Grantee Research Project Results

2000 Progress Report: Environmental Policy and Endogenous Technical Change: A Theoretical & Empirical Analysis

EPA Grant Number: R826610
Title: Environmental Policy and Endogenous Technical Change: A Theoretical & Empirical Analysis
Investigators: Opaluch, James J. , Grigalunas, Thomas A. , Jin, Di
Institution: University of Rhode Island , Woods Hole Oceanographic Institution
EPA Project Officer: Chung, Serena
Project Period: October 1, 1998 through September 30, 2001
Project Period Covered by this Report: October 1, 1999 through September 30, 2000
Project Amount: $325,000
RFA: Decision-Making and Valuation for Environmental Policy (1998) RFA Text |  Recipients Lists
Research Category: Environmental Justice

Objective:

The objectives of this project are to:

• Develop a deeper understanding of the relationship between technical change and alternative environmental policies that accounts for environmental inputs and depletion of natural capital stocks.

• Use a case study to measure historic rates of technical change, accounting for environmental inputs.

• Compare ex ante, engineering estimates of the costs of complying with environmental regulations to actual, ex post performance that includes innovative means of achieving standards, like process change.

• Estimate benefits of environmental policies that provide increased flexibility and that encourage innovation.

• Simulate the long run, relationship between productivity change and environmental protection.
  

Progress Summary:

The majority of the accomplishments fall under four general tasks: (1) data collection and management, (2) creation of indexes for new technological discoveries, (3) measurement of technical change, and (4) the theoretical model of technical change with complementary technologies. Progress on each of these tasks is briefly discussed below.

Data Collection and Management. Over the past year, we have largely completed our task of data collection, and have expended considerable efforts cleaning, merging, and otherwise preparing the data for analysis. We have collected data for production of oil and natural gas at the well level for the Gulf of Mexico from 1948 through 1998. This is a truly huge data set. We aggregated the data to the field level, as we believe that production efficiency is most appropriately measured in terms of production from each field. Considerable effort was required to identify, collect, clean, and merge this production data with other data, including estimates of total resources, changes in resource estimates over time due to new discoveries, drilling distance for exploratory wells, drilling distance for development wells, etc. We also have linked various geophysical measures for fields, such as water depth, porosity, etc.

Creation of Indexes for New Technological Discoveries. One important goal of our project is to separate out technical change associated with specific technical discoveries with the more routine "learning by doing" (e.g., Arrow, 1962). This also relates to the two primary models of endogenous technical change?Romer's model, which looks at the creation of new technologies, and Aghion's model, which looks at incremental improvements in the quality of existing technologies.

We use an index of technical discoveries developed by Moss (1993), and used by Cuddington and Moss (2000) to measure technological inputs. Thus, the coefficient on this "input" is meant to capture the effect of discrete new technological discoveries, while the residual measure of technical change can be attributed to an accumulation of more modest improvements in production technology, attributable to "learning by doing," for example. We have updated Moss' index through 1998 using the same methodology. We then apply this technology index as a measure of technological change attributable to new technological discoveries.

Measurement of Technical Change. With the data largely in place, we have begun to carry out analyses of efficiency with these data. We completed a preliminary application of Data Envelopment Analysis (e.g., Fare, Grosskopf, et al., 1994) to construct efficiency measures. Simultaneously, we focused our efforts on two alternative methods?stochastic efficiency frontiers (e.g., Aigner, Lovell, et al., 1977) and more usual regression techniques, such as ordinary least squares and two stage least squares.

The Stochastic Efficiency Frontier (SEF) approach employs econometric methods to estimate efficient production frontiers using an asymmetric error term, rather than the usual econometric approach of estimating parameters that represent "average" (or representative) levels. Thus, the goal is to estimate parameters of the efficient frontier, or the maximum levels of production with a given set of inputs. The methods can be applied to estimating primal functions, such as the production function, or dual functions, such as the profit or cost function.

The SEF model is based on decomposing the random error term, e, into two parts?v and u. The component v is a normal error component that captures the usual sources of random variation (e.g., measurement error). This symmetric component of the error term also measures shifts that might be associated with random fluctuations in the efficient frontier. For example, in agriculture, random fluctuations in weather imply that the maximum amount of yield that comes from a given set of inputs will vary randomly. During years of bad weather, the maximum yield will be lower than it will be during years of average weather or years of ideal growing weather. The second component of the error term, u, is an asymmetric component that is a one-sided error term (e.g., a truncated normal or an exponential distribution), which measure deviations from the production frontier that arise due to errors in maximizing. So, a firm might use too high (or too low) a labor-capital ratio, or a firm might drill too many wells (or not enough wells) in its search for oil on a specific structure. A divergence from the optimal in either direction leads to a reduction relative to the efficiency frontier, hence the one-directional error component.

Maximum likelihood methods are used to estimate the coefficients of the model, including the variances of the two error components. This provides two pieces of information. First, the estimated coefficients provide an estimate of the efficiency frontier that can be used to construct the Malmquist indexes of productivity change. Second, the relative sizes of the two error variances indicate the extent to which inefficiency (distance from the efficient frontier) is an important factor in the data. The greater the variance of the asymmetric component, u, the greater is the estimated inefficiency of some observations relative to others. As the error variance on u goes to zero, inefficiency goes to zero, and all firms are on the efficient frontier. In this case, all random deviations are shifts in the frontier or measurement error.

We have completed our initial estimation of the Stochastic Efficiency Frontiers and have some preliminary results. However, additional analyses must be conducted before we can report any results or conclusions. The next steps in this process are to: (1) more fully specify the models, and (2) construct estimates of efficiency and technical change.

In the mean time, we are refining our models to complete our DEA work. We anticipate that our analysis of Stochastic Efficiency Frontiers and Data Envelopment Analysis will be largely completed over the next few months. This analysis will provide our Malmquist indexes of technical change and efficiency change, which are major components of the empirical work of our project. We also will use our technology indexes to distinguish between specific technical discoveries from more routine "learning by doing."

We also came to the realization that oil and gas prices might be important explanatory variables for efficiency. That is, in periods when prices are high, even low productivity fields may be profitable to exploit; when prices are low, the less productive fields would more likely be shut down. Thus, prices are likely to be important explanatory variables that are negatively related to the average and minimum efficiency of observed fields. We collected oil and gas prices over the time horizon, and integrated prices into our database.

Theoretical Model of Technical Change. We also have made progress in the theoretical model of technical change, focusing on complementary technologies, as discussed in previous progress reports. Production occurs at two levels. First, there are intermediate outputs, which are products of firms that make technical discoveries and hold the associated patent for that level of technology. Second, these intermediate outputs serve as inputs in the production of the final output, which is consumed.

The theoretical modeling of this process takes place in five steps:

1. Specify the technology for final output

2. Determine demand for intermediate outputs as a function of the current level of technology and output price.

3. Determine optimal pricing strategy for intermediate outputs by monopoly patent holder of technology for those intermediate outputs.

4. Determine expected profit function for patents holders.

5. Determine optimal investment in R&D for discovery of new technologies.

We have pursued two avenues for final output technology: a model of Hicks neutral change and a model of input-biased technical change. Both models are based on a Cobb-Douglas production technologies of the form:

where Y is the final output, Xij(t) represents the level of the intermediate outputs, and is the associated coefficient in the production function for final outputs. The model incorporates both the Romer approach, where technical change is embodied in discovery of new technologies, all of which operate simultaneously, and the Agion and Howitt approach, where productivity change represents incremental improvements in existing technologies, and the improvements replace the older methods (creative destruction). Within the context of our model, increases in n(t) represents discovery of new technologies, and increases in qij(t) represents incremental improvements in existing technologies.

Hicks neutral technical change is modeled by assuming the are constant, and Xij(t) = q(t)xij(t), where X is the "effective" input, including a technical change component, x is the actual physical level of the input, and q(t) represents the level of technology. Input biased technical change is modeled by assuming Xij(t) =xij(t), so that the effective input equals the physical input, but aij(t) = q(t), so that the exponent changes with new discoveries.

Both models demonstrate externalities in productivity, where advances in productivity for one input increase the productivity of other inputs. However, the theoretical results show that the monopolist owner of new technologies is not able to capture rents under Hicks neutral technical change, while at least part of the rent can be captured with input biased technical change, hence providing incentives for R&D activities.

Future Activities:

The next steps in the research include completing each of the above tasks, and developing this work into a formal simulation model for the industry. Now that we have the OnFront software, we will complete the DEA analysis, followed by the Stochastic Efficiency Frontier analysis. These will be used to specify the production frontier and to characterize inefficiency within the industry. Next, we will measure technical and efficiency change in the offshore oil industry over the study period (1948-1998). Once this is complete, we will provide separate estimates of technical change for discrete new discoveries, versus refinements and "learning by doing." Finally, this will be used to specify our industry simulation model and to explore how alternative policies would affect technical change within the industry.

References:

Aigner DJ, Lovell CAK, Schmidt P. Formulation and estimation of stochastic frontier production function models. Journal of Econometrics 1977;6(1):21-37.

Arrow KJ. The economic implications of learning by doing. Review of Economic Studies 1962;29(80):155-173.

Cuddington JT, Moss DL. Technological change, depletion, and the U.S. petroleum industry. American Economic Review 2001;91(4):1135-1148.

Fare R, Grosskopf S, Norris M, Zhang Z. Productivity growth, technical progress, and efficiency change in industrialized countries. American Economic Review 1994;84(1):66-83.

Moss DL. Measuring technical change in the petroleum industry: a new approach to assessing its effect on exploration and development. Presented at the National Economic Research Associations, Working Paper #20, 1993.

Journal Articles:

No journal articles submitted with this report: View all 8 publications for this project

Supplemental Keywords:

petroleum, productivity, innovative technology, economics, technical change, offshore oil., RFA, Scientific Discipline, Economic, Social, & Behavioral Science Research Program, Economics & Decision Making, Engineering, Economics, decision-making, benefits assessment, theoretical analysis, economic benefits, compliance costs, cost benefit, technical innovation, ecosystem valuation, SIC 1300, innovative pollution control, economic incentives, endogenous technical change, empirical analysis, offshore oil, decision making

Relevant Websites:

http://www.uri.edu/cels/enre/

Progress and Final Reports:

Original Abstract

1999 Progress Report

Final Report