Entrepreneur: Start & Grow Your Business

Asset pricing in created markets.


by Newell, Richard G.^Papps, Kerry L.^Sanchirico, James N.

Although recent experience with the sulfur dioxide trading program in the United States has changed many perceptions, there are still questions about how well tradable permit systems for environmental pollution, greenhouse gases, agricultural production, and natural resources can work in practice. Such skepticism is in part warranted by the limited number of ex post assessments on the performance of created markets. Because building the necessary institutions can require significant political and economic costs, it is imperative to develop an empirical record of the performance of created markets in practice.

One area where market-based systems are subject to a significant degree of skepticism is in the management of ocean fisheries. One such system is individual fishing quotas, in which the total catch is capped and shares of the catch are allocated. An individual transferable quota (ITQ) system results when transfer of the shares is permitted. Over time, the least efficient fishermen should find it more profitable to sell their quota rather than fish it, both reducing excess capacity and increasing the efficiency of vessels operating in the fishery.

For ITQs to address the common pool problem in practice, it is important that quota markets are competitive and convey appropriate price signals. Price signals sent through the quota market are an essential source of information on the expected profitability of fishing and an important criterion for decisions to enter, exit, expand, or contract individual fishing activity. Quota prices also send signals to policymakers about the economic and biological health of a fishery. Arnason (1990) showed that under the assumption of competitive markets, monitoring the effect of changing the total allowable catch (TAC) on quota prices could be used to determine the optimal TAC.

In a previous study, Newell, Sanchirico, and Kerr (2005) (hereafter NSK) investigate the performance of ITQ markets using the most comprehensive dataset gathered to date for the largest system of its kind in the world. The panel dataset from New Zealand covers 15 years of transactions across the 33 species that were in the program as of 1998 and includes price and quantity data on transactions in more than 150 fishing quota markets. Markets exist in New Zealand both for selling the perpetual right to a share of a stock's TAC, as well as for leases of that right to catch a given tonnage in a particular year. NSK found that market activity appears sufficiently high to support a reasonably competitive market for most of the major quota species and that price dispersion has decreased over time. Investigating the asset and lease markets separately, they find evidence of economically rational behavior in each of the quota markets and their results show an increase in quota asset prices, consistent with increased profitability.

We extend the analysis of NSK by econometrically examining the relationship between the annual lease and sale prices in the perpetual quota asset markets. A notable exception to the virtually nonexistent literature examining quota prices in fisheries is the paper by Batstone and Sharp (2003), which investigates the relationship between fishing quota sale and lease prices and changes in the total allowable catch for the New Zealand red snapper fishery (region 1). Batstone and Sharp (2003) find support for the relationship proposed by Arnason (1990). Other related research in fisheries includes Karpoff (1984a, 1984b, 1995) and Huppert, Ellis, and Noble (1996), who look at the relationship between license prices and fishery rents in Alaska salmon fisheries.

With competitive markets, rational asset pricing theory suggests that the price of an income-producing asset in period t, [p.sub.t], should be determined by the real per-period profits from the asset, [[pi].sub.t], and the real discount rate, [r.sub.t]:

(1) [p.sub.t] = [[infinity].summation over (j=0)] [E.sub.t([[pi].sub.t] + j)]/ [[PI].sup.j.sub.k=0] (1 + [E.sub.t]([r.sub.t] + k)])

where E(*) is the expectations operator. In our setting, equation (1) states that the current quota asset price should be equal to the present discounted value of all future expected earnings, where the lease prices represent the annual flow of profits from holding quota. The price of the quota asset, therefore, will vary across fish stocks and over time based on changes in expected future lease prices or changes in the expected discount rate over time.

Under the simplifying assumption that expected lease prices and discount rates remain constant in the future, the price of the asset would simply equal the lease price divided by the discount rate, or [p.sub.t] = [[pi].sub.t]/[r.sub.t]. The expected rate of return from holding fishing quota (or dividend-price ratio) would be equal to [[pi].sub.t]/[p.sub.t]. Figure 1 supports the basic structure of such a relationship in New Zealand fishing quota, with the dividend-price ratio tracking both the level and the trend in New Zealand short-term interest rates over the sample period. For example, at the same time the dividend-price ratio fell by about half from 13% to 7%, the interest rate as measured by New Zealand Treasury bills fell from 10% to about 4% in real terms. Overall, the quota dividend-price ratio is about 2%-3% higher than the risk-free rate on average. Figure 2 likewise suggests a close, relatively linear association between asset and lease prices (in logs). The level of the average asset price is also approximately 10 times the lease price over the sample period, roughly equal to the present value of a perpetuity discounted at 10%.

[FIGURES 1-2 OMITTED]

Figure 1 also shows that there is considerable cross-sectional variation in the dividend-price ratio across fish stocks markets, where the upper and lower plus signs represent the 25th and 75th percentiles. Why might such variation exist? One reason could be that if fishers are risk averse they will prefer fish stocks with lower variance, other things equal. This effect is consistent with a higher discount rate, or higher required rate of return for riskier stocks. Such volatility could be associated with natural variation in stock abundance and economic variability in costs and fish prices. Another explanation could be differences in the expected growth rate of profits over time (Melichar 1979), possibly due to differences in output price growth, changes in fish populations, or other factors affecting costs such as cost rationalization due to quota trading.

Using panel data econometric techniques on an updated NSK dataset, we estimate models that relate the asset price of quota to their annual lease (or rental) price and observed determinants of the growth rate and volatility of rents. Within this framework, we explore the relationship between asset and lease prices, as well as whether differences in asset prices are due to differential risks associated with holding quota across fish stocks and/or different expected growth rates in fishery rents in those stocks. These data are uniquely qualified to address these questions, because of the relatively long time series, breadth of markets, and cross-sectional heterogeneity, as the market characteristics are diverse across both economic and ecological dimensions (see table 1 for a list of species included). For example, in 2000 the export value of these species ranges from about NZ$700 per ton for jack mackerel to about NZ$40,000 per ton for rock lobster. (1)

Consistent with asset pricing theory, we find a statistically (and economically) significant relationship between asset prices and contemporaneous lease prices. Stocks with a higher degree of biological volatility tend to have lower asset prices, and stocks that have rising returns or falling costs from fishing are found to have higher asset prices, ceteris paribus. Taken together, these results suggest that the price signals generated by the ITQ system are a good indication of the future profitability of individual fishing quota stocks. (2) The magnitude of some interrelationships is muted relative to what the theory suggests, possibly due to measurement error.

Our analysis also contributes to the extensive literature investigating asset prices by utilizing micro-level trading data across multiple (related) markets to measure the relationships embedded in equation (1), and the relative importance of the different factors behind the heterogeneity in figure 1. Nonfishery studies relevant to ours that investigate agricultural land prices and farming rents (e.g., Melichar 1979; Alston 1986; Falk 1991; Clark, Fulton, and Scott 1993; Just and Mirinowski 1993) or agricultural production quota (e.g., Barichello 1996; Wilson and Sumner 2004) typically focus on aggregate data and/or concentrate on a single market. For example, Falk (1991) models farmland prices in Iowa using aggregate price and rent data, and Wilson and Summer (2004) analyze the market for diary quota in California. The same holds for Batstone and Sharp (2003), who investigate a single quota market. Clark, Fulton, and Scott (1993) argue that a cross-sectional comparison of land markets can help illuminate the factors important in understanding the empirical relationship in equation (1).

In the next section, we provide a selected review of the literature modeling asset prices and dividends. This is followed by a description of the design of the ITQ system in New Zealand, paying particular attention to market characteristics. We then develop an empirical model that is appropriate to a multiple-asset setting like the New Zealand fishing quota market. We discuss the empirical specification, data sources, time-series properties of the data, estimation approach, and results, before we conclude by summarizing our findings.

Modeling Asset Prices and Dividends

The literature exploring the relationship between asset prices, dividends, and other relevant factors (e.g., firm size) is extensive. A thorough literature review is therefore beyond the scope of this article, and interested readers should consult Cochrane (2001) and Campbell, Lo, and MacKinley (1997) or the review articles by LeRoy (1989), Fama (1991, 1998), and Campbell (2000). (3)

Simplifying equation (1) under the assumption that the expected discount rate follows a martingale process yields (4)

(2) [p.sub.t] = [[infinity].summation over s=0] [E.sub.t]([[pi].sub.t+s])/ [(1 + [r.sub.t].sup.s+1].

Equation (2) illustrates how the asset price is dependent on the expected future stream of earnings, so that information available at time t along with type of expectation process is important in modeling the relationship between asset prices and dividends. For example, if one assumes that expected future earnings are constant, then [E.sub.t]([[pi].sub.t+s]) = [E.sub.t]([pi]). Huppert, Ellis, and Noble (1996) model and find support for an adaptive expectations process where [E.sub.t]([pi]) = [beta][[pi].sub.t-1] + (1 - [beta]) [E.sub.t-1]([pi]) with [beta] [member of [0, 1], and Karpoff (1984b) models a myopic process where [beta] = 1. Wilson and Sumner (2004) find support for a second-order adaptive expectation process in California dairy quota prices. Just and Miranowski (1993) test myopic, adaptive, and rational expectation regimes and find that farmland price data support myopic expectations. Falk (1991) finds a similar result. Orazem and Miranowski (1986) provide an empirical strategy for testing competing hypotheses of expectations regimes when direct measures of expectations are unavailable. Applied to farm acreage allocation decisions as a function of expected commodity prices, it yielded little evidence for favoring any of the three regimes.

If future profits (lease prices) grow at a constant rate g, then [[pi].sub.t] = (1 + g) [[pi].sub.t-1] + [[epsilon].sub.t], where [[epsilon].sub.t] is a white noise error term. Taking expectations and solving equation (2) forward in time with g < r, the asset price follows

(3) [p.sub.t] = [[pi].sub.t]/[r.sub.t] - g.

Equation (3) is the dynamic "Gordon growth model" (Campbell, Lo, and MacKinley 1997) that forms the basis of the majority of studies on the relationship between asset prices and dividends.

Due to a divergence between simple present-value relationships and empirical observations on agricultural land prices and rents during the 1970s and 1980s, a number of authors have extended this basic structure to include other factors, such as taxes (e.g., Robison, Lins, and VenKataraman 1985; Alston 1986), changes in risks (Barry 1980), and credit market constraints (Shalit and Schmitz 1982). Instead of investigating these many factors separately, Just and Miranowski (1993) develop a detailed structural model of the determinants of asset prices, which is a function of inflation, taxes, credit market imperfections, transaction costs, and risk aversion.

Others have focused on estimating a reduced form that is consistent with equation (2). For example, Burt (1986) argues that movements in asset prices may occur because of continued adjustment to past changes in returns, implying that the price does not adjust instantaneously to changes in expected future returns. In addition, expectations of future rents may be based on past, as well as current, values of [[pi].sub.t]. He approximated the effect of both sources of dynamic behavior by using a multiplicative distributed lag specification for [[pi].sub.t], with a restriction that the lag coefficients sum to unity.

Background on NZ ITQ System

We include a brief review of the New Zealand ITQ system with special attention to the elements that are most relevant for our analysis. For further history and institutional detail, see Batstone and Sharp (1999), Yandle (2001), NSK, and the references cited therein.

The New Zealand government passed the Fisheries Amendment Act in 1986, creating a national ITQ system. The system initially covered seventeen inshore species and nine offshore species, which together expanded to a total of forty-five species by 2000. Under the system, the New Zealand Exclusive Economic Zone (EEZ) is geographically delineated into quota management regions for each species based on the location of major fish populations. Rights for catching fish are defined in terms of fish stocks that correspond to a specific species taken from a particular quota management region. In 2000, the total number of fishing-quota markets stood at 275, ranging from 1 for the species hoki to 11 for abalone. As of the mid-1990s, the species managed under the ITQ system accounted for more than 85% of the total commercial catch taken from New Zealand's EEZ and from our calculations had an estimated market capitalization of about NZ$3 billion.

The New Zealand Ministry of Fisheries sets a TAC for each fish stock based on an intertemporal biological assessment (including the prior year's catch level) and other relevant environmental, social, and economic factors. The TACs are legislated to maintain the fish population at a level (or move it to a level) that will support the largest possible annual catch (i.e., maximum sustainable yield), after an allowance for recreational and other noncommercial fishing. Not all species have their TACs adjusted for noncommercial uses, especially those in the offshore sector where there is little if any recreational fishing (see table 1).5 Most TACs remain constant from year to year and for many fish stocks (especially those of low value) there are no formal stock assessments (Annala 1996). When a TAC needs to be adjusted there is no automatic process, and the appropriate level of the adjustment is discussed with the quota owners (Sanchirico et al. 2006).

Individual quota were initially allocated to fishermen free of charge as fixed annual tonnages in perpetuity based on their average catch level over two of the years spanning 1982-1984. Beginning with the 1990 fishing year, however, the government switched from quota rights based on fixed tonnages to quota denominated as a share of the TAC. Compliance and enforcement is undertaken through a detailed set of reporting procedures that track the flow of fish from a vessel to a licensed fish receiver (on land) to export records, along with an at-sea surveillance program including onboard observers.

Given the uncertainty around the quantity and composition of catch, a fisherman's quota holdings represent a mix of ex ante and ex post leases, as well as asset purchases and sales to cover actual catch. Although there are no official statistics, the general belief is that brokers handle a majority of the transactions between small and medium-sized quota (with a fee between 1% and 3% of the total value of the trade paid by the seller) and larger companies typically have quota managers on staff and engage in bilateral trades with other large companies. Whether ex ante or ex post transactions, fishing quota are generally tradable only within the same fish stock, and not across regions or species or years, although there have been some minor exceptions. (6) The quota rights can be broken up and sold in smaller quantities and any amount may be leased or subleased any number of times. Virtually all leases are for one year or less. There are also legislative limits on aggregation for particular stocks and regions, and limitations on foreign quota holdings.

NSK find that the quota markets are active, with about 140,000 leases and 23,000 quota asset sales occurring between economically distinct private entities between 1986 and 2000--an annual average of about 9,300 leases and 1,500 asset sales. Market participation has also increased over time with around 70% of quota owners taking part in a market transaction in 2000. Although some individual quota markets are thin, these tend to be of low economic importance in the size and value of the catch. The annual number of leases has risen ten-fold between 1986 and 2000, and the median percentage of total quota that are leased in these markets has risen consistently, from 9% in 1987 to 44% in 2000. At the same time, the total number of quota asset sales declined from a high of about 3,200 sales in 1986 (when initial quota allocations for most species took place), leveling off to around 1,000 sales in the late 1990s. The median shows a similar decline, with the percentage of total outstanding quota sold per year being as high as 23% at the start of the program, gradually decreasing in subsequent years to around 5% in the late 1990s. This pattern of asset sales is consistent with a period of rationalization and reallocation proximate to the initial allocation of quota, with sales activity decreasing after the less profitable producers have exited.

Empirical Analysis of Fishing Quota Asset Prices

Empirical Model

Our empirical assessment of the relationship between quota asset prices and expected future profits from fishing quota is based directly on the dynamic Gordon growth model (equation (3)). Within this framework, we explore possible explanations for the heterogeneity in quota asset prices across the different fishing quota markets, as illustrated in figure 1. Potential reasons for the heterogeneity include different growth rates of profits due to expected changes in revenues or costs, or because fish stocks are associated with different risk premia.

It is straightforward to allow for different asset prices, profits, and expected growth rates of profits across fish stocks, i. To investigate different risk premia, we follow the methods employed in Alston (1986) and Cochrane (1992) by decomposing the discount rate into a real market interest rate, [[??].sub.t], and an asset-specific risk premium, [[theta].sub.i]. Formally, this leads to

(4) [p.sub.i,t] = [[pi].sub.i,t]/[[??].sub.t] + [[theta].sub.i] - [g.sub.i].

In fishing quota markets, a major difference in risks stems from ecological volatility, whereby some fish stocks have more variable populations from one year to the next. Because search costs depend on the stock size and location, greater fluctuations in population abundance could lead to greater harvest and cost uncertainty.

In our setting, another important issue arises when considering the application of equation (4), which, for simplicity, assumes continuous growth into the indefinite future. In particular, fishing quota markets are created to address the "tragedy of the commons," and our analysis includes a period over which there was a market-based transition away from regulated open access conditions. Typically, when quota markets are created, fishing capital and labor inputs are distorted and fish populations are depleted due to years of operating under regulated open access conditions. An implication of this is that there will likely be a divergence between the current lease-asset ratio and the longer-term equilibrium, at least early on in the market, because at that time the contemporaneous lease price is not a good indicator of future profitability. This means that the asset price of a stock anticipating rationalization would initially be relatively high compared to its lease price. This divergence would decline over time as the stock achieved its anticipated profit increases and higher lease prices. Figure 1 suggests support for this hypothesis, as the difference between the 25th and 75th percentiles follows a downward trend.

Why might the divergence decrease over time? Initially, trades of the perpetual right to fish will occur as high-cost fishers find it more profitable to sell their quota rather than fish it. The gains from trade and elimination of excess fishing capital should result in cost savings. In addition, in many fisheries the cost function is likely to be stock-dependent, so that costs increase as the fish stock size falls and it becomes harder to find the fish (i.e., searching costs increase). As a result, if the TACs are set to allow stock recovery, then the gains due to stock rebuilding will also be incorporated in the expectation of future costs. The ability to time fishing trips to higher product prices rather than being forced to operate in short seasons, along with the shift from maximizing quantity to maximizing quality, will also feature in near-term expectations of future revenue growth. (7) These effects will likely dissipate over time as the potential gains are realized, where the rate of dissipation is an empirical question.

We modify equation (4) to account for these transitory effects by including a multiplicatively separable function, [psi](*), representing the transition associated with ITQs:

(5) [p.sub.i,t] = [[pi].sub.i,t]/[[??].sub.t] + [[theta].sub.i] - [g.sub.i] [psi](*).

We expect [psi](*) to be greater than one, because asset prices in ITQ markets will initially be above the levels predicted by the long-run relationships due to short-run expected profitability gains. Furthermore, we expect it to be larger for stocks with greater short-term gains, but to be decreasing over time, as asset prices should converge to the long-run relationship after some interval of time, holding everything else equal. The arguments of [psi](*) can include, for example, time since the market was created, and variables that represent gains from trade and fish stock recovery.

Empirical Specification and Data

After adding and subtracting 1 in the denominator of equation (5) (see footnote 12), we take a logarithmic approximation. We also approximate ln[psi](*) by [[beta].sub.5][s.sub.ijy] + [[beta].sub.6][a.sub.ij] +[[beta].sub.7][a.sub.ij[t.sub.y], where s is a measure of expected future cost declines due to reallocation of fishing effort through trading, a indicates the effect of expected future cost reductions on increases in fish stock abundance, and t is an annual time index. (8) Specifically, the relationship we bring to the quota asset price data is

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where p is the quarterly average quota asset price, [pi] is the contemporaneous quota lease price (as a measure of the annual profits from fishing), [??] is the real interest rate, ln[theta] is proxied by each species natural mortality rate (a measure of risk), and g is proxied by a measure of expected future growth in the output price of fish species i. (9) We also include a dummy variable (d) for shellfish stocks (i.e., abalone, rock lobster, and scallops), a set of quarterly fixed effects ([[alpha].sub.q]), a set of yearly fixed effects ([[alpha].sub.y]), a fish-stock-specific effect ([v.sub.ij]), whose specification varies depending on the estimation approach (e.g., fixed or random effects), and an independent and identically distributed error term ([epsilon]). Species are denoted by the subscript i and regions by j, so that each ij combination indexes a different fishing quota market. Time is indexed by quarter q of year y.

The model and accompanying discussion above imply the following hypotheses for the model: [[beta].sub.1] > 0, [[beta].sub.2] < 0, [[beta].sub.3] < 0, [[beta].sub.4] > 0, [[beta].sub.5] > 0, [[beta].sub.6 > 0, and [[beta].sub.7] < 0. Strict interpretation of the logarithmic approximation given by equation (6) further implies the following hypotheses about the specific magnitudes of certain coefficients: [[beta].sub.1] = 1, [[beta].sub.1] [approximately equal to] -(1 + r)/(r + [theta] - g), [[beta].sub.3] [approximately equal to] -[theta]/(r + [theta] - g), and [[beta].sub.4] [approximately equal to] (1 + g)/(r + [theta] -g), where each of the variables in these formulae is taken to be its mean value (the point of approximation). We do not impose these as restrictions, but rather consider them when interpreting the findings below.

We estimate equation (6) using the comprehensive panel dataset described in detail in NSK, which was constructed using information from New Zealand government agencies and other sources. We include 152 fish stocks representing 32 different species that had entered the New Zealand ITQ system by 1998. The data cover 14 years from the 1987-1988 fishing year until the end of the 2000-2001 fishing year. All monetary figures were adjusted for inflation to year 2000 New Zealand dollars, using the producer price index (PPI) from Statistics New Zealand. Table 2 gives descriptive statistics for the 4,120 observations that comprise the estimation sample; the included variables exhibit a large degree of variation.

As described above, the quota asset and lease prices are quarterly averages for each species-region specific fish stock quota market, based on more than 140,000 underlying lease transactions and more than 23,000 asset transactions. (10) The real market interest rate, [??], is the 90-day New Zealand Treasury bill rate, adjusted for inflation using the New Zealand CPI. As a measure of variation in the risk premium across species, ln[theta], we use each species' natural mortality rate. Species with higher mortality rates have population sizes that are typically more variable due to fewer age classes, which we argue leads to increasingly greater uncertainty in the amount of fish likely to be caught with a given level of effort. As a consequence, there is greater uncertainty in the profits from fishing high-mortality species, and we would therefore expect higher mortality rates to have a negative effect on quota asset prices. (11) We base g on the historic growth rate in output prices, where output prices are based on the export price per greenweight ton using data from Statistics New Zealand over the period 1986-2001, deflated using the NZ PPI (see NSK). (12)

Empirically, the components of the approximation to the [psi](*) function are as follows. To represent expected future profit increases due to reallocation of fishing effort through trading, s, we use the annual percentage of quota assets sold for each fish stock, normalized by dividing by each stock's average percentage sold. The hypothesis is that reallocation of quota assets is an indication of expected future profits from that trade, most likely through cost reductions.

Improvements in profits through cost reductions can also occur as a result of improvements in fish stock abundance, and associated increases in the catch-per-unit-effort. We represent this feature using a dummy variable, a, that indicates whether each stock faced significant reductions upon implementation of the ITQ program. (13) We expect that fisheries plagued by excess capacity and overfishing prior to the implementation of the ITQ system that also faced significant reductions in allowable catch at the outset of the ITQ program would experience greater increases in profitability through stock rebuilding and cost rationalization than fish stocks without a high degree of overfishing, everything else being equal. Thus, we would expect the coefficient on a to be positive, indicating that for a given lease price, the asset price will be higher for stocks with fish stock rebuilding plans in place.

Over time, however, the gains from such improvements should be realized, implying that future gains will be lower. We capture this effect by interacting a with a time trend, hypothesizing that over time the lease price will rise as stocks improve, and the effect on the asset price of additional future gains will diminish. Under these conditions, we would expect the coefficient on the interaction of a and t to be negative.

Estimation Approach

Time-Series Properties of Data

Before considering estimation of equation (6), it is essential to determine the time series properties of the asset and lease price series. If either one or both of the series are nonstationary, then standard regression techniques will be susceptible to the problem of spurious regression.

While testing for unit roots in panels is a relatively new enterprise, there are several tests available to researchers (see Banjeree [1999] for more information on the tests). We employ three tests, all of which can be thought of as panel data extensions or pooled versions of the Dickey-Fuller test (or Augmented Dickey-Fuller test when lags are included). Full details are given in the supplemental appendix (Newell, Papps, and Sanchirico 2007). In all three cases, we reject the hypothesis of a unit root in both the asset and lease price series at the 1% level. The same result holds when the tests are repeated using species-level (rather than stock-level) data. The agreement in the time series properties of the asset and lease prices satisfies, at least at the panel level, a necessary condition of the present-value model (Falk 1991). (14)

We also test for the possibility of n