in-use evaporative rate actually experienced in 1975. 3.1.3 Case 3: Standards Plus Certification with Parameter Adjustment To represent the effect of EPA parameter adjustment on the effective- ness of certification, the mileage regressions described for case 2 were repeated on only those 1975 vehicles in the EF data base which were within manufacturer specifications on idle CO, idle RPM, and timing.1 The re- gression results are presented in table 4. The high and low estimates again were based on 95 percent confidence intervals around each of these regression lines at 50,000 miles. Under the assumption that parameter adjustment would not affect evaporative emissions, the case 2 estimates for evaporative effectiveness were used for case 3. 3.1.4 Cases 4 and 5: Standards Plus Selective Enforcement Audit The effectiveness of Selective Enforcement Audit (SEA) was estimated with the use of 1975 assembly-line data provided voluntarily by the manu- facturers to EPA1s Mobile Source Enforcement Division. Using these data LA vehicle was considered to be within manufacturer specifications if its idle CO was less than 0.5 percent, and if it was within 150 RPM of the manufacturer's specification for idle RPM and within 2 degrees of specifi- cation for timing. There were 314 such vehicles in the data base. Note that the requirement that idle CO be less than 0.5 percent eliminated the need for the kind of composite regression used for case 2. ------- Emission rate for a given FIGURE 1: DETERMINATION OF BEST ESTIMATE AND RANGE OF VALUES FOR AVERAGE EMISSION RATES FOR CASE 2 (STANDARDS PLUS CERTIFICATION) ------- 16 TABLE 4: REGRESSION ANALYSIS OF IN-USE EMISSION RATES AS A FUNCTION OF MILEAGE FOR CASE 3 Relation: e.. (m) = vi.j m + (i = 1> 2, 3) where e. (m) = average in-use emissions of pollutant i by a vehicle at mileage m; m = vehicle mileage, in tens of thousands; and u.j > n.j = regression coefficients Results: i ni R2 Value Significance Value Significance 1 (HC) 0.21532 0.0000 0.76435 0.0000 0.06059 2 (CO) 1.1438 0.1510 12.071 0.0000 0.00660 3 (N0X) 0.23168 0.0010 2.2729 0.0000 0.03395 ------- 17 and 1975 EF data for 282 vehicles, regression analyses (summarized in table 5) were performed to estimate the relationships between assembly- line emission levels (with deterioration factors applied) and levels of vehicles in use (converted to 50,000 mile levels using the mileage regres- sion described in section 3.1.2). The best estimate of effectiveness then assumed that manufacturer design, production engineering, and quality control would result in vehicles which just met the standards at the assembly line. The regression relationships were used to convert these assembly-line levels to levels for vehicles in use, and these in-use levels were further reduced to account for the effect of rejections by SEA. This effect was determined by assuming that an SEA failure would result in the emissions of failing vehicles in a class being reduced to the average emission levels of passing vehicles in the same class. The voluntarily submitted data were then used to estimate the average saving per failing class1 corresponding to this assumption, and these savings were converted to in-use reductions using the regression relations given in table 5. To apply these reductions to overall emissions, it was assumed that (1) 20 test orders per year would be conducted; (2) 50 percent of vehicles in a tested class would have been sold (and not subject to the reductions2) by the time of the test order; (3) 100 percent of classes would fail for the case 4 estimate, while 75 percent would fail for the case 5 estimate; and 1Failing classes were identified in the data by assuming a 40 percent AQL for each pollutant. 2In cases Involving recall as well as SEA, 1t was assumed that these vehicles would be recalled (see section 3.1.10). ------- 18 TABLE 5: REGRESSION ANALYSIS OF IN-USE EMISSIONS AS A FUNCTION OF ASSEMBLY LINE EMISSIONS Re lation: where ij ij Zij Yi,6i Y . Z. . + 5 • 1 1J 1 |