- Email: [email protected]
Commodity price volatility under regulatory changes and disaster
Journal of Empirical Finance 38 (2016) 355–361
Contents lists available at ScienceDirect
Journal of Empirical Finance journal homepage: www.elsevier.com/locate/jempfin
Commodity price volatility under regulatory changes and disaster☆ Akbar Marvasti a,b,⁎, Antonio Lamberte c a b c
U.S. Department of Commerce, Southeast Fisheries Science Center, NOAA,75 Virginia Beach Dr., Miami, FL 33149, United States Department of Economics, University of Miami, United States U.S. Department of Commerce, Southeast Regional Office, NOAA 263 13th Avenue South, St. Petersburg, FL 33701, United States
a r t i c l e
i n f o
Article history: Received 19 December 2015 Received in revised form 7 July 2016 Accepted 9 July 2016 Available online 21 July 2016 JEL classification: C22 C52 Q22 Keywords: Price volatility GARCH Time series Regulatory change Disaster
a b s t r a c t We find that the EGARCH model best describes the dynamics of U.S. Gulf of Mexico red snapper daily dockside prices and find their reaction to shocks to be asymmetric, though news has an impact on volatility level in a direction contrary to that of financial asset prices. We also find that volume contains useful information for predicting volatility. However, unlike financial asset prices, though consistent with fish commodities prices, red snapper price volatility diminishes when the volume is high. Also, the effect of expected changes on transaction volume is more dominant than that of unexpected changes. Explicitly accounting for oil spill closures and the Individual Fishing Quotas (IFQ) program in other species as variance shift parameters significantly reduces volatility and improves the market efficiency response to shocks. Published by Elsevier B.V.
1. Introduction The fishing industry provides some important commodities and this paper studies the price and volume of red snapper and the effects of environmental factors, quotas and other government regulations on their price and supply. In particular, we examine price of red snapper, which is an important fish commodity commonly found in the Gulf of Mexico (GOM). The price and trading volume of red snapper reflect production uncertainty and patterns of volatility that have similarities with those in well-organized financial markets and are interestingly enough related but different to the usual assets considered in finance. However, previous work by Chambers and Bailey (1996) and Deaton and Laroque (1992) have presented theories of commodity prices that imply a positive correlation between price and volatility that is similar to other financial returns. Studies examining fish commodity prices and market volatility are very limited. However, Bose (2004) finds persistent volatility in the prices of several fish species in Australia and a negative correlation between the variance of prices and trade volumes. While Oglend and Silkveland (2008)document evidence of increased volatility when prices for Norwegian salmon are high. The
☆ The authors are grateful for helpful guidance from Richard Baillie, one of the Journal editors. We are also thankful to Jessica Stephen and Andy Strelchek of NOAA for the data. The opinions expressed herein are those of authors and do not necessarily reflect the views of NOAA. ⁎ Corresponding author at: U.S. Department of Commerce, Southeast Fisheries Science Center, NOAA,75 Virginia Beach Dr., Miami, FL 33149, United States. E-mail address: [email protected] (A. Marvasti).
http://dx.doi.org/10.1016/j.jempfin.2016.07.008 0927-5398/Published by Elsevier B.V.
356
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
authors also use trade volume as an exogenous variable in the variance equation and find a negative correlation between trade volume and price volatility, which they attribute to the availability of inventory. In this paper, we examine the pricing and volatility of red snapper and how they are affected by some relevant exogenous variables. For this purpose, we apply the standard GARCH model and its extensions to the log of the daily price of red snapper. Our models are similar to those employed by Baillie and Myers (1991) for a variety of commodities; but we also include volume of trading and dummy variables to investigate the impact of various exogenous shocks. We also examine nonlinearity in the dynamic of red snapper prices and investigate the asymmetry in the effects of exogenous shocks to the red snapper fishing industry. In particular we assess the effects of oil spill shocks to the red snapper commodity. The rest of this paper is structured as follows: Section 2 provides institutional background and briefly describes the red snapper individual fishing quota (IFQ-RS) program. Section 3 discusses the stylized facts of price and other variables, as well as their dynamic properties. Section 4 presents various econometric models to which we apply variations of the GARCH model and discusses the empirical results. Section 5 summarizes the key findings and concludes. 2. Red snapper IFQ system and other regulatory controls As one of the most important fishery species in the GOM, red snapper stocks have been subjected to quota management since 1990 with the establishment of a total allowable catch (TAC), allocating 51% of the quota to the commercial sector and the remaining 49% to the recreational sector. The commercial red snapper quota was reduced in 1991 over concerns of overfishing. This quota reduction brought about some fishery closures and also market gluts as fishermen attempted to harvest as much as they could before the quota was reached and the fishery closed; see Waters (2001). Subsequent management measures, including an endorsement system in 1993 that was later converted to a two-tier license limitation system in 1998, a 10-day open season each month, and a quota increase in 1996, alleviated the derby fishing problem. Nonetheless, dockside prices remained low until January 1, 2007, when the IFQ-RS program was implemented, replacing the two-tier license limitation program. This program effectively expanded the commercial harvest season length from approximately 85 days a year to year-round. Since then, prices appear to have stabilized at higher levels, consistent with the general expectation from the IFQ-RS program, and the derby fishing problem has been essentially eliminated.1 Independent of the IFQ program, the commercial TAC has continued to change based on the status of the stock, fluctuating between 2.297 million pounds (gutted weight) in 2007 to 3.713 million pounds in 2012 (Fig. 1). For example, the TAC was reduced by 23% to address continuing overfishing in 2008 and was maintained at that level until the end of 2009. However, the typical short-term market gluts associated with the fishing derby did not recur. Although red snapper prices rose during this period in reaction to the reduced quota, the price reaction was not as severe as in the early 1990s. The industry has recently gone through two other notable regulatory changes. In January 2010, the IFQ-RS program was complemented by an IFQ program for the grouper-tilefish species (IFQ-GT), many of which are harvested with red snapper. The IFQ-GT program appears to have contributed to a geographic expansion of red snapper landings by allowing the trading of share allocations between these two programs and may possibly have affected price volatility for red snapper. This geographic expansion has been aided by the rebuilding of red snapper stocks in West Florida, as well as a generally increasing population. The IFQ-GT is thus believed to have led to an increase in both the diversity of the catch composition in the GOM and the number of vessels harvesting red snapper (NOAA, 2013). Another notable event for the industry was the April 20, 2010 Deepwater Horizon oil spill, which was one of the largest in U.S. history. The extent of the potential damage to the fishery was a serious environmental disaster causing the government to close various areas of the GOM to fishing between May 2 and December 30, 2010. This paper estimates the average percentage of the GOM area that was closed for each day of that period due to the Deepwater Horizon oil spill (Fig. 2). The peak closure was reached on June 3, 2010, when approximately 36% of the GOM was closed to fishing. Evidence in this paper indicates that the volatility of red snapper prices is directly affected by the oil spill over the regular market volatility. 3. Data and properties of the time series Our analysis is based on daily red snapper prices during the period of January 3, 2007 through September 30, 2015, which are obtained from the National Oceanographic and Atmospheric Administration (NOAA).2 Since 2007, the commercial red snapper fishing season has been open throughout the year; although the actual landings under the IFQ-RS program started on January 3, 2007. The IFQ-RS monitoring system electronically records all daily commercial red snapper landings and the corresponding (ex-vessel) price per pound. Under the red snapper IFQ program, commercial vessels may land at any time; but fish may not be offloaded between 6 P.M. and 6 A.M. A landing transaction is then completed by the IFQ dealer recording the date, location of transaction, weight and actual ex-vessel value of fish landed. The number of dealers purchasing red snapper throughout the 1 Initially, the IFQ shares were issued based on the number of red snapper landings reported under each participant's license during a specific time period. IFQ shares for Class 1 license holders were based on landings in the best ten consecutive years in 1990–2004. For historical captain's license holders, the IFQ shares were based on seven years of landings in 1998–2004. For Class 2 license holders, the IFQ shares were based on landings in the best five years in 1998–2004. A total of 546 entities, natural or juridical persons, qualified for the initial IFQ shares, in which 0.0001% was the lowest share issued and 6.0203% the highest. The number of shareholders has dropped since the beginning of the program, from 554 to 407 in 2012. 2 Daily prices have only been available since the institution of the IFQ-RS program in 2007.
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
357
Fig. 1. Commercial Quota for GOM Red Snapper (Source: NOAA, 2013).
Fig. 2. Percentage of GOM Closed Due to the DWH Oil Spill.
GOM since the inception of the red snapper program has varied between 66 and 96. Currently there are 316 IFQ dealer's facilities in the GOM. Therefore, each dealer may have multiple facilities for landing and/or the processing of fish. The distribution of the facilities and dealers tends to be proportionate to the length of the coast lines in each state. Accordingly, Florida, which has the longest coast line in the GOM, holds approximately 73% of the facilities. While most dealers operate in only one state, a few have facilities in more than one state. Most dealers purchase fish from 3 to 10 different fishermen (NOAA, 2013). The daily price used in this study is the average of recorded transactions by all dealers in the GOM market. Therefore, the composition of the dealers in each day can differ as the fish stock availability can change with water temperature, currents, and availability of baits. There were 131 days in the sample period when no data is available and is either due to no landings occurring on these days, or because dealers waited until the following day to report the transactions. This leaves a sample size of 3062 days with a mean price of $3.96 per pound and a variance of $0.53. The distribution of the price is slightly skewed to the left (see Fig. 3). Since the log of the series turns appears to be stationary, the unconditional long-run variance of the price is constant. Similar to the price series, the distribution of the landings is skewed negatively. This suggests that during the high season, when the stocks are plentiful, the average landing is typically high, while the fishing trips are less productive during the low season. Fig. 4 shows the log of landings (volume), which also shows a slight upward trend. The depictions of both the price and landings data show that there is more volatility below the trend than above it. Descriptive statistics for the log of price and other variables are in Table 1. The stylized facts of the log of the red snapper price series in terms of volatility clustering and fat tails are less severe than most other economic and financial time series. We also explore the effects of some exogenous variables on the volatility of red snapper prices using NOAA data to study the effects of quotas, the Deepwater Horizon oil spill closure, and the volume of transactions (landings). We considered two variations of the quota data: either the log of the series or a dummy for the quota reduction that takes the value of one when the quota is reduced relative to its previous level and zero otherwise.3 As a result of the 2010 oil spill, rather large areas of the GOM were closed to fishermen. We use a dummy variable capturing the significance of the oil closure event in our model; the variable takes the value of one during the closure period and zero otherwise. We also consider the average percentage of the GOM
3 We did not consider a quota increase dummy as well because it turns out be very similar, though inversed, to the dummy for quota reduction, except that it lasts six months.
358
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
Fig. 3. Log of Red Snapper Price.
Fig. 4. Log of Landings (Volume).
area closed each day as an alternative to the dummy variable. Finally, a dummy variable is created to represent the introduction of the IFQ-GT; this variable is equal to one for the time period after January 2010 and zero otherwise. Contrary to most financial assets, red snapper prices appear to be stationary based on the Augmented Dickey Fuller (ADF) test results. The log of quota and log of oil spill area closures appear to be I(1) series and are first differenced. Bai Perron tests are applied to test for structural breaks in the log of price of red snapper.4 The sup F statistic is 23.37 and WD max F is 38.69; both are significant at the 0.01% level. One of the relevant breaks is on May 29, in 2010, which is 27 days after the start of the oil spill, when more than 24% of the GOM was closed to fishing. The actual peak of the closure is on June 3, which is within the 95% confidence interval of this break.
4. Modeling the dynamics of prices Based on the unit root test results, we model the log of price as yt = lnPt and investigate the dynamics of the price series. The autocorrelations and partial autocorrelations of the red snapper series indicate quite persistent autocorrelations, while the partial autocorrelations are not significantly different from zero after lag seven. The use of BIC suggests that an AR(7) model is the most appropriate description of the conditional mean. Specification tests reveal the presence of ARCH effects, which led to the following lagged dependent variable regression:
yt ¼ ϕ0 þ
7 X
GT
ϕk yt−k þ λ1 vt þ λ2 Q Rt þ λ3 Ot þ λ4 IFQ t þ εt ;
ð1Þ
k¼1
4 The following Bai-Perron multiple structural change test is applied for the existence, number, and timing of breaks in the price volatility: yt =x′β+z ′δ t t j +μt, where t=Tj−1 +1, . . . ,Tj for j=1, . . . ,m+1. Here yt is the dependent variable, xt(p...1) and zt(q..1) are vectors of covariates, β and δj(j=1, ...,m+1) are coefficient vectors, and μt is the error term. More detailed information can be found in Bai and Perron (2003).
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
359
Table 1 Descriptive Statistics. Variables
Mean
Standard deviation
Skewness
Kurtosis
J.B. statistics
Log of red snapper price Log of volume (gutted landings in pounds) Log of oil closure days Log of quota Quota reduction dummy (Reduction = 1, Otherwise = 0) IFQ-GT introduction (IFQ-GT Present = 1, Otherwise = 0)
1.36 8.66 1.92 15.01 0.27 0.67
0.19 1.34 1.27 0.31 0.44 0.47
−0.79 −1.36 3.40 0.38 1.07 −0.71
7.10 5.99 13.18 2.11 2.14 1.50
2463.71 (0.00) 2585.93 (0.00) 19,131.70 (0.00) 175.33 (0.00) 676.12 (0.00) 542.14 (0.00)
Sample: January 3, 2007 to October 30, 2015 (N = 3193).
q p r X X ε X ε 2 2 log δt ¼ ω0 þ β j log δt− j þ α i t−i þ γ k t−k þ θ1 V t þ δ δt−k t−i j¼1 i¼1 k¼1
ð2Þ
θ2 QRt þ θ3 Ot þ θ4 IFQ t þ νt :
The additional variables in this model include Vt, the transaction volume; QRt, a dummy variable representing the quota reduction; Ot, the log of the percentage of the daily GOM oil spill closure; and IFQGT, a dummy variable for the introduction of the grouper-tilefish IFQ program. In contrast to the mean equation, there are no a priori sign expectations for the effects of the exogenous variables in the variance equation. After testing different error distribution assumptions, we select a normal error distribution model as the best fit. Various ARMA-GARCH models are then estimated using quasi-maximum likelihood estimation (QMLE) as in Bollerslev and Wooldridge (1992). The estimated models are presented in Table 2. Our estimate of the presence of volatility depends on whether explanatory variables are included. Before adding any exogenous variables to the basic model, we test the asymmetric effect of innovations on the conditional variance. When, the Engle and Ng (1993) leverage effect test is applied, the t-value for the lagged residuals turns out to be negative and statistically highly significant. This suggests that the sign of the current period shock provides helpful information in predicting conditional volatility. We then estimate the basic EGARCH model with asymmetric effects, finding that the parameter estimates suggest a high degree of persistence in the conditional variance (The results are omitted for brevity.). Price volatility persistence is not uncommon in fishery: Oglend and Silkveland (2008) estimate the degree of volatility persistence in the salmon market at 0.81, while Buguk et al. (2003) arrive at a degree of volatility persistence of 0.98 for the catfish market and 0.38 for the menhaden market. We then proceed to enhance our basic model by including volume, which reduces volatility.5 The results are reported in Table 2 as model 1. The response to new information, α1, has increased, while volatility decay speed, β1, is significantly diminished relative to the basic (unreported) model. Unlike some commodity studies (e.g. Clark, 1973; Lamoureux and Lastrapes, 1990; Fleming et al., 2006), the GARCH coefficient also remains statistically significant. Typically, a larger volume of transactions tends to reduce variance in financial data (e.g., Deaton and Laroque, 1992; Chambers and Bailey, 1996; Bessembinder and Seguin, 1993; Fleming et al., 2006). Clark (1973) argues that volume and volatility tend to be very close or the same. We find trade volume to be negatively correlated with the variance of red snapper price. Oglend and Silkveland (2008), in their analysis of salmon price volatility behavior, attribute the negative correlation between trade volume and price volatility to the availability of inventories, which might also be a valid argument here. The authors argue that large volume increases inventories, thereby smoothing out price fluctuations. The coefficient of γ1 suggests that the leverage effect is present in the conditional variance. In other words, the impact of news is not symmetric, and there is a tendency for more of an increase in volatility following positive shocks than negative shocks of the same magnitude, unlike the typical asymmetric results with financial assets. Model 2 considers the effect of trade volume on price by separating volume into expected (θE1) and unexpected (θU 1 ) components. The model for expected volume is based on daily seasonal dummies and a linear time trend. The estimated AR-EGARCH model with predicted and unpredicted trade volume measures is presented, with the results showing a negative correlation in the conditional variance equation for both components of trade volume. Surprises in trading volume have a larger effect on price volatility than the forecastable component, which is consistent with the findings of Bessembinder and Seguin (1993), who also show both components of volume have a positive effect on price volatility. However for the red snapper series in this study there appears to be a negative effect; which is consistent with the findings of Oglend and Silkveland (2008). In our result, the unanticipated change in trade volume that will move the red snapper price by one unit is approximately 0.45, which is within the market depth range for financial assets reported by Bessembinder and Seguin (1993). The effects of quotas and the oil spill closures are represented by dummy variables, which together with IFQ-GT are included as explanatory variables in the conditional mean and variance equations of the estimated models presented as model 3 in Table 2. The addition of these variables reduces the skewness and kurtosis of the residuals and improved the fit of the model. We discover that quota reductions have no statistically significant effect on the mean price, while they have a positive effect on the variance. 5 A high degree of volatility persistence may be due to failure to account for the presence of structural changes, as has been argued by Engle and Bollerslev (1986) and Gray (1996). The topic has been further discussed and investigated in several empirical studies (e.g., Lamoureux and Lastrapes, 1990; Simonato, 1992; Hillebrand, 2005; Galeano and Tsay, 2010). In our study, the inclusion of red snapper trading volume in the variance function is likely to tone down the degree of volatility persistence in the price.
360
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
Table 2 AR(7)-EGARCH Models with Asymmetry Parameter Estimates with Trade Volumes, and Other Exogenous Variables. Parameters
Model 1
Model 2
Model 3
Model 4
Model 5
0.146⁎⁎⁎ (0.017) 0.086⁎⁎⁎ (0.016) 0.056⁎⁎⁎
0.120⁎⁎⁎ (0.016) 0.092⁎⁎⁎ (0.016) 0.075⁎⁎⁎
0.119⁎⁎⁎ (0.017) 0.059⁎⁎⁎ (0.015) 0.041⁎⁎⁎
0.118⁎⁎⁎ (0.017) 0.058⁎⁎⁎ (0.015) 0.041⁎⁎⁎
0.122⁎⁎⁎ (0.017) 0.058⁎⁎⁎ (0.015) 0.038⁎⁎⁎
ϕ6
(0.016) 0.108⁎⁎⁎ (0.015) 0.077⁎⁎⁎ (0.016) 0.131⁎⁎⁎
(0.015) 0.099⁎⁎⁎ (0.014) 0.090⁎⁎⁎ (0.016) 0.114⁎⁎⁎
(0.016) 0.091⁎⁎⁎ (0.015) 0.0617⁎⁎⁎ (0.015) 0.108⁎⁎⁎
(0.016) 0.090⁎⁎⁎ (0.015) 0.062⁎⁎⁎ (0.015) 0.108⁎⁎⁎
(0.016) 0.090⁎⁎⁎ (0.015) 0.060⁎⁎⁎ (0.015) 0.107⁎⁎⁎
ϕ7
(0.017) 0.190⁎⁎⁎
(0.017) 0.178⁎⁎⁎
(0.017) 0.160⁎⁎⁎
(0.017) 0.159⁎⁎⁎
(0.017) 0.160⁎⁎⁎
λ2
(0.018) –
(0.018) –
λ3
–
–
(0.017) 0.001 (0.007) –
λ3 ,t+56
–
–
(0.017) 0.001 (0.007) −0.093⁎⁎⁎ (0.016) –
(0.017) 0.007 (0.006) −0.085⁎⁎⁎ (0.015) –
λ4
–
–
Mean Equation: ϕ1 ϕ2 ϕ3 ϕ4 ϕ5
−0.099⁎⁎⁎ (0.017) 0.053⁎⁎⁎
–
–
0.054⁎⁎⁎ (0.007) –
θ1
0.254⁎⁎⁎ (0.055) 0.279⁎⁎⁎ (0.063) −0.423⁎⁎⁎
0.442⁎⁎⁎ (0.073) 0.195⁎⁎⁎ (0.107) –
0.173⁎⁎⁎ (0.054) 0.209⁎⁎⁎ (0.062) −0.437⁎⁎⁎
0.176⁎⁎⁎ (0.053) 0.206⁎⁎⁎ (0.062) −0.437⁎⁎⁎
0.158⁎⁎⁎ (0.055) 0.215⁎⁎⁎ (0.063) −0.434⁎⁎⁎
θE1
(0.027) –
(0.026) –
(0.027) –
(0.025) –
θU 1
–
–
–
–
θ2
–
θ3
–
−
0.630⁎⁎⁎ (0.125) 0.346⁎⁎⁎
0.647⁎⁎⁎ (0.127) –
0.626⁎⁎⁎ (0.122) −0.330⁎⁎⁎
θ3,t+56
–
−
θ4
–
–
γ1
0.148⁎⁎⁎ (0.031) 1531.82 −0.97 0.04 (prob. 0.85)
0.234⁎⁎⁎ (0.076) 1784.75 −1.13 0.44 (prob. 0.51)
δ2t Variance equation: α1 β1
Log Likelihood BIC ARCH LM F-stat.
−0.383⁎⁎⁎ (0.062) −0.445⁎⁎⁎ (0.108) –
(0.144) – 0.881⁎⁎⁎ (0.113) 0.122⁎⁎⁎ (0.032) 1662.68 −1.04 0.79 (prob. 0.37)
(0.006) –
0.059⁎⁎⁎ (0.007) −0.460⁎⁎⁎ (0.149)
0.404⁎⁎⁎ (0.151) 0.887⁎⁎⁎ (0.114) 0.124⁎⁎⁎ (0.032) 1667.19 −1.04 0.82 (prob. 0.37)
(0.141) – 0.876⁎⁎⁎ (0.112) 0.116⁎⁎⁎ (0.033) 1665.72 −1.04 1.43 (prob. 0.23)
Note: The error term is assumed to follow normal distribution. Heteroskedasticity consistent standard errors are in parentheses; ⁎⁎⁎ denotes significance at 1%; degrees of freedom (d.f.) is the estimated tail spread of the t distribution.
The oil spill closures, on the other hand, have reduced the mean of red snapper price, perhaps by cutting supply, but the probable level of anxiety they caused appears to have caused an increase in price volatility. Finally, the introduction of the IFQ-GT has both increased the mean and price volatility. This may be due to interactions between the species' landings and vessels/permits variables. We extend our investigation of the role of the oil spill event by applying the two relevant oil spill structural break dates identified in Section 3. Both the May 29 and June 3 breaks produced similar results to that of the initial oil spill closure dummy variable, though the magnitude of the coefficients for this variable increase slightly in both the mean and the variance equations. For brevity, only the results from the May 29 break are reported as model 4.6
6 The log difference of the average percentage of the GOM area that was closed each day, as an alternative measure of the oil spill, showed a positive effect on the mean of price, but no effect on the variance.
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
361
The effect of changes in the price risk of red snapper on its mean price is tested by including the variance of the log of GARCH in the mean equation in model 5. It turns out that changes in the riskiness of purchasing red snapper have a negative effect on the mean, similar to the effect of risk in financial assets. 5. Summary and conclusion This study provides a better understanding of the price risk that fishermen typically face, affecting their profitability and possibly their investment decisions. Analysis of price volatility is also important to the retail seafood market and policymakers. We examine the behavior of red snapper dockside prices for the first few years of the IFQ system for this fishery. The autocorrelation function suggests stationarity and highly seasonal behavior. However, other stylized facts of the log of the red snapper price series are similar to most other economic and financial time series in terms of volatility clustering, fat tails, and volatility mean reversion. We employ an array of GARCH models; and conclude that the EGARCH model describes the dynamics of red snapper price better than does a standard linear model, but could not reject the presence of asymmetry in the red snapper price response to shocks. Analyses of financial data have typically found a link between the volume of transactions and price volatility. We find that red snapper price volatility is diminished when trade volume is high, which is contrary to the findings from financial assets, but is consistent with the behavior of fish commodity prices. However, the market depth is not significant when other exogenous variables are absent in the model. In other words, changes in the volume of transactions do not cause as much change in price volatility as they do in financial asset prices. We can only speculate whether the association between red snapper price volatility and volume is due to inventory changes, geographic expansion, or increase in the number of fishermen landing red snapper on highvolume days. Using the regularity pattern for transaction volume during the week, we separate the expected changes from the unexpected changes in the volume of transactions. It turns out that while transaction volume provides useful information regarding price volatility, the effect of expected changes in transaction volume is more pronounced than that of unexpected shocks. We also find that reductions in quota have increased the red snapper prices volatility. On the other hand, the 2010 oil spill closures in the GOM have had a dampening effect on the conditional mean while increasing price volatility. Furthermore, the debut of the closely related IFQ-GT program in 2010 is associated with an increase in red snapper price volatility. Finally, we discover that volatility has not been constant over the study period and that there is a positive correlation between price and volatility, as measured by the GARCH variance. References Bai, J., Perron, P., 2003. Computation and analysis of multiple structural change models. J. Appl. Econ. 18, 1–22. Baillie, R.T., Myers, R.J., 1991. Bivariate GARCH estimation of the optimal commodity futures hedge. J. Appl. Econ. 6, 109–124. Bessembinder, H., Seguin, P.J., 1993. Price volatility, trading volume, and market depth: evidence from futures markets. J. Financ. Quant. Anal. 28, 21–39. Bollerslev, T., Wooldridge, J.M., 1992. Quasi-maximum likelihood estimation and inference in dynamic models with time varying covariances. Econ. Rev. 11, 143–172. Bose, S., 2004. Price volatility of south-east fishery's quota species: an empirical analysis. Int. Econ. J. 18, 283–297. Buguk, C., Hudson, D., Hanson, T., 2003. Price volatility spillover in agriculture: an examination of U.S. catfish markets. J. Agric. Resour. Econ. 28, 86–99. Chambers, M.J., Bailey, R.E., 1996. A theory of commodity price fluctuations. J. Polit. Econ. 104, 924–957. Clark, P.K., 1973. A subordinated stochastic process model with finite variance for speculative prices. Econometrica 41, 135–155. Deaton, A., Laroque, G., 1992. On the behaviour of commodity prices. Rev. Econ. Stud. 59, 1–23. Engle, R.F., Bollerslev, T., 1986. Modelling the persistence of conditional variances. Econ. Rev. 5, 1–50. Engle, R.F., Ng, V., 1993. Measuring and testing the impact of news on volatility. J. Financ. 48, 1749–1778. Fleming, J., Kirby, C., Ostdiek, B., 2006. Stochastic volatility, trading volume, and the daily flow of information. J. Bus. 79, 1551–1590. Galeano, P., Tsay, R.S., 2010. Shifts in individual parameters of a GARCH model. J. Financ. Economet. 8, 122–153. Gray, S.F., 1996. Modeling the conditional distribution of interest rates as a regime-switching process. J. Financ. Econ. 42, 27–62. Hillebrand, E., 2005. Neglecting parameter changes in GARCH models. J. Econ. 129, 121–138. Lamoureux, C., Lastrapes, W., 1990. Persistence in variance, structural change, and the GARCH model. J. Bus. Econ. Stat. 8, 225–233. National Oceanic and Atmospheric Administration (NOAA), 2013. Red Snapper Individual Fishing Quota Program 5-Year Review. Southeast Regional Office, National Marine Fisheries Service, St. Petersburg, Florida. Oglend, A., Silkveland, M., 2008. The behaviour of salmon price volatility. Mar. Resour. Econ. 23, 507–526. Simonato, J.G., 1992. Estimation of GARCH process in the presence of structural change. Econ. Lett. 40, 155–158. Waters, J.R., 2001. Quota management in the commercial red snapper fishery. Mar. Resour. Econ. 16, 65–78.
Contents lists available at ScienceDirect
Journal of Empirical Finance journal homepage: www.elsevier.com/locate/jempfin
Commodity price volatility under regulatory changes and disaster☆ Akbar Marvasti a,b,⁎, Antonio Lamberte c a b c
U.S. Department of Commerce, Southeast Fisheries Science Center, NOAA,75 Virginia Beach Dr., Miami, FL 33149, United States Department of Economics, University of Miami, United States U.S. Department of Commerce, Southeast Regional Office, NOAA 263 13th Avenue South, St. Petersburg, FL 33701, United States
a r t i c l e
i n f o
Article history: Received 19 December 2015 Received in revised form 7 July 2016 Accepted 9 July 2016 Available online 21 July 2016 JEL classification: C22 C52 Q22 Keywords: Price volatility GARCH Time series Regulatory change Disaster
a b s t r a c t We find that the EGARCH model best describes the dynamics of U.S. Gulf of Mexico red snapper daily dockside prices and find their reaction to shocks to be asymmetric, though news has an impact on volatility level in a direction contrary to that of financial asset prices. We also find that volume contains useful information for predicting volatility. However, unlike financial asset prices, though consistent with fish commodities prices, red snapper price volatility diminishes when the volume is high. Also, the effect of expected changes on transaction volume is more dominant than that of unexpected changes. Explicitly accounting for oil spill closures and the Individual Fishing Quotas (IFQ) program in other species as variance shift parameters significantly reduces volatility and improves the market efficiency response to shocks. Published by Elsevier B.V.
1. Introduction The fishing industry provides some important commodities and this paper studies the price and volume of red snapper and the effects of environmental factors, quotas and other government regulations on their price and supply. In particular, we examine price of red snapper, which is an important fish commodity commonly found in the Gulf of Mexico (GOM). The price and trading volume of red snapper reflect production uncertainty and patterns of volatility that have similarities with those in well-organized financial markets and are interestingly enough related but different to the usual assets considered in finance. However, previous work by Chambers and Bailey (1996) and Deaton and Laroque (1992) have presented theories of commodity prices that imply a positive correlation between price and volatility that is similar to other financial returns. Studies examining fish commodity prices and market volatility are very limited. However, Bose (2004) finds persistent volatility in the prices of several fish species in Australia and a negative correlation between the variance of prices and trade volumes. While Oglend and Silkveland (2008)document evidence of increased volatility when prices for Norwegian salmon are high. The
☆ The authors are grateful for helpful guidance from Richard Baillie, one of the Journal editors. We are also thankful to Jessica Stephen and Andy Strelchek of NOAA for the data. The opinions expressed herein are those of authors and do not necessarily reflect the views of NOAA. ⁎ Corresponding author at: U.S. Department of Commerce, Southeast Fisheries Science Center, NOAA,75 Virginia Beach Dr., Miami, FL 33149, United States. E-mail address: [email protected] (A. Marvasti).
http://dx.doi.org/10.1016/j.jempfin.2016.07.008 0927-5398/Published by Elsevier B.V.
356
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
authors also use trade volume as an exogenous variable in the variance equation and find a negative correlation between trade volume and price volatility, which they attribute to the availability of inventory. In this paper, we examine the pricing and volatility of red snapper and how they are affected by some relevant exogenous variables. For this purpose, we apply the standard GARCH model and its extensions to the log of the daily price of red snapper. Our models are similar to those employed by Baillie and Myers (1991) for a variety of commodities; but we also include volume of trading and dummy variables to investigate the impact of various exogenous shocks. We also examine nonlinearity in the dynamic of red snapper prices and investigate the asymmetry in the effects of exogenous shocks to the red snapper fishing industry. In particular we assess the effects of oil spill shocks to the red snapper commodity. The rest of this paper is structured as follows: Section 2 provides institutional background and briefly describes the red snapper individual fishing quota (IFQ-RS) program. Section 3 discusses the stylized facts of price and other variables, as well as their dynamic properties. Section 4 presents various econometric models to which we apply variations of the GARCH model and discusses the empirical results. Section 5 summarizes the key findings and concludes. 2. Red snapper IFQ system and other regulatory controls As one of the most important fishery species in the GOM, red snapper stocks have been subjected to quota management since 1990 with the establishment of a total allowable catch (TAC), allocating 51% of the quota to the commercial sector and the remaining 49% to the recreational sector. The commercial red snapper quota was reduced in 1991 over concerns of overfishing. This quota reduction brought about some fishery closures and also market gluts as fishermen attempted to harvest as much as they could before the quota was reached and the fishery closed; see Waters (2001). Subsequent management measures, including an endorsement system in 1993 that was later converted to a two-tier license limitation system in 1998, a 10-day open season each month, and a quota increase in 1996, alleviated the derby fishing problem. Nonetheless, dockside prices remained low until January 1, 2007, when the IFQ-RS program was implemented, replacing the two-tier license limitation program. This program effectively expanded the commercial harvest season length from approximately 85 days a year to year-round. Since then, prices appear to have stabilized at higher levels, consistent with the general expectation from the IFQ-RS program, and the derby fishing problem has been essentially eliminated.1 Independent of the IFQ program, the commercial TAC has continued to change based on the status of the stock, fluctuating between 2.297 million pounds (gutted weight) in 2007 to 3.713 million pounds in 2012 (Fig. 1). For example, the TAC was reduced by 23% to address continuing overfishing in 2008 and was maintained at that level until the end of 2009. However, the typical short-term market gluts associated with the fishing derby did not recur. Although red snapper prices rose during this period in reaction to the reduced quota, the price reaction was not as severe as in the early 1990s. The industry has recently gone through two other notable regulatory changes. In January 2010, the IFQ-RS program was complemented by an IFQ program for the grouper-tilefish species (IFQ-GT), many of which are harvested with red snapper. The IFQ-GT program appears to have contributed to a geographic expansion of red snapper landings by allowing the trading of share allocations between these two programs and may possibly have affected price volatility for red snapper. This geographic expansion has been aided by the rebuilding of red snapper stocks in West Florida, as well as a generally increasing population. The IFQ-GT is thus believed to have led to an increase in both the diversity of the catch composition in the GOM and the number of vessels harvesting red snapper (NOAA, 2013). Another notable event for the industry was the April 20, 2010 Deepwater Horizon oil spill, which was one of the largest in U.S. history. The extent of the potential damage to the fishery was a serious environmental disaster causing the government to close various areas of the GOM to fishing between May 2 and December 30, 2010. This paper estimates the average percentage of the GOM area that was closed for each day of that period due to the Deepwater Horizon oil spill (Fig. 2). The peak closure was reached on June 3, 2010, when approximately 36% of the GOM was closed to fishing. Evidence in this paper indicates that the volatility of red snapper prices is directly affected by the oil spill over the regular market volatility. 3. Data and properties of the time series Our analysis is based on daily red snapper prices during the period of January 3, 2007 through September 30, 2015, which are obtained from the National Oceanographic and Atmospheric Administration (NOAA).2 Since 2007, the commercial red snapper fishing season has been open throughout the year; although the actual landings under the IFQ-RS program started on January 3, 2007. The IFQ-RS monitoring system electronically records all daily commercial red snapper landings and the corresponding (ex-vessel) price per pound. Under the red snapper IFQ program, commercial vessels may land at any time; but fish may not be offloaded between 6 P.M. and 6 A.M. A landing transaction is then completed by the IFQ dealer recording the date, location of transaction, weight and actual ex-vessel value of fish landed. The number of dealers purchasing red snapper throughout the 1 Initially, the IFQ shares were issued based on the number of red snapper landings reported under each participant's license during a specific time period. IFQ shares for Class 1 license holders were based on landings in the best ten consecutive years in 1990–2004. For historical captain's license holders, the IFQ shares were based on seven years of landings in 1998–2004. For Class 2 license holders, the IFQ shares were based on landings in the best five years in 1998–2004. A total of 546 entities, natural or juridical persons, qualified for the initial IFQ shares, in which 0.0001% was the lowest share issued and 6.0203% the highest. The number of shareholders has dropped since the beginning of the program, from 554 to 407 in 2012. 2 Daily prices have only been available since the institution of the IFQ-RS program in 2007.
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
357
Fig. 1. Commercial Quota for GOM Red Snapper (Source: NOAA, 2013).
Fig. 2. Percentage of GOM Closed Due to the DWH Oil Spill.
GOM since the inception of the red snapper program has varied between 66 and 96. Currently there are 316 IFQ dealer's facilities in the GOM. Therefore, each dealer may have multiple facilities for landing and/or the processing of fish. The distribution of the facilities and dealers tends to be proportionate to the length of the coast lines in each state. Accordingly, Florida, which has the longest coast line in the GOM, holds approximately 73% of the facilities. While most dealers operate in only one state, a few have facilities in more than one state. Most dealers purchase fish from 3 to 10 different fishermen (NOAA, 2013). The daily price used in this study is the average of recorded transactions by all dealers in the GOM market. Therefore, the composition of the dealers in each day can differ as the fish stock availability can change with water temperature, currents, and availability of baits. There were 131 days in the sample period when no data is available and is either due to no landings occurring on these days, or because dealers waited until the following day to report the transactions. This leaves a sample size of 3062 days with a mean price of $3.96 per pound and a variance of $0.53. The distribution of the price is slightly skewed to the left (see Fig. 3). Since the log of the series turns appears to be stationary, the unconditional long-run variance of the price is constant. Similar to the price series, the distribution of the landings is skewed negatively. This suggests that during the high season, when the stocks are plentiful, the average landing is typically high, while the fishing trips are less productive during the low season. Fig. 4 shows the log of landings (volume), which also shows a slight upward trend. The depictions of both the price and landings data show that there is more volatility below the trend than above it. Descriptive statistics for the log of price and other variables are in Table 1. The stylized facts of the log of the red snapper price series in terms of volatility clustering and fat tails are less severe than most other economic and financial time series. We also explore the effects of some exogenous variables on the volatility of red snapper prices using NOAA data to study the effects of quotas, the Deepwater Horizon oil spill closure, and the volume of transactions (landings). We considered two variations of the quota data: either the log of the series or a dummy for the quota reduction that takes the value of one when the quota is reduced relative to its previous level and zero otherwise.3 As a result of the 2010 oil spill, rather large areas of the GOM were closed to fishermen. We use a dummy variable capturing the significance of the oil closure event in our model; the variable takes the value of one during the closure period and zero otherwise. We also consider the average percentage of the GOM
3 We did not consider a quota increase dummy as well because it turns out be very similar, though inversed, to the dummy for quota reduction, except that it lasts six months.
358
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
Fig. 3. Log of Red Snapper Price.
Fig. 4. Log of Landings (Volume).
area closed each day as an alternative to the dummy variable. Finally, a dummy variable is created to represent the introduction of the IFQ-GT; this variable is equal to one for the time period after January 2010 and zero otherwise. Contrary to most financial assets, red snapper prices appear to be stationary based on the Augmented Dickey Fuller (ADF) test results. The log of quota and log of oil spill area closures appear to be I(1) series and are first differenced. Bai Perron tests are applied to test for structural breaks in the log of price of red snapper.4 The sup F statistic is 23.37 and WD max F is 38.69; both are significant at the 0.01% level. One of the relevant breaks is on May 29, in 2010, which is 27 days after the start of the oil spill, when more than 24% of the GOM was closed to fishing. The actual peak of the closure is on June 3, which is within the 95% confidence interval of this break.
4. Modeling the dynamics of prices Based on the unit root test results, we model the log of price as yt = lnPt and investigate the dynamics of the price series. The autocorrelations and partial autocorrelations of the red snapper series indicate quite persistent autocorrelations, while the partial autocorrelations are not significantly different from zero after lag seven. The use of BIC suggests that an AR(7) model is the most appropriate description of the conditional mean. Specification tests reveal the presence of ARCH effects, which led to the following lagged dependent variable regression:
yt ¼ ϕ0 þ
7 X
GT
ϕk yt−k þ λ1 vt þ λ2 Q Rt þ λ3 Ot þ λ4 IFQ t þ εt ;
ð1Þ
k¼1
4 The following Bai-Perron multiple structural change test is applied for the existence, number, and timing of breaks in the price volatility: yt =x′β+z ′δ t t j +μt, where t=Tj−1 +1, . . . ,Tj for j=1, . . . ,m+1. Here yt is the dependent variable, xt(p...1) and zt(q..1) are vectors of covariates, β and δj(j=1, ...,m+1) are coefficient vectors, and μt is the error term. More detailed information can be found in Bai and Perron (2003).
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
359
Table 1 Descriptive Statistics. Variables
Mean
Standard deviation
Skewness
Kurtosis
J.B. statistics
Log of red snapper price Log of volume (gutted landings in pounds) Log of oil closure days Log of quota Quota reduction dummy (Reduction = 1, Otherwise = 0) IFQ-GT introduction (IFQ-GT Present = 1, Otherwise = 0)
1.36 8.66 1.92 15.01 0.27 0.67
0.19 1.34 1.27 0.31 0.44 0.47
−0.79 −1.36 3.40 0.38 1.07 −0.71
7.10 5.99 13.18 2.11 2.14 1.50
2463.71 (0.00) 2585.93 (0.00) 19,131.70 (0.00) 175.33 (0.00) 676.12 (0.00) 542.14 (0.00)
Sample: January 3, 2007 to October 30, 2015 (N = 3193).
q p r X X ε X ε 2 2 log δt ¼ ω0 þ β j log δt− j þ α i t−i þ γ k t−k þ θ1 V t þ δ δt−k t−i j¼1 i¼1 k¼1
ð2Þ
θ2 QRt þ θ3 Ot þ θ4 IFQ t þ νt :
The additional variables in this model include Vt, the transaction volume; QRt, a dummy variable representing the quota reduction; Ot, the log of the percentage of the daily GOM oil spill closure; and IFQGT, a dummy variable for the introduction of the grouper-tilefish IFQ program. In contrast to the mean equation, there are no a priori sign expectations for the effects of the exogenous variables in the variance equation. After testing different error distribution assumptions, we select a normal error distribution model as the best fit. Various ARMA-GARCH models are then estimated using quasi-maximum likelihood estimation (QMLE) as in Bollerslev and Wooldridge (1992). The estimated models are presented in Table 2. Our estimate of the presence of volatility depends on whether explanatory variables are included. Before adding any exogenous variables to the basic model, we test the asymmetric effect of innovations on the conditional variance. When, the Engle and Ng (1993) leverage effect test is applied, the t-value for the lagged residuals turns out to be negative and statistically highly significant. This suggests that the sign of the current period shock provides helpful information in predicting conditional volatility. We then estimate the basic EGARCH model with asymmetric effects, finding that the parameter estimates suggest a high degree of persistence in the conditional variance (The results are omitted for brevity.). Price volatility persistence is not uncommon in fishery: Oglend and Silkveland (2008) estimate the degree of volatility persistence in the salmon market at 0.81, while Buguk et al. (2003) arrive at a degree of volatility persistence of 0.98 for the catfish market and 0.38 for the menhaden market. We then proceed to enhance our basic model by including volume, which reduces volatility.5 The results are reported in Table 2 as model 1. The response to new information, α1, has increased, while volatility decay speed, β1, is significantly diminished relative to the basic (unreported) model. Unlike some commodity studies (e.g. Clark, 1973; Lamoureux and Lastrapes, 1990; Fleming et al., 2006), the GARCH coefficient also remains statistically significant. Typically, a larger volume of transactions tends to reduce variance in financial data (e.g., Deaton and Laroque, 1992; Chambers and Bailey, 1996; Bessembinder and Seguin, 1993; Fleming et al., 2006). Clark (1973) argues that volume and volatility tend to be very close or the same. We find trade volume to be negatively correlated with the variance of red snapper price. Oglend and Silkveland (2008), in their analysis of salmon price volatility behavior, attribute the negative correlation between trade volume and price volatility to the availability of inventories, which might also be a valid argument here. The authors argue that large volume increases inventories, thereby smoothing out price fluctuations. The coefficient of γ1 suggests that the leverage effect is present in the conditional variance. In other words, the impact of news is not symmetric, and there is a tendency for more of an increase in volatility following positive shocks than negative shocks of the same magnitude, unlike the typical asymmetric results with financial assets. Model 2 considers the effect of trade volume on price by separating volume into expected (θE1) and unexpected (θU 1 ) components. The model for expected volume is based on daily seasonal dummies and a linear time trend. The estimated AR-EGARCH model with predicted and unpredicted trade volume measures is presented, with the results showing a negative correlation in the conditional variance equation for both components of trade volume. Surprises in trading volume have a larger effect on price volatility than the forecastable component, which is consistent with the findings of Bessembinder and Seguin (1993), who also show both components of volume have a positive effect on price volatility. However for the red snapper series in this study there appears to be a negative effect; which is consistent with the findings of Oglend and Silkveland (2008). In our result, the unanticipated change in trade volume that will move the red snapper price by one unit is approximately 0.45, which is within the market depth range for financial assets reported by Bessembinder and Seguin (1993). The effects of quotas and the oil spill closures are represented by dummy variables, which together with IFQ-GT are included as explanatory variables in the conditional mean and variance equations of the estimated models presented as model 3 in Table 2. The addition of these variables reduces the skewness and kurtosis of the residuals and improved the fit of the model. We discover that quota reductions have no statistically significant effect on the mean price, while they have a positive effect on the variance. 5 A high degree of volatility persistence may be due to failure to account for the presence of structural changes, as has been argued by Engle and Bollerslev (1986) and Gray (1996). The topic has been further discussed and investigated in several empirical studies (e.g., Lamoureux and Lastrapes, 1990; Simonato, 1992; Hillebrand, 2005; Galeano and Tsay, 2010). In our study, the inclusion of red snapper trading volume in the variance function is likely to tone down the degree of volatility persistence in the price.
360
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
Table 2 AR(7)-EGARCH Models with Asymmetry Parameter Estimates with Trade Volumes, and Other Exogenous Variables. Parameters
Model 1
Model 2
Model 3
Model 4
Model 5
0.146⁎⁎⁎ (0.017) 0.086⁎⁎⁎ (0.016) 0.056⁎⁎⁎
0.120⁎⁎⁎ (0.016) 0.092⁎⁎⁎ (0.016) 0.075⁎⁎⁎
0.119⁎⁎⁎ (0.017) 0.059⁎⁎⁎ (0.015) 0.041⁎⁎⁎
0.118⁎⁎⁎ (0.017) 0.058⁎⁎⁎ (0.015) 0.041⁎⁎⁎
0.122⁎⁎⁎ (0.017) 0.058⁎⁎⁎ (0.015) 0.038⁎⁎⁎
ϕ6
(0.016) 0.108⁎⁎⁎ (0.015) 0.077⁎⁎⁎ (0.016) 0.131⁎⁎⁎
(0.015) 0.099⁎⁎⁎ (0.014) 0.090⁎⁎⁎ (0.016) 0.114⁎⁎⁎
(0.016) 0.091⁎⁎⁎ (0.015) 0.0617⁎⁎⁎ (0.015) 0.108⁎⁎⁎
(0.016) 0.090⁎⁎⁎ (0.015) 0.062⁎⁎⁎ (0.015) 0.108⁎⁎⁎
(0.016) 0.090⁎⁎⁎ (0.015) 0.060⁎⁎⁎ (0.015) 0.107⁎⁎⁎
ϕ7
(0.017) 0.190⁎⁎⁎
(0.017) 0.178⁎⁎⁎
(0.017) 0.160⁎⁎⁎
(0.017) 0.159⁎⁎⁎
(0.017) 0.160⁎⁎⁎
λ2
(0.018) –
(0.018) –
λ3
–
–
(0.017) 0.001 (0.007) –
λ3 ,t+56
–
–
(0.017) 0.001 (0.007) −0.093⁎⁎⁎ (0.016) –
(0.017) 0.007 (0.006) −0.085⁎⁎⁎ (0.015) –
λ4
–
–
Mean Equation: ϕ1 ϕ2 ϕ3 ϕ4 ϕ5
−0.099⁎⁎⁎ (0.017) 0.053⁎⁎⁎
–
–
0.054⁎⁎⁎ (0.007) –
θ1
0.254⁎⁎⁎ (0.055) 0.279⁎⁎⁎ (0.063) −0.423⁎⁎⁎
0.442⁎⁎⁎ (0.073) 0.195⁎⁎⁎ (0.107) –
0.173⁎⁎⁎ (0.054) 0.209⁎⁎⁎ (0.062) −0.437⁎⁎⁎
0.176⁎⁎⁎ (0.053) 0.206⁎⁎⁎ (0.062) −0.437⁎⁎⁎
0.158⁎⁎⁎ (0.055) 0.215⁎⁎⁎ (0.063) −0.434⁎⁎⁎
θE1
(0.027) –
(0.026) –
(0.027) –
(0.025) –
θU 1
–
–
–
–
θ2
–
θ3
–
−
0.630⁎⁎⁎ (0.125) 0.346⁎⁎⁎
0.647⁎⁎⁎ (0.127) –
0.626⁎⁎⁎ (0.122) −0.330⁎⁎⁎
θ3,t+56
–
−
θ4
–
–
γ1
0.148⁎⁎⁎ (0.031) 1531.82 −0.97 0.04 (prob. 0.85)
0.234⁎⁎⁎ (0.076) 1784.75 −1.13 0.44 (prob. 0.51)
δ2t Variance equation: α1 β1
Log Likelihood BIC ARCH LM F-stat.
−0.383⁎⁎⁎ (0.062) −0.445⁎⁎⁎ (0.108) –
(0.144) – 0.881⁎⁎⁎ (0.113) 0.122⁎⁎⁎ (0.032) 1662.68 −1.04 0.79 (prob. 0.37)
(0.006) –
0.059⁎⁎⁎ (0.007) −0.460⁎⁎⁎ (0.149)
0.404⁎⁎⁎ (0.151) 0.887⁎⁎⁎ (0.114) 0.124⁎⁎⁎ (0.032) 1667.19 −1.04 0.82 (prob. 0.37)
(0.141) – 0.876⁎⁎⁎ (0.112) 0.116⁎⁎⁎ (0.033) 1665.72 −1.04 1.43 (prob. 0.23)
Note: The error term is assumed to follow normal distribution. Heteroskedasticity consistent standard errors are in parentheses; ⁎⁎⁎ denotes significance at 1%; degrees of freedom (d.f.) is the estimated tail spread of the t distribution.
The oil spill closures, on the other hand, have reduced the mean of red snapper price, perhaps by cutting supply, but the probable level of anxiety they caused appears to have caused an increase in price volatility. Finally, the introduction of the IFQ-GT has both increased the mean and price volatility. This may be due to interactions between the species' landings and vessels/permits variables. We extend our investigation of the role of the oil spill event by applying the two relevant oil spill structural break dates identified in Section 3. Both the May 29 and June 3 breaks produced similar results to that of the initial oil spill closure dummy variable, though the magnitude of the coefficients for this variable increase slightly in both the mean and the variance equations. For brevity, only the results from the May 29 break are reported as model 4.6
6 The log difference of the average percentage of the GOM area that was closed each day, as an alternative measure of the oil spill, showed a positive effect on the mean of price, but no effect on the variance.
A. Marvasti, A. Lamberte / Journal of Empirical Finance 38 (2016) 355–361
361
The effect of changes in the price risk of red snapper on its mean price is tested by including the variance of the log of GARCH in the mean equation in model 5. It turns out that changes in the riskiness of purchasing red snapper have a negative effect on the mean, similar to the effect of risk in financial assets. 5. Summary and conclusion This study provides a better understanding of the price risk that fishermen typically face, affecting their profitability and possibly their investment decisions. Analysis of price volatility is also important to the retail seafood market and policymakers. We examine the behavior of red snapper dockside prices for the first few years of the IFQ system for this fishery. The autocorrelation function suggests stationarity and highly seasonal behavior. However, other stylized facts of the log of the red snapper price series are similar to most other economic and financial time series in terms of volatility clustering, fat tails, and volatility mean reversion. We employ an array of GARCH models; and conclude that the EGARCH model describes the dynamics of red snapper price better than does a standard linear model, but could not reject the presence of asymmetry in the red snapper price response to shocks. Analyses of financial data have typically found a link between the volume of transactions and price volatility. We find that red snapper price volatility is diminished when trade volume is high, which is contrary to the findings from financial assets, but is consistent with the behavior of fish commodity prices. However, the market depth is not significant when other exogenous variables are absent in the model. In other words, changes in the volume of transactions do not cause as much change in price volatility as they do in financial asset prices. We can only speculate whether the association between red snapper price volatility and volume is due to inventory changes, geographic expansion, or increase in the number of fishermen landing red snapper on highvolume days. Using the regularity pattern for transaction volume during the week, we separate the expected changes from the unexpected changes in the volume of transactions. It turns out that while transaction volume provides useful information regarding price volatility, the effect of expected changes in transaction volume is more pronounced than that of unexpected shocks. We also find that reductions in quota have increased the red snapper prices volatility. On the other hand, the 2010 oil spill closures in the GOM have had a dampening effect on the conditional mean while increasing price volatility. Furthermore, the debut of the closely related IFQ-GT program in 2010 is associated with an increase in red snapper price volatility. Finally, we discover that volatility has not been constant over the study period and that there is a positive correlation between price and volatility, as measured by the GARCH variance. References Bai, J., Perron, P., 2003. Computation and analysis of multiple structural change models. J. Appl. Econ. 18, 1–22. Baillie, R.T., Myers, R.J., 1991. Bivariate GARCH estimation of the optimal commodity futures hedge. J. Appl. Econ. 6, 109–124. Bessembinder, H., Seguin, P.J., 1993. Price volatility, trading volume, and market depth: evidence from futures markets. J. Financ. Quant. Anal. 28, 21–39. Bollerslev, T., Wooldridge, J.M., 1992. Quasi-maximum likelihood estimation and inference in dynamic models with time varying covariances. Econ. Rev. 11, 143–172. Bose, S., 2004. Price volatility of south-east fishery's quota species: an empirical analysis. Int. Econ. J. 18, 283–297. Buguk, C., Hudson, D., Hanson, T., 2003. Price volatility spillover in agriculture: an examination of U.S. catfish markets. J. Agric. Resour. Econ. 28, 86–99. Chambers, M.J., Bailey, R.E., 1996. A theory of commodity price fluctuations. J. Polit. Econ. 104, 924–957. Clark, P.K., 1973. A subordinated stochastic process model with finite variance for speculative prices. Econometrica 41, 135–155. Deaton, A., Laroque, G., 1992. On the behaviour of commodity prices. Rev. Econ. Stud. 59, 1–23. Engle, R.F., Bollerslev, T., 1986. Modelling the persistence of conditional variances. Econ. Rev. 5, 1–50. Engle, R.F., Ng, V., 1993. Measuring and testing the impact of news on volatility. J. Financ. 48, 1749–1778. Fleming, J., Kirby, C., Ostdiek, B., 2006. Stochastic volatility, trading volume, and the daily flow of information. J. Bus. 79, 1551–1590. Galeano, P., Tsay, R.S., 2010. Shifts in individual parameters of a GARCH model. J. Financ. Economet. 8, 122–153. Gray, S.F., 1996. Modeling the conditional distribution of interest rates as a regime-switching process. J. Financ. Econ. 42, 27–62. Hillebrand, E., 2005. Neglecting parameter changes in GARCH models. J. Econ. 129, 121–138. Lamoureux, C., Lastrapes, W., 1990. Persistence in variance, structural change, and the GARCH model. J. Bus. Econ. Stat. 8, 225–233. National Oceanic and Atmospheric Administration (NOAA), 2013. Red Snapper Individual Fishing Quota Program 5-Year Review. Southeast Regional Office, National Marine Fisheries Service, St. Petersburg, Florida. Oglend, A., Silkveland, M., 2008. The behaviour of salmon price volatility. Mar. Resour. Econ. 23, 507–526. Simonato, J.G., 1992. Estimation of GARCH process in the presence of structural change. Econ. Lett. 40, 155–158. Waters, J.R., 2001. Quota management in the commercial red snapper fishery. Mar. Resour. Econ. 16, 65–78.