HAZARDOUS WASTE & HAZARDOUS MATERIALSVolume 11, Number 2, 1994Mary Ann Liebert, Inc., Publishers

Measuring Toxics Use Reductionin the Metal Intensive IndustriesTIMOTHY J. GREINER,1 JOSEPH E. PALUZZI,1

LORI COLOMBO,2 SANDRA WYMAN,2 and CHRIS WHICHARD2'Massachusetts Office of Technical Assistance

100 Cambridge StreetBoston, MA 02202

2The Environmental Careers Organization286 Congress StreetBoston, MA 02210

ABSTRACT

This study examined toxics use reduction measurement among metal intensive manufacturers(SIC 34-39) located in Massachusetts. Of the 110 study firms chosen from a stratified randomsample, less than half said they documented and measured their toxics use reduction progress.During on-site visits with twenty-eight of these firms, roughly half measured their progressaccurately. Toxics use reduction measurement errors were most frequently due to one or a

combination of five causes: use of poor normalization factors, inaccurate math calculations, materialsaccounting mistakes, incorrect chemical balances, and poor production unit definitions. Inreconciling these errors, chemical input data was found to be more accurate than byproduct andemission data. Using chemical input data, the study documented over 2.4 million pounds of toxicsuse reduction at 28 metal-intensive firms. To adjust reductions to the level of business activity,linear regression techniques were used to choose normalization measures which correlated withchemical use. The study found that for 20 chemicals, there was only a 5% difference betweennormalized and non-normalized measures of toxics use reduction.

1.0 INTRODUCTION

In 1989, the Office of Technical Assistance (OTA) started a three-year, EPA-funded technicalassistance project for metal intensive manufacturers in Central Massachusetts1. Through the CentralMassachusetts Pollution Prevention Project (Central Mass Project), OTA assisted firms in reducinghazardous chemical usage and waste generation through source reduction methods2.

Upon completion of the three year project, which lasted from July 1989 to July 1992, OTA

'EPA originally awarded a grant to the Massachusetts' Department of Environmental Management Officeof Safe Waste Management. Shortly thereafter, the Massachusetts' s Legislature created the Office of TechnicalAssistance (OTA). Subsequently, all grant responsibilities and personnel were transferred from the Departmentof Environmental Management to OTA.

2Another aspect of the Central Mass Project was OTA's participation in the Massachusetts Departmentof Environmental Protection Blackstone Project.

345

hired an independent contractor, the Environmental Careers Organization to evaluate the CentralMass Project. Three questions were asked in the evaluation: (1) What value did industry find inOTA services such as workshops and site visits? (2) What effect did OTA's coordination withregulatory agencies have on industry's perception of OTA as a non-regulatory entity? (3) How didfirms measure toxics use reduction? This article examines the third question only; it does notexamine OTA's efficacy or lack there of. Research results examining the first two questions willbe reported in future publications.

This article is comprised of six sections. The introduction is contained in Section 1. Section2 reviews results from a telephone survey where 110 metal-intensive manufacturers in Massachusettsanswered questions concerning toxics use reduction measurement practices. Section 3 presents theresults of toxics use reduction (TUR) efforts at 28 metal-intensive manufacturers. Section 4 describesthe statutory definition of toxics use reduction measurement under the Massachusetts Toxics UseReduction Act of 1989 and reviews the accuracy of such measurements at 13 companies. Section5 describes and evaluates a statistical technique used in the study for normalizing TUR to the levelof production. Lastly, Section 6 discuss the study's findings and recommends topics for futureresearch.

2.0 TELEPHONE SURVEY

Do relatively small metal-intensive manufacturers measure their toxics use reduction progress?The telephone survey examined the extent of TUR measurement and served as a screeningmechanism to identify firms for more comprehensive on-site interviews. Researchers developed a

list of 293 candidate firms, each of which were telephoned. Of the 293 firms called, 110 firmscompleted the telephone survey. One-hundred eighty three of the 293 firms either refused toparticipate, simply did not respond, or only partially completed the telephone survey. Partialinterviews were omitted from the data analysis.

Characteristics of Telephone Survey Firms

The 110 firms that completed the telephone survey were primarily small firms. Two thirdshad $10 million or less in annual sales and employed 50 or fewer employees. Telephoneinterviewers called each firm and requested to speak with the person:

'...in charge of, or having significant responsibility for pollutionprevention, including toxics use reduction and waste minimization.."

Seventy-nine percent of the persons interviewed were upper-level managers, plant managers,presidents, and owners. The 110 companies engaged in roughly 400 processes, comprised of 67.8%metal working operations, 13.7% metal finishing operations, and 9.9% ancillary operations such as

painting, assembly and soldering.

Findings

In the telephone interview, 83 of the 110 firms (75%) said they were doing toxics usereduction. Of those doing TUR, 36.1% said they documented and measured their TUR changes, 58%did not, and 6% did not know 3. Firms that reported under the Massachusetts Toxics Use Reduction

3The term "document and measure" was not defined by the interviewer during the telephone interview.Firms may have had different interpretations of the term's meaning.

346

Act (TURA) were significantly more likely to have documented and measured their TUR changes(see Figure 1 below). It is surprising that nearly 50% of TURA filers interviewed said they did notdocument their changes, since such firms are required to report annually on TUR progress.

D Yes

TURA FilersN=41

No

48.8%

43.9%

Don't Know

Non TURA FilersN=69

42.0%-KsÄp/^-l 8.8%

7.3%

|TV-18.£

^^^-39.2%

FIGURE 1:Are TUR Changes Documented and Measured?

During the telephone interview, researchers asked those firms that had not documented ormeasured their TUR progress, why they hadn't done so. One-half said they had no reason tomeasure their TUR changes. Twenty percent cited a lack of time and/or staff. Figure 2 reviewsthese and other reasons firms did not measure their toxics use reduction changes.

Percent of RespondentsN=52

50.0%

20.0%10.0%

LegendÜ No reason to measure Lack of time/staff

Q Only minimal effort¡g Lack of knowledgeH Lack of instruments

Too soon to measure

FIGURE 2:Why Are Changes Not Documented and Measured?

347

3.0 ON-SITE INTERVIEWS

The telephone interviews provide a general sense of TUR measurement practices amongMassachusetts metal-intensive manufacturers. To obtain more detailed information, on-site interviewswere performed. During the telephone survey, 83 of 110 companies claimed they had done TUR.Through a second, more in-depth interview, researchers found that 56 of the 83 firms had actuallydone TUR. The other 17 firms had actually done different types of waste management, such as

recycling or treatment. Of 56 the firms that had actually done TUR, 28 firms measured then-progress and were willing participate in an on-site interview. Box 1 summarizes the results of thisscreening process.

BOX 1: TUR Measurement Interviews

110 Firms telephone-surveyed in metal-intensive industry.83 Company representatives interviewed by telephone claimed to do TUR.56 Firms actually did "True TUR" and had measurement data to back it up.28 Qualified firms agreed to be interviewed.13 Firms of 28 interviewed report under TURA.

Results for 28 Firms

This section examines the toxics use reduction achievements of the 28 firms that completedon-site interviews. In compiling the results, researchers encountered poor materials accountingpractices, use of inappropriate normalization measures, and chemicals for which no materials balancehad been performed. The study sought to use byproduct and emissions indices, but found byproductsand emissions difficult to measure since they are derived from end-of-year materials balanceestimates. These difficulties threatened to push the study beyond its budget and time-line. To dealwith the paucity and inaccuracy of data, two simpler metrics based upon chemical inputs were

developed. Inputs were generally well documented in purchasing, chemical inventory, and processrecords and more readily available than by-product and emissions data.

The two simpler metrics, termed the unadjusted input reduction index (UnlRI) and thenormalized input reduction index (IRI), are defined in the Box 2 below (1). The unadjusted inputreduction index gives a percent decrease or increase in inputs over time. The normalized inputreduction index adjusts the decrease or increase in inputs by the level of production through the useof the unit-of-product. Each unit-of-product was chosen such that it correlated with chemical use.

BOX 2: Input Reduction Indices

Unadjusted Input Reduction Index (UnlRI) = 100 x [(A-B)/A]

where: A = input in base year B = input in current year

Normalized Input Reduction Index (IRI) = 100 x [(A-B)/A]where: A = input in base year B = input in current year

unit of product unit of product

348

Researchers measured reductions in inputs for 47 chemicals at 28 firms. These measurementsare summarized in Box 3 below. The 47 chemicals were comprised of solvents (57%), acids (30%),and metals and miscellaneous chemicals (13%). For 7 of the 47 chemicals, firms were in the processof making TUR changes. For these 7 chemicals, the amount of TUR was estimated. Box 3 detailsboth the mean and median unadjusted input reduction indices. Since a huge reduction by a singlecompany can have a significant influence on the mean reduction per chemical, the median providesa better measure of the typical chemical reduction. In cases where chemicals were eliminated fromuse, through techniques such as substitution, the unadjusted input reduction index (UnlRI) equals100%. From 28 companies, working on 47 chemicals, companies reduced chemical inputs by2,431,815 pounds through toxics use reduction techniques4.

BOX 3: Chemical Reductions

#ofChemicals

Mean %UnlRI

Median %UnlRI

Mean lbsreduced

Median lbsreduced

Total lbsreduced

Acids 14 46 54 114,731 62,325 1,606,237Solvents 27 64 77 11,261 7,280 304,045Mise 83 99 115,369 4,075 461,477Metals 25 25 30,028 30,028 60,056TOTAL 47 59 63 51,741 9,365 2,431,815

For 20 of the 47 chemicals, researchers identified a unit-of-product and therefore calculatethe normalized input reduction index (IRI). For these 20 chemicals, the net difference between theunadjusted pounds of chemical reduced and the normalized pounds of chemical reduced was only5%. Three of the 20 chemicals were eliminated from production and therefore had an IRI of 100%.

For 37 of the 47 chemicals discussed above, materials balances were performed to determinethe total amount of byproduct reduction by media. As Box 4 on the following page shows, thegreatest byproduct reductions were water-based, due to several firms that eliminated or substantiallyreduced acid usage. TUR resulted in substantial air byproduct reductions as well, due to the effortsof many firms to reduce or eliminate solvent usage. Consumers also benefitted from the changes.Over 48,000 pounds of toxics are no longer incorporated into product. The reductions listed in Box4 only partially demonstrate the benefits of TUR. Benefits such as improved worker health andsafety and reduced risk of spills during the transport of toxic raw materials and wastes areunaccounted for, but important.

4.0 TUR MEASUREMENT ACCURACY

The results presented in the previous section provide a snapshot of the TUR activity amongMassachusetts metal-intensive manufacturers. These results also gave information on the accuracyof reports filed under the Massachusetts Toxics Use Reduction Act of 1989 (TURA). This section

"Under TURA, toxics use reduction is achieved through the following techniques: input substitution,production reformulation, production unit redesign or modification, production unit modernization, improvedoperation and maintenance and integral recycling or extended use of toxics.

349

reviews TURA reporting requirements and presents the findings on the accuracy of 13 firms whichfile annual toxics use reduction reports.

BOX 4: Material Balance by Media5(pounds of chemical)

TotalDecreasein Input

Decreasein Air

Byproduct

Decreasein Water

Byproduct

Decrease inHazardous

WasteByproduct

Increase inByproduct-

as-Product6

Decrease inUnaccounted

forByproduct

Decreasein

ProductOutput

Decreasein OffsiteRecycle

TotalDecrease in

Output

2,377,235 639,606 1,525,290 149,328 (3,504) 2,893 48,138 15,484 2,377,235

Statutory Requirements

The Massachusetts Toxics Use Reduction Act (TURA), enacted in July, 1989, defines a

methodology for measuring toxics use reduction progress. TURA defines toxics use reduction as:

"in-plant changes in production processes or raw materials that reduce, avoid, or

eliminate the use of toxic or hazardous substances or generation of hazardousbyproducts per unit-of-product, so as to reduce risks to the health of workers,consumers, or the environment, without shifting risks to the health of workers,consumers, or parts of the environment.

TURA requires firms to report their annual use, byproducts, and emissions of listedsubstances in pounds7. TURA defines byproducts as "all nonproduct outputs of toxic or hazardoussubstances generated by a production unit, prior to handling, transfer, treatment or release." The Actdefines emissions as "a release of a toxic or hazardous substance to the environment or a transferof a toxic or hazardous substance in waste to an off-site location. TURA requires firms to reporton toxics use reduction performance using two metrics: the byproduct reduction index (BRI) andthe emissions reduction index (ERI). These indices represent the percentage change in byproductsand emissions per unit-of-product compared to a base year. Box 5 on the following page containsthe formulas for calculating byproduct or emission reduction indices.

TURA requires reporting firms to calculate BRI and ERI indices for any listed chemical usedin excess of a facility-wide threshold. BRI's and ERI's are based each production unit within whicha listed chemical is used. To calculate a BRI and ERI for each production unit, the amount ofbyproduct and emission are normalized using a measure of production known as the unit-of-product.Firms choose a unit-of-product they believe to be most suitable to their production process. Intheory, such production level measures should be correlated with each production unit's toxics use.

5Results for 26 companies, 37 chemicals. Results from 2 companies, 5 chemicals not included since firmswould not take the time to collect the data.

6TURA permits firms to classify certain byproduct management techniques as toxics use reduction. TURAdefines byproduct-as-product as, "use of byproduct without further treatment when the byproduct would haveotherwise been released, treated, or shipped off-site for recycling/reuse (1)."

7TURA defines listed substances as chemicals covered under Superfund Amendment and ReauthorizationAct, Title III and the Comprehensive Environmental Response and Compensation Liability Act.

350

BOX 5: Byproduct/Emission Index

Byproduct Reduction Index (BRI) = 100 x [(A-B)/A]

where: A = byproduct generated in base year B = byproduct generated in current yearunit of product unit of product

Emission Reduction Index (ERI) = 100 x [(A-B)/A]

where: A = emission generated in base yearunit of product

B = emission generated in current yearunit of product

A positive byproduct or emission reduction index indicates the percent reduction inbyproducts or emissions from a base year. A sample calculation of BRI and ERI indices aids inclarifying how firms measure toxics use reduction in Massachusetts. Box 6 below depicts an

example TUR measurement for a metal degreasing operation.

BOX 6: Example TUR Measurement

In 1988, Canton Circuits used 22,000 pounds per year of 1,1,1-trichloroethane in a vapor degreaser toremove oil from the 16,000 metal circuit boxes it manufactured. The 1,1,1-trichloroethane was not usedelsewhere in Canton's operations. Fugitive emissions from the degreaser were vented outside of theplant. Periodically, Canton cleaned out sludge from the degreaser and shipped it off-site to be recycled.In 1989, after installing a tank cover and an additional set of cooling coils to reduce fugitive emissions,Canton used 15,000 pounds of 1,1,1-trichloroethane to degrease 18,000 circuit boxes. Under TURA,Canton would measure progress for the degreaser production unit as follows:

Production Unit: The production unit must include all chemical use, storage, and transfer operations.In this case, the degreaser and all 1,1,1-trichloroethane storage, waste management,and operations areas comprise the production unit.

Unit-of-Product: Canton may choose their own unit-of-product, such as the number of circuit boxesor square feet of metal. In this case, Canton chose the number of circuit boxes.

Byproducts: Since virtually no 1,1,1 is incorporated into the product nor reacted during thedegreasing process, the amount of 1,1,1 byproduct equals the amount of 1,1,1 use.

Emissions: Since no 1,1,1-trichloroethane byproducts are treated or destroyed on-site, theamount of 1,1,1 emissions equals the amount of 1,1,1 use.

Byproduct and emission reduction indices are calculated below (calculations in thousands):Year Byproduct (lbs) Emission flbs) Circuit Boxes1988 22,000 22,000 16,0001989 15,000 15,000 18,000

Byproduct Reduction Index: 100 x {[22/16]-

[15/18]} / [22/16] = 40%Emissions Reduction Index: 100 x {[22/16]

-

[15/18]} / [22/16] = 40%

351

Findings: Byproduct and Emission Reduction Index Accuracy

TURA reports for thirteen firms (all of the TURA filers among the 28 firms receiving on-siteinterviews) were audited for accuracy. In many cases, researchers reviewed multiple chemicals ata single facility, examining a total of 28 chemicals at 13 companies. In each personal interview,materials balances, unit-of-product choice, and reduction indices were examined using the companiesown documentation. For the purpose of the study, a firm's BRI's and ERI's were consideredaccurate if they came within + 10% of the researchers reconstructed estimate. As Box 7 belowshows, roughly half of the BRI and ERI indices were found to be accurate.

BOX 7: BRI/ERI Accuracy

For 28 TURA Chemicals Analyzed at 13 Firms

TURA Index Calculations BRI ERI TOTAL

Accurate 13 13 26Inaccurate 11 9 20Not Calculated 4a 4b 8Other 0 2C 2

a Because a BRI compares current year by-product generation with abase year, the BRI has no meaning in the first year of reporting.b Some chemicals, such as acids that are neutralized in waste watertreatment, have zero emissions and consequently no ERI.c Two firms chose not to submit the ERI for their reported chemical.

The reasons for inaccurate indices in order of frequency were: (1) first time filers(inexperience), (2) estimated or inaccurate data, (3) math errors, (4) poor unit-of-product choice, (5)poor segmentation of production units, and (6) unaccounted for chemical reactions. First time filers,estimated data, calculation errors, and unaccounted for chemical reactions were primarily due to thenewness of TURA regulations and the use of poor materials accounting techniques. Poor unit-of-product choice, which is related to poor segmentation of production units, stems from the lack of aformal technique to chose a good unit-of-product. One technique for improving unit-of-productchoice is discussed in the next section.

5.0 EVALUATING THE UNIT-OF-PRODUCT

Four of the eleven inaccurate BRI's documented in this study (see Box 7 above) were dueto poor unit-of-product choice. The unit-of-product adjusts a production unit's yearly byproduct andemissions generation to that year's level of production, making year-to-year comparisons possible.When a firm uses an uncorrelated unit-of-product, it may report BRI and ERI increases or decreasesthat bear little relation to actual changes in chemical use efficiency.

The Massachusetts TUR program gives firms general guidance for selecting a unit-of-product.The state recommends that the unit-of-product "must increase/decrease in a consistent way aschemical usage increases or decreases" (2). This guidance is similar to guidance found in theliterature (3)(4)(5) and that given for pollution prevention reporting requirements in SARA 313 (6).For the purpose of this study, a formal technique for choosing a unit-of-product was developed.

352

Finding a Suitable Unit-of-product

To evaluate the choice for unit-of-product, researchers used linear regression to test forcorrelation between a firms chemical usage and their unit-of-product. A strong correlation indicatesa good unit-of-product choice. No correlation indicates a poor choice. The five-step methodologyfor conducting such a test is outlined below:

1.2.

3.

4.

5.

Draw a flow diagram, tracing the chemical's path through the process.Gather time-consistent data (i.e. every month) on

• chemical inputs to the process• independent variables that relate to the process (i.e amp hours, sales, square feet plated)

Gather this data over time periods where:• the process performed consistently• there were no changes to the process (including TUR changes)

Test the relationship between chemical inputs and independent variables• using an F-test or t-test, determine whether correlation exists• analyze the data for anomalies such as influential observations

Repeat the analysis (steps 2-4) for time periods after a TUR change to ensure the correlation stillexists.

This test may be used when time-consistent unit-of-product and chemical use data areavailable. An example helps to clarify the usefulness of this technique. Data was collected fromone firm on chemical usage versus sales revenue over 19 quarters. The regression, performedat the facility site using a Hewlett Packard calculator, confirms that acid use and sales are highlycorrelated (p < 0.0005) with an R2 value of 0.8609. The raw data, the mean regression line and95% confidence intervals for the mean are presented in Figure 3 below.

MeanRegressionLine

Facility Data Upper 95%ConfidenceInterval

"°- Lower 95*ConfidenceInterval

700 900 1,100 1,300 1,500 1,700 1,900 2,100

Sales ($Thousands)

FIGURE 3: Regression Test for CorrelationSales vs Acid Use

353

The graph indicates that one data point may have extreme influence on the analysis. Thisis confirmed through statistical analysis of the Cook's distance for the observation with salesequal to $1.78 million. Although we did not do so, the firm could review their records for thistime period to determine the cause of this influential observation.

Even with the influential observation, the analysis shows that sales is a good indicator ofacid use. This regression test was performed on data prior to a TUR change to the manufacturingprocess. As soon as sufficient time periods have passed since the change, the test should bereapplied to insure the TUR change did not alter the correlation between acid use and sales. Thedata were also tested for autocorrelation (7), to test whether the time-dependent nature of the dataoverestimated the data's fit to the simple regression model. In the presence of autocorrelation,a regression estimate tends to overestimate the correlation between independent and dependentvariables. In this case, the Durbin-Watson d-statistic test (8) showed the data were notautocorrelated, confirming the initial findings that acid use and sales are significantly correlated.

The strength of the correlation between chemical use and the unit-of-product is veryimportant. In this example, the relationship between acid use and sales is so uncertain that onecould almost eliminate acid use or almost double it and still be within pounds of acid usedpredicted by the regression model (the 95% confidence interval for acid use at $1,019,069 annualsales (the mean sales value) is 2,718 + 2,572 pounds, or 145 to 5,291 pounds). Clearly, onewould prefer a more strongly correlated unit-of-product, achieved by choosing a unit-of-productother than annual sales such as pounds of metal etched. If such data were available, theregression results of sales and pounds etched could be compared and the most significantlycorrelated one chosen as the unit-of-product. But even without this further analysis, performinga regression to choose a unit-of-product is preferable to simply "guestimating" the best unit-of-product since doing so may generate BRI's and ERI's that bear no relation to production processchanges.

6.0 DISCUSSION AND FUTURE RESEARCH

The metal intensive industry in Massachusetts appears to be preventing a sizeable amountof pollution. This study identified over 2.4 million pounds of toxics eliminated by 28 firms inthe time period ranging from 1988 to 1992.

In performing the study, researchers found TUR measurement complex and time-consuming. Performing a materials balance on a single chemical took 8-40 hours to complete;collecting the information to run a quick correlation check on unit-of-product took 4-10 hoursper chemical. Even with this time investment, inconsistent quality of materials accounting datamade accurate byproduct, emissions, and unit-of-product estimates difficult. For example, thechemical composition of waste water discharges are monitored periodically; these periodicsnapshots of the waste stream may not capture a production process' true byproducts andemissions. Finding normalization measures that correlate with a production units chemical useis similarly difficult. Normalization measures depend on the ability to collect accurate, time-dependent data on inputs and production variables.

This research showed that one method for simplifying TUR measurement is to examineinput reductions as opposed to byproducts and emissions. With a few notable exceptions suchas cases where input concentrations are highly variable (e.g. oil refining), input reductionappraisals will measure TUR progress better than byproduct and emissions measures. Inputs areeasier to quantify; most firms already track chemical inputs to production processes. And unlikebyproducts, which are typically calculated at the end of the year as a derivative in the materialbalance calculation, inputs may be monitored on a weekly or monthly basis to provide feedbackon process characteristics such as material use efficiency. Such feedback may induce firms thatdo not measure their toxics use to do so regularly, and those that do measure, to measure more

accurately.

354

This study also investigated the utility of normalizing reductions for the level of businessactivity. For input reductions in 20 chemicals, normalizing for production levels had no

appreciable effect (less than 5%) on the total amount of toxics eliminated. These findingscollaborate with the findings of other researchers (4). Normalizing at the firm level is stillessential to determining a firm's year-to-year progress. But for pollution preventionmeasurements at the industry, state, or national level, normalizing for production levels may notbe necessary.

Among the many TUR measurement issues warranting future research, two issues standout from this study. First, the study was unable to determine which steps in the TURmeasurement process contributed most to TUR measurement error. For example, was inaccuratebyproduct accounting more likely to contribute to larger TUR measurement errors than inaccurateunit-of-product accounting? Such a sensitivity analysis could assist firms in efforts to improveTUR measurement accuracy.

Second, the study did not perform a rigorous evaluation of the use of regressiontechniques in determining correlation between a production unit's unit-of-product and its toxicsuse. Regression analysis is a tool firms can use to find normalization measures that correlate tochemical use. These regression techniques range from simple linear regression to more complexmultiple regression and autocorrelation analysis. In many cases, simple linear regression maywork very well in identifying a strongly correlated unit-of-product. But in some cases, linearregression will be unable to identify a single factor that is an accurate predictor of chemical use.In these cases, multiple regression may aid in finding predictors with greater accuracy. Inaddition, autocorrelation analysis may be needed since both simple and multiple regressionoverestimate the correlation between toxics use and the unit-of-product when the data underanalysis is autocorrelated. But both multiple regression and autocorrelation analysis requiresoftware capable of statistical analysis. Teaching such analysis techniques to industry would bea complex task. Although its unclear whether these techniques could be successfully used bysmall and medium sized firms, larger firms may find them more useful than simply "guestimat-ing" normalization factors that may bear no relationship to chemical use or waste generation.

The extensive time requirements, lack of quality data, and complexity in materialsaccounting estimates and unit-of-product choices makes measuring TUR difficult. In this study,roughly half of the BRI's and ERI's reviewed were inaccurate. Despite this difficulty, few ofthe mostly small- and medium-sized firms interviewed in this study devoted personnel solely toenvironmental, health and safety responsibilities. However, collecting accurate TUR data is keyto assessing an individual firm's progress and the progress of the Commonwealth towards its1997 state-wide goal to reduce toxic-byproducts 50% from a 1987 baseline. We believe it isincumbent on the Massachusetts TURA program to increase its educational-outreach andassistance programs to aid manufacturers in accurately measuring their TUR performance.Educational programs could have a substantial impact on the three most frequent sources of BRIand ERI errors found in this study: inexperience, use of estimated data, and calculation errors.

REFERENCES

1. Harriman ED, Markarian J, Naparstek JS, Stolecki JW. Measuring Progress in Toxics UseReduction. Capstone Project Report. Hazardous Materials Management Program. Departmentof Civil Engineering. Tufts University. Prepared for the Commonwealth of MassachusettsDepartment of Environmental Protection. 1991.

2. Massachusetts Department of Environmental Protection. DEP TURA Reporting Package.1991.

355

3. Serious Reduction of Hazardous Waste: For Pollution Prevention and Industrial Efficiency.Office of Technology Assessment. 1986. OTA-ITE-317.

4. National Research Council. Tracking Toxic Substances at Industrial Facilities. NationalAcademy Press. 1990.

5. Baker RD, Dunford RW, and Warren JL. Alternatives for Measuring Hazardous WasteReduction. Center for Economics Research. Research Triangle Institute. Research TrianglePark, NC. For The Hazardous Waste Research and Information Center. Illinois Department ofEnergy and Natural Resources. 1991.

6. EPA Office of Pollution Prevention and Toxics. 1991 Toxic Chemical Release InventoryReporting Form R and Instructions. 700-K-92-002. 1992.

7. Ostrom CW. Time Series Analysis: Regression Techniques. Sage University Paper serieson Quantitative Applications in the Social Sciences. Series no. 07-009. Sage Publications.Beverly Hills and London. 1976.

8. Durbin J and Watson GS. Testing for serial correlation in least-squares regression.Biometrika 37:409-428 and 38:159-178. 1950 and 1951.

Address reprint requests to:

Timothy J. GreinerMassachusettsOfficeofTechnical Assistance100 Cambridge StreetBoston, MA 02202

356