Estimating the impact of COVID-19 pandemic on alternative semi-intensive shrimp (Penaeus vannamei) production schedules in Mexico: a stochastic bioeconomic approach.

This study uses a stochastic bioeconomic approach to estimate the COVID-19 pandemic economic impact on shrimp farming in Mexico. Seeding-harvesting schedules - March-June, May-August, and August-November - were analyzed using shrimp prices and production costs corresponding to 2017-2019 (pre-pandemic) and 2020 (pandemic). The analyses estimated net revenue varied within 597.97-2758.88 USD$ ha-1 and 1262.40-1701.32 USD$ ha-1 under the pre-pandemic and pandemic scenarios, respectively. Significant decreases (38%) were estimated in net revenue values in March-June and May-August under the pandemic scenario. However, probability distributions estimated that uncertainty on the expected net revenues was not affected by the pandemic conditions, and the probability of losing was null or negligible in all the cases. Unfavorable conditions under the pandemic also required significantly higher break-even production for March-June (25.7%) and May-August (28.5%) schedules. The cost of post-larvae was the most important economic factor influencing net revenue. To conclude, although the operating conditions during the pandemic were conducive to worsening the economic outcome, no evidence still exists that uncertainty and economic risk increased compared with pre-pandemic conditions.

Introduction

COVID-19 pandemic (from now on, “pandemic”) has been causing major problems worldwide and significant damage to the global economy. The seafood sector, like most industries, has to deal with a negative outlook on demand and several obstacles in supply (INFOPESCA, 2021). Global shrimp and salmon production was expected to be the most affected by the pandemic (FAO and Eclac, 2020).

Several reports have been prepared regarding how the pandemic has affected aquaculture activities worldwide. Kakoolaki et al (2020) estimated its potential socio-economic effects on the world shrimp aquaculture sector during its early stages. Waiho et al. (2020) evaluated the pandemic impact on Malaysia aquaculture sector and discussed the potential effectiveness of the coping strategies implemented. The pandemic impact in Bangladesh was assessed by Rafiquzzaman (2020). Islam et al. (2021) discussed the case of Bangladesh to suggest a set of immediate and long-term changes to policy and plans to recover aquaculture and fishery sectors from the pandemic. Lebel et al. (2021) studied the effects of public health responses to the pandemic on aquaculture farmers in the Mekong Region. A sectoral assessment of the pandemic impact on aquaculture in India was conducted by Kumaran et al. (2021). Akter and Khan (2021) showed shrimp farming as a case study of the pandemic economic impacts on southwestern Bangladesh.

According to Betanzo-Torres (2020), fish farmers and seafood traders in Mexico reported that sales during Easter week (April 5–11, 2020) were down 60% compared with the previous year and relatively low prices due to low demand and tourism activity shutdown. Martínez-Cordero et al. (2020) surveyed tilapia producers in Mexico, indicating that most of the farmers scheduled harvesting on Easter and Easter week during the pandemic, but sales dropped 50% on average, and tilapia prices decreased about 15–17% due to the confinement.

The reports mentioned above have been prepared based on surveys of the regional or sectoral impact of the pandemic on aquaculture, but estimations of the quantitative economic impact caused at the individual farms level have remained unaddressed. This study focuses on assessing such impact on the typical shrimp production of farmers operating in the state of Nayarit, Mexico, which, together with the states of Sinaloa and Sonora, constitute the most important shrimp producers in Mexican aquaculture CONAPESCA Comisión Nacional de Acuicultura y Pesca (2018). A bioeconomic approach was used for evaluating the pandemic impact from a detailed dynamical perspective rather than an overall static description.

The stock model presented by Ruiz-Velazco et al. (2013) and Estrada-Pérez et al. (2015) was used to analyze the seeding-harvesting schedules: March–June, May–August, and August–November. Surveys among shrimp farmers, post-larvae hatchery managers, and feed manufacturers were conducted to update production costs and shrimp prices before (2017, 2018, and 2019) and during (2020) the COVID-19 pandemic. In addition, stochastic elements were incorporated into the analyses to account for the uncertainty and economic risk involved in the expected outcomes. To our knowledge, no such study has been previously conducted and reported in the literature.

Materials and methods

Database

Primary production data from semi-intensive shrimp farms were used for stock model calibration as described by Ruiz-Velazco et al. (2013) and Estrada-Pérez et al. (2015). Table 1 summarizes the mean values of the general operating conditions.

Table 1

Mean (± standard deviation) values of rearing conditions and production variables of semi-intensive shrimp Penaeus vannamei farming during alternative seeding-harvesting schedules

Variable	Schedule
	March–June	May–August	August–November
Temperature (°C)	31.02 ± 0.63	31.98 ± 0.15	31.14 ± 0.68
Dissolved oxygen (mg L−1)	4.86 ± 0.75	3.90 ± 0.23	4.33 ± 0.39
Stocking density (post-larvae m−2)	15.07 ± 1.53	26.22 ± 5.67	20.14 ± 5.24
Rearing duration (weeks)	10.50 ± 1.10	11.94 ± 1.13	12.86 ± 1.77
Yield (t ha−1)	1.13 ± 0.12	1.80 ± 0.48	1.6 ± 0.37

Stock model

The model proposed by Ruiz-Velazco et al. (2013) predicts shrimp biomass as a function of time as the product of shrimp individual mean weight and surviving shrimp population. The following equation in the model by these authors serves to predict shrimp growth:

where w is the mean individual weight predicted after t time units; w is initial weight; w is final weight; k is a growth coefficient; and h is the number of time units at harvest time.

On the other hand, the authors use the negative exponential equation to calculate surviving shrimp population (n):

where is initial population and z is the instantaneous mortality rate.

Multiple regression equations and stochastic elements

Using multiple regression analyses, Ruiz-Velazco et al. (2013) established relationships between the stock model parameters and water quality and management. This investigation incorporates the residual errors obtained from those analyses as stochastic elements in the bioeconomic model. Table 2 shows the equations and error distributions obtained from the analyses.

Table 2

Multiple regression equations used to predict the values of the stock model parameters as a function of water quality and management variables. Normal distributions (nml) with zero mean values were used to generate error values (ε) for parameters and variables. The standard deviations of the distributions are within parentheses. A distribution was fitted to coefficient k values directly after no significant result was obtained from the corresponding regression analysis

Equation	Error distribution
Stock model parameters:
wf =  − 25.2436 + 0.9678 T + 0.5329 RD + ε1	ε1 = nml (1.39)
k = nml (0.83, 0.06)
z = 0.0916 − 0.0081 DO + 0.0010 D − 0.0034 RD + ε2	ε2 = nml (0.01)
Water quality and management variables:
T = 0.049 D + 30.36 + ε3	ε3 = nml (0.61)
DO = 6.51 − 0.036 D − 0.1175 RD + ε4	ε4 = nml (0.60)

T mean pond water temperature, RD rearing duration, DO dissolved oxygen, D stocking density

The multiple linear regression procedure available in Stata 17 (Stata Corp, College Station, TX) was used for the analyses, which deals automatically with multicollinearity according to the methods described by Rencher (2002). Tests for violation of linear regression assumptions available in Stata 17 indicated data adequacy.

Feed conversion ratio model

The feed conversion ratio (FCR) showed an increasing linear trend throughout the rearing period, and the linear equation predicted it as a function of time (t) according to:

where a and b are regression coefficients whose values are in turn calculated using the multiple regression equation:

and,

where T is pond water temperature; nml is the normal distribution with mean and standard deviation within parentheses; and ε5 = nml (0.0, 0.05).

Economic model

This model calculates net revenue (nr, in USD$ with 19.13 MXN per USD$ exchange rate) as a function of time (t) according to:

where income (i) is the product of shrimp biomass from the stock model and farm-gate shrimp price, and c are the costs considered for analysis, namely:

where cfeed is the cost for feed; cfer for fertilizer; ce for energy; cPL for post-larvae; cpp for pond preparation; cl for labor; cma for maintenance; cmi for miscellaneous; and ch for harvesting. The benefit/cost ratio (B/C) was also calculated as: . Sale shrimp prices and unitary costs were obtained from personal communications (Tables 3, 4, and 5). The same parameter values of the stock model, regression equations, and FCR model were used for 2017–2019 and 2020.

Table 3

Shrimp sale price mean values (USD$ kg−1) used for analyses during pre-pandemic and pandemic scenarios

Shrimp size (g)	Pre-pandemic	Pandemic
7	3.53	3.66
8	3.44	3.44
9	3.72	3.51
10	3.90	3.84
11	4.10	3.73
12	4.35	3.93
13	4.34	4.16
14	4.29	4.20

Shrimp prices were obtained by personal communications (Encarnación Torres, Productora y Comercializadora de Camarón, J. Rangel Becerra Zepeda, December 2020)

Table 4

Unitary cost estimates. Shrimp post-larvae and feed costs were the only ones differing among the pre-pandemic and pandemic operating conditions

Costs
Post-larvae (USD$ thousand−1)	4.50 (pre-pandemic), 5.34 (pandemic)
Feed (USD$ kg−1)	0.82 (pre-pandemic), 0.92 (pandemic)
Fertilizers (USD$ kg−1)	0.95
Energy (USD$ kwh−1)	0.15
Pond preparation (USD$ ha−1 yr−1)	115.38
Labor (USD$ ha−1 yr−1)	2352.00
Maintenance (USD$ ha−1 yr−1)	481.15
Miscelaneous (USD$ ha−1 yr−1)	157.31
Harvesting (USD$ shrimp kg−1)	0.10

Post-larvae and feed prices were obtained by personal communications (Gonzalo Abundis, Policultivos Intensivos de Nayarit, February 2021) (Venancio Torres, Alimentos Balanceados Camaronay, March 2021)

Table 5

Minimum (Min), maximum (Max), and most likely (ML) program evaluation and review technique (PERT) distribution parameter estimates, and error normal distributions (nml: mean, standard deviation) used for stochastic values calculation of feed and post-larvae unitary costs and shrimp sale prices during the pre-pandemic and pandemic scenarios. ML = (6 mean-Min–Max)/4

	Pre-pandemic				Pandemic
Distributions	PERT			NORMAL	PERT
Prices	Min	Max	ML	nml	Min	Max	ML
Feed	0.79	0.86	0.81		0.82	0.94	0.93
Post-larvae				0.0045, 0.0005	0.0051	0.0056	0.0053
Shrimp (7 g)
M-J	3.02	3.77	3.40		3.00	3.70	3.34
M-A	3.54	3.78	3.66		3.51	3.70	3.63
A-N	Not available				Not available
Shrimp (8 g)
M-J	3.16	3.95	3.55		3.16	3.95	3.55
M-A	3.35	3.40	3.38		3.35	3.38	3.36
A-N	3.23	3.52	3.41		3.23	3.52	3.41
Shrimp (9 g)
M-J	3.16	3.95	3.59		3.46	3.65	3.55
M-A	3.71	4.39	3.97		3.24	3.55	3.40
A-N	3.43	3.73	3.62		3.50	3.66	3.59

Post-larvae, feed, and shrimp prices were obtained by personal communications: Gonzalo Abundis, Policultivos Intensivos de Nayarit, February 2021; Venancio Torres, Alimentos Balanceados Camaronay, March 2021; Encarnación Torres, Productora y Comercializadora de Camarón, J. Rangel Becerra Zepeda, December 2020

M-J March–June, M-A May–August, A-N August–November

Random variability in feed and post-larvae unitary costs and shrimp sale prices for the pre-pandemic scenario was incorporated as follows: (1) regression analyses served to determine whether ascending or descending trends existed during the period; (2) when no significant trend was detected (p 0.05), cost or price values were pooled, and a Program Evaluation and Review Technique (PERT) distribution was used with the minimum, maximum and most likely parameters estimated from the pooled set values; (3) when a significant trend was determined, a normal error distribution resulting from the regression analysis was used to add random error values to cost or price (Table 4). Given that the pandemic scenario corresponded to a single year, random variability in prices was incorporated using PERT distributions as described above with the parameters estimated from data costs and prices collected during 2020 (Table 5).

Stochastic break-even analysis

The following formula calculates the production yielding break-even situations (BP) (Parkin, 1996):

where FC are the fixed costs; UI and UVC are the unit income (shrimp price/kg) and unit variable cost/kg of shrimp produced.

Monte Carlo simulation and statistical and sensitivity analyses

The bioeconomic model was programmed in Excel 2019, and probability distributions for nr and break-even production under the different scenarios were inferred with Monte Carlo simulation using @RISK v.8.1 (Palisade Corp., Ithaca, NY, USA). Following Ayyub (2014), the expected nr value measures average return, whereas the standard deviation measures dispersion in nr, reflecting uncertainty associated with the outcome. According to the author, the coefficient of variation (ratio standard deviation/mean) is thus a measure of uncertainty per unit value of the expected nr. This study adopts such a coefficient (CV) as an uncertainty indicator. Zar (2010) explains the t-test and Z-test used in this investigation to determine whether the pandemic situation significantly affected the economic benefits (mean nr), uncertainty (CV), and break-even production (BP) when compared with the pre-pandemic situation, setting significance at α = 0.05. On the other hand, the loss probability estimated from the nr probability distribution served as an economic risk indicator.

A sensitivity analysis was also conducted to estimate the contribution to uncertainty of the bioeconomic model parameters and water quality variables. This analysis used the multiple regression routine in @Risk 8.1 (Palisade, Ithaca, NY, U.S.A.) with “mapped” regression coefficients indicating the amount nr is modified per standard deviation unit of each model parameter, thus contributing to its uncertainty.

Results

According to net revenue values, the best harvesting times for the pre-pandemic period were estimated at the end of March–June and May–August schedules (i.e., 11 weeks; Fig. 1). However, during the pandemic period, the best harvesting time was 10 weeks in both schedules (Fig. 1). In August–November schedule, no differences were estimated regarding harvesting time (11 weeks) in both scenarios (Fig. 1). The B/C ratio was maximized simultaneously as nr, except for May–August and August–November schedules in the pre-pandemic situation.

Fig. 1

Mean net revenue, costs, income, and benefit–cost ratio (B/C) as a function of rearing time for the alternative seeding-harvesting schedules under the pre-pandemic (17–19) and pandemic (20) scenarios

Figure 2 shows the probability distributions of nr values inferred from the Monte Carlo simulation for the three schedules at their best harvesting times under the pre-pandemic and pandemic scenarios. Net revenue varied from 1597.97 to 2758.88 USD$ ha−1 and from 1262.40 to 1701.32 USD$ ha−1 under the pre-pandemic and pandemic scenarios. The t-test indicated that during the pandemic scenario, significant decreases (38%) were estimated in net revenue values in March–June and May–August schedules. In contrast, no significant differences existed in the August–November schedule between the pre-pandemic and pandemic conditions. On the other hand, the Z-test indicated that mean CV values did not vary significantly among the schedules, and the probability of losing was null or negligible in all the cases. The inferred probability distributions for the B/C ratio when its maximum values occurred indicated better results for the pre-pandemic period (1.46–1.69) than those resulting from the pandemic scenario (1.38–1.41).

Fig. 2

Probability distributions of net revenue for the alternative seeding-harvesting schedules under the pre-pandemic (17–19) and pandemic (20) scenarios. The values corresponding to the dashed lines indicate the 95% probability intervals. Different letters indicate significant differences among the scenarios for each schedule separately. SD standard deviation, CV coefficient of variation, LP loss probability

Table 6 summarizes the results obtained from the inferred probability distributions for BP. The pandemic conditions required a significantly higher mean BP for March–June (25.7%) and May–August (28.5%) schedules, while no significant difference existed for the August–November schedule. Obtaining mean BP, however, does not impede the possibility of losing. The probability of loss was very similar in all the cases and slightly varied around 55.8%. The probability distributions showed that to minimize loss probability to 5%, BPs 20.7 to 29.5% higher than the mean BP were necessary. Mean BP represented 23.6 to 36.8% of the total mean productive capacity, while the percentage corresponding to the BP required for 5% loss probability ranged from 29.3 to 45.8%. Except for the August–November schedule, the mean total production percentages required for mean BP and 5% loss probability were higher under the pandemic scenario.

Table 6

Stochastic break-even production analysis results for the alternative seeding-harvesting schedules under pre-pandemic and pandemic conditions. Different letters indicate significant differences for each schedule between both conditions

Schedule	Mean total production (kg ha−1)	Mean BP (kg ha−1)	LP for mean BP (%)	BP for 5% LP (kg ha−1)	MPP-MBP (%)	MPP-5% LP
March–June (pre-pandemic)	1114.3	326.5a	56.6	404.1	29.3	36.3
March–June (pandemic)	1114.3	410.4b	54.5	495.5	36.8	44.5
May–August (pre-pandemic)	1794.0	423.2a	56.4	526.4	23.6	29.3
May–August (pandemic)	1794.0	543.9b	55.7	694.0	30.3	38.7
August–November (pre-pandemic)	1400.2	495.0a	56.7	641.2	35.4	45.8
August–November (pandemic)	1400.2	486.1a	54.8	599.1	34.7	42.8

BP break-even production, LP loss probability, MPP-MBP mean total production percentage required for mean BP, MPP-5% LP mean total production percentage required for 5% LP

The sensitivity analysis indicated that shrimp final weight and mortality rate were the main parameters consistently affecting nr variability (Table 7). The post-larvae cost was the most notably economic parameter contributing to nr variance, while pond water temperature, dissolved oxygen, and feed conversion ratio showed intermediate importance. Shrimp sale prices and feed costs were of lower importance (Table 5).

Table 7

Results of the sensitivity analysis of net revenue to the stochastic variability of the bioeconomic model parameters for the alternative seeding-harvesting schedules under pre-pandemic and pandemic scenarios. The values indicate the net revenue/ha change per parameter standard deviation unit

Schedules
March–June				May–August				August–November
Pre-pandemic Parameter		Pandemic Parameter		Pre-pandemic Parameter		Pandemic Parameter		Pre-pandemic Parameter		Pandemic Parameter
Wf	423.7	Wf	344.8	Z	− 575.3	Wf	430.5	Z	− 392.8	Z	− 417.1
Z	− 409.0	Z	− 307.6	Wf	532.8	Z	− 418.4	Wf	376.9	Wf	401.4
cPL	− 224.3	T	165.15	aF	− 284.3	aF	− 250.5	cPL	− 224.2	aF	− 227.6
T	198.4	aF	− 156.9	T	255.8	T	211.4	aF	− 201.9	T	193.2
SP	166.5	DO	124.96	DO	234.7	DO	170.1	T	181.8	DO	169.2
DO	166.2	k	− 59.82	cPL	− 225.4	k	− 131.0	DO	160.1	bF	− 56.8

Wf shrimp final weight, k shrimp growth coefficient, Z shrimp mortality rate, T temperature, DO dissolved oxygen, aF and bF regression coefficients for FCR prediction, SP shrimp sale price, cfeed feed cost, cPL post-larvae cost

Discussion

This investigation indicates that the conditions prevailing during the pandemic impaired the economic performance of semi-intensive shrimp production during March–June and May–August 2020. Nonetheless, it is worth noting that despite the worsening economic conditions under the pandemic scenario, the probability of having losses was null or negligible.

Under the pandemic scenario in March–June and May–August 2020, the rearing time when maximum nr values were obtained showed that harvesting a week earlier would have been preferable compared with the pre-pandemic scenario 2017–2019. At the end of the production cycles, income diminished during the pandemic because shrimp prices decreased in June (3.7–6.6%) and August (2.4–14.7%) compared with the pre-pandemic operating conditions. In contrast, in the pandemic scenario, shrimp prices rose 7.3–7.8% during November, without affecting when maximum nr occurred because higher post-larvae (18.6%) and feed (12.2%) costs compensated for price increases.

The B/C ratio is a relative index that measures the efficiency of gaining per monetary unit invested, while nr is an absolute gain indicator, and their maximization may occur at different times during the production cycle. Such a case occurred in some pre-pandemic schedules, although the pandemic conditions were conducive to maximizing both indexes simultaneously.

The results showed that the conditions during the pandemic scenario did not lead to higher uncertainty (i.e., higher CV) or economic risk (i.e., loss probability) than the previous operating conditions. Nevertheless, the prevailing conditions during the pandemic impaired net revenues and B/C ratio. Shrimp prices diminished due to decreased demand for farmed shrimp, particularly during the first part of the pandemic (Solano, personal communication, 2021). Rafiquzzaman (2020), Islam et al. (2021), and Lebel et al. (2021) reported similar situations for Bangladesh and the Mekong Region, where the lockdown forced by the pandemic led to consumption and price reduction of aquaculture products. During the pandemic, the low market shrimp price has been deemed the major issue for shrimp farmers in Southwestern Bangladesh (Akter and Khan, 2021). However, this investigation found that during the pandemic, shrimp prices recovered in November and exceeded those observed in the previous years.

Increases in pelleted feed price during the pandemic also contributed to impairing net revenues and B/C ratio mainly due to rises in fishmeal (11.0%, Torres, personal communication, 2021) and fish oil (12.0%, SalmonExpert, 2021) used for manufacturing feed. Feed suppliers have embarked on a strategy to substitute fishmeal with other inputs containing vegetable protein to sustain feed prices and be competitive (Torres, personal communication, 2021). On the other hand, Mexican hatchery broodstock, feed, and live feed used for shrimp post-larvae production are priced in USD$ (Jaime, personal communication, 2021). The exchange rate rose approximately 7.2% during the pandemic, thus significantly increasing post-larvae unitary costs. Lebel et al. (2021) referred that higher input prices adversely impacted aquaculture farms in the Mekong Region during the pandemic.

When break-even production was calculated, the results obtained were consistent with those obtained for nr and B/C, indicating that the worse conditions occurred in March–June and May–August during the pandemic. Adverse economic conditions resulting from lower shrimp prices and increased post-larvae and feed costs required higher yields to reach break-even situations than the necessary production under the pre-pandemic scenario. However, no significant differences were estimated in break-even production in August–November between both scenarios because, as explained previously, increased shrimp price in November during the pandemic compensated for higher post-larvae and feed costs.

Overall, the sensitivity analysis indicated that the number of parameters contributing significantly to nr variance was higher under the pandemic scenario, resulting in a lower influence of each parameter on nr variance per parameter unit standard deviation than in the pre-pandemic scenario. The parameter relevance was consistent in both scenarios, particularly those strongly or moderately influencing nr variance. Consequently, although the difference in shrimp prices and costs determined the economic outcome in both scenarios, their contributions to uncertainty were of little importance when analyzed within the context of each scenario.

The sensitivity analysis also showed that the most influential economic parameter was post-larvae unitary cost. As mentioned previously, hatcheries producing Penaeus vannamei post-larvae in Mexico determine their sale prices based on the exchange rate between the MXN and USD$ since many of their inputs come from abroad and are priced in USD$. Thus, the high sensitivity on nr to stochastic variability of post-larvae cost reflects the uncertainty associated with the exchange rate.

Several sensitivity analyses have previously shown that shrimp sale price is the most important economic factor influencing net revenue (Hernandez-Llamas and Zarain-Herzberg, 2011; Hernández-Llamas et al., 2013; Moreno‐Figueroa et al. 2019; Estrada-Perez et al. 2020; Ruiz-Velazco et al. 2021). High variability in shrimp prices and price grading by size are deemed the leading causes of shrimp price relevance by those authors. However, in this investigation, shrimp price was not the most influential economic parameter because its values were relatively stable from 2017 to 2019 and in 2020, and the differences in price among shrimp sizes were minor.

Conclusion

This study concludes that operating conditions during the COVID-19 pandemic were conducive to worsening the economic performance of semi-intensive farms, particularly during the March–June and May–August schedules. However, in the short term, loss probability was null or negligible, and no evidence existed that uncertainty and economic risk increased compared with the previous farming conditions. Nevertheless, further studies should be conducted to update the pandemic impact.