Projects

Price Volatility

Price Volatility, Inventory and Industry Consolidation

Full Description

Research Strategy and Methodology

In our existing project, we have collected a vast amount of data on the pulp and paper industry from various sources. Moreover, we have carefully examined and processed the new data from Forest Products Laboratory (FPL). The FPL data provide detailed information at mill level for every mill in the U.S. from 1970 to 2000. With the data, we will be able to construct rich econometric models to conduct a wide range of analyses.

In respond to Sloan Foundation’s preference for observation-based methodology, our approach will be empirical, i.e., to use first hand data from the industry to conduct econometric analyses. Moreover, in addition to the data sets we have collected, we intend to visit 3-4 companies in order to get industry insights on inventory management, price volatility, consolidation, vertical integration, and other non-price topics.

In order to bridge our project with the industry, we will ask industry experts Mr. Robert Guide and Mr. Gary Helik to serve as industry advisor to this project. Mr. Guide is a former director of containerboard sector at the AF&PA, and Mr. Helik is a director of the North America pulp and paper division at the Traditional Financial Services. We will ask more people from the industry to serve our industry advisory panel as our project moves forward.

Our approach to each research questions specified in the above section will be discussed as follows.

I. Modeling Price Volatility

In order to study the dynamics of price volatility, we will utilize several measures of volatility, including absolute or squared log price differentials, the range defined as the difference between the highest and lowest log price during a discrete sampling interval, a moving average of lagged log price changes, and the implied volatility estimates obtained from time series models. We will utilize weekly and monthly data on pulp and paper prices (or their PPI).

By examining the time series of log linerboard price changes, we find that periods of large price changes are followed by periods of relatively stable prices. This indicates that markets are sometimes tranquil and sometimes turbulent. This property of prices is referred to as volatility clustering. Given such time-varying volatility dynamics, we aim to model volatility by utilizing Generalized Autoregressive Conditional Heteroscedasticity (GARCH) family of models. To our best knowledge, no study has explored the volatility dynamics by using GARCH models in pulp and paper industry.

Econometric modeling of the time-varying volatility occurred relatively recently in 1980s. The Autoregressive Conditional Heteroscedasticity (ARCH) models, introduced first by Engle (1982) and modified by Bollerslev (1986) and labeled as Generalized ARCH (GARCH) model and their extensions, have become popular both among practitioners and researchers. GARCH models are able to describe certain properties of economic time series, such as volatility clustering and excess kurtosis. GARCH family of models allows us to model persistence and serial correlation in volatility dynamics parsimoniously. The GARCH processes approximate volatility dynamics in moving averages of lagged squared errors and lagged autoregressions of variances. Today’s conditional variance functions are linked to the lagged conditional variances and lagged squared errors in a linear fashion. They are flexible enough to model joint dynamics of conditional mean, say prices and the conditional variances in a market. The GARCH family of models has been used in exchange rate, stock and other commodity markets extensively both to model volatility dynamics and to forecast the volatility and future prices. A survey of this extant literature can be found in Diebold and Lopez (1995), Bollerslev, Chou, and Kroner (1992), and Diebold (2004).

In addition to use of GARCH family of models to model volatility dynamics, we will also model jointly the pulp and paper prices and their volatilities by estimating ARIMA-GARCH models for prices and volatilities. Since these models will incorporate any dynamics and information in the conditional moments of prices, we can generate better forecasting models for the future prices.

Moreover, by comparing the volatility dynamics in pulp and paper price to that of other commodities that have well functioning futures markets, we can obtain useful information in terms of hedging against risk and on the need and development of futures market in the paper industry.

II. Inventory and Short Term Price Changes

Academic research has noticed the relationship between inventory and price as early as 1960s. Hay (1970) studied price, inventory, and production, and used the data from U.S. lumber and paper industry to estimate their relationship. This study is mainly on the long term relationship between price and inventory, not the short term causality relationship we will be investigating. Since then, there have been no major study on price/inventory issue in the pulp and paper industry, although many studies on other industries. Recent studies on this issue utilize micro level data to estimate the probability of price changes, such as Carlson and Dunkelberg (1989) and McIntosh et al (1993). Our approach will follow these studies while using industry level data.

In particular, based on monthly data from the containerboard sector (linerboard in particular), we noticed that inventory changes generally occur before price changes, for example, inventory goes up first, and then price start to fall (see the graph below). Based on industry analysts, a good indicator of price change is the ratio of inventory to shipments. In fact, a formal test based on Granger’s causality test shows that inventory does causes price changes and thus can be treated as a leading indicator of price change. Therefore, it is desirable to build a model of price change based on inventory to predict the probability of price change in a short term. For example, we can use the inventory at the beginning of the month to predict the probability of price increase or decrease at the end of the month. The predicted probability can be useful for production planning, downtime scheduling, and other short term operation needs.

In order to estimate the probability of price change in response to inventory changes, we will apply probit models. We will use monthly data on containerboard price and inventory (including inventory at mills and at box plants) for estimation. Two types of probit models will be estimated, one is for price increase and the other is for price decrease. The results from the two models will help to investigate whether the effect of inventory on price increase and decrease is symmetric.

Moreover, probit models with the same specification will be estimated for output changes. By comparing the effects of inventory on price change and on output change, we can get some important insight on how the industry responds to inventory changes. If the inventory effect on price decrease is stronger than that on output, it demonstrates that the industry would rather take a hit on price instead of actively planning on production (such as taking down time) to maintain the inventory as a more efficient level. Hence, a more efficient production planning mechanism is beneficial for the producers.

Depending on the results, there are a number of ways to extend the single equation probit models to more sophisticated ones. More specifically, based on the results, we can estimate the price model and the output model jointly in an equation system, if it turns out that price and output are closely related in such a short time frame after controlling for inventory. Another possible extension is to move beyond the binary choice probit model to multinomial logit model, which can investigate three price scenarios, increase, decrease, and no change, in the same model.

III. The Effect of Industry Consolidation on Price, Price Volatility, and Costs

There is a vast literature about the effect of market concentration on price and price-cost margin. This literature is reviewed by, among others, Weiss (1989), Schmalensee (1989), and Werden (1991). However, the existing studies mostly cover airline, banking, advertising, gasoline, and grocery retailing industries; or they are based on cross industry data. One study by Koller II and Weiss (1989) investigates price-concentration relationships in the cement industry, which is relatively more closely related to the pulp and paper industry. Additionally, studies on market concentration and price volatility are mostly focused on electricity market, for example, Robinson and Baniak (2002) and Newbery (1998). There is no major study on these issues for the pulp and paper industry.

In this research dimension, we will estimate the effect of consolidation on price, profit margin and price volatility. We will use annual data from 1970 to 1997 for all three sectors: pulp, paper, and paperboard. The data will come from two sources, one is from Census and the other is from the FPL. Most existing studies (in other industries) used the Census data. However, since the census data are collected every five years, these studies needed to estimate annual data using interpolation. The FPL data are annual data at mill level, thus we can generate true values for all year for the model. In this sense, the FPL data we have should be more accurate.

Moreover, since we have data on all three sectors in the pulp and paper industry, we can use the fixed effects model for panel data in the estimation. In general, panel data are more powerful in dealing with the problem caused by omitted variables in regression analyses, compared to either cross-sectional data or time-series data. Using the panel data, we will estimate a price model, a profit margin model, and a price volatility model to investigate the effect of market concentration resulting from industry consolidations. The profit margin is calculated from value added and material costs, and the data are available from the Census. One measure of price volatility is annualized price variance. There are a number of measures of market concentration, and we will primarily use the concentration measure based on the top four producers. These models will be controlled for capital intensity, imports intensity and business cycles in order to identify the effect of market concentration. All models will be estimated using both the Census data and the FPL data to compare the difference.

Based on the results from all these models, we can get a lot of useful information. More specifically, if consolidation does not show a significant effect on price, but it shows a positive and significant effect on profit margin, this result indicates that consolidation helps lowering the cost and thus improves production efficiency. If the results show that consolidation has a significant effect in reducing price volatility, this is a desirable outcome as well. In any case, as the outcome of industry consolidation is still largely unknown, the findings will be of great interest to both researchers and industry players.

IV. The Effect of Price Volatility on Vertical Integrations

The final research question in the proposed project is to investigate the effect of price volatility on vertical integration. We will specifically look into the decision of a paper or a paperboard company to integrate backwards with a pulp mill. Because it is costly for upstream firms (paper mills) to negotiate with input suppliers downstream (pulp), paper mills have a profit incentive to vertically integrate with downstream firms when integration is expected to reduce transactions costs, that is, the vertically integrated firm reaps transactions economies. Further, Williamson (1975) argues that transactions costs between non-integrated firms are greater the more concentrated the market (reflecting a less competitive environment), the more closely are capital assets to the productive activity (e.g. a desktop computer can be used in many productive activities whereas a Fourdriner machine can only be used in papermaking), and the more uncertain the environment (since this entails the negotiation of more contingencies in a contractual agreement). Building upon the work of Ohanian (1993, 1994) and Melendez (2002), and after controlling for concentration-related and asset-related factors that affect transaction costs, we will investigate whether price uncertainty, as measured by its volatility, has any effect on the decision for a paper mill to vertically integrate with a pulp mill.

Primary data for this analysis will be a panel of pulp and paper mills for the period 1975 – 2000, provided by the Forest Products Laboratory in Madison, WI. During the past 20 years or so, the proportion of integrated paper mills in the US has remained relatively stable, ranging from 39% to 47% during the period, and equal to 42% in 1995. Also during this period, integrated mills represented about 80% of total paper capacity in the US and 93% of pulp capacity. However, between 1975 – 1980 and 1985 – 1990, there was a net increase of 24 and a net decrease of 9 integrated firms, respectively. More recently, the industry saw a further net decrease of 7 integrated firms in the 1990 – 1995 period. Thus, although there is relative stability in the proportion of vertically integrated firms, in total, these data indicate that there has been a significant amount of integrating activity in the industry.

This part of the study will contribute to our understanding of the role of price and price dynamics in firms' decisions to vertically integrate and provide important insights on how price uncertainty is expected to change industry structure. With these data, we will estimate a series of econometric models in order to understand the varied effects that price volatility has on a firm's decision to vertically integrate. First, we will use a probit model to investigate whether a company is vertically integrated and to estimate the effect that changes in price volatility have on the probability of vertical integration and whether these effects differ by firm size or by market concentration. Second, we will explore whether price uncertainty has symmetric effects, that is, whether the effect of upward price volatility on vertical integration is similar to downward price volatility. Third, we will explore whether the effects of price volatility have been stable over time or whether firms have greater incentives to integrate during particular phases of the business cycle. And fourth, we will use a Tobit model to investigate the degree of vertical integration measured by the ratio of paper or paperboard capacity over pulp capacity in a company.

Desired Outputs and Contributions to Theory and Research

In the proposed project, we will study price volatility and some important relationships including price-inventory, price-consolidation, and price volatility-vertical integration. These issues are directly linked to economic theory and have great academic value. In the existing research literature, most of these issues have not been investigated in the context of pulp and paper industry. Therefore, we will contribute to research and theory by examining these questions in this specific industry.

Moreover, the study will require sophisticated (but appropriated) and even some new techniques (for example, the price volatility analysis), or will require to view the same question from a different angle (for example, the price-inventory relation, and the relationship between price volatility and consolidation/vertical integration). Therefore, we will contribute to the academic literature not only by examining a particular industry but also by novel ideas and methodologies applied in the investigations.

The desired outputs to research include four potential research papers listed below. These papers will be the core for several Master theses, and will be submitted to peer-reviewed academic journals.

A research paper on the dynamics of price volatility for pulp price and paper price.

A research paper about the short-term effect of inventory on price change and production change.

A research paper on the effect of industry consolidation on price, profit margin and price volatility.

A research paper about the effect of price volatility as a transaction cost on vertical integrations.

Desired Outputs and Contributions to Industry/Other Groups

As discussed above, every research question in the proposed project has important implications for the industry. In particular, the study of price volatility will improve price forecasting and provide useful information on price risks. Thus, it can aid the industry’s need in transforming their traditional business models by, for example, incorporating modern forecasting tools in decision making and developing means to hedge against risks. The price-inventory model can generate estimate of the probability of a price change at a particular point of time, and hence should be helpful in inventory management and downtime planning. The evaluation of the effect of industry consolidation and the study of vertical integration will provide some very important information for industry players regarding strategic decisions on mergers and acquisitions. It should be extremely valuable for the industry to know whether consolidation has solved or will solve the problems faced by the industry, and whether an increased degree of integration will be more cost and logistic efficient.

Based on our findings, we will produce a series of shorter papers (e.g. executive summaries) and presentations in order communicate our results to industry. We will pursue this goal through a variety of formats, such as write papers for trade journals and present our findings at industry conferences.

The desired outputs to the industry will include the following reports or executive summaries:

A report on the dynamics of the volatility of pulp and paper price; and how the volatility dynamics can be used to improve price forecasting; and what the industry can do to hedge the risks.

A report on predicting short-term price changes using inventory as a leading indicator; and on how the information can be used to help planning production, such as taking down time.

A report on the effect of industry consolidation on price, profit margin, and price volatility; and whether consolidation has solved the problems facing the industry.

A report on the factors affecting vertical integration, especially the effect of price volatility on the probability and degree of vertical integration; and whether an increased degree of integration will be more cost and logistic efficient.