Data appendix.

*All tables and equations refer to the main text of article where all references are listed as well.

Steps:

  1. Aggregating the 73 sectors of the national I/O table to the 26 sectors for which regional statistics exist.

Inputs:

  • National I/O Table for 2005 (73x73 sectors) (INE, 2009)
  • Regional statistics on GDP of Murcia region (26 sectors) (CREM, 2010)
  • Sector definitions as per the CNAE93 accounting system (INE, 2009)

Calculations:

  • Adding together sectors in rows and columns of the I/O table to reduce the table from 73x73 to 26x26 cells. Calculations are performed on the table with inter-industry relations in base prices. Total inputs (and intermediate consumption categories) and total outputs (and final demand categories) are aggregated as well.

Estimations and assumptions:

  • None

Outputs:

  • National I/O Table for 2005 (26 x 26 sectors)
  1. Constructing regional I/O table based on location quotients

Inputs:

  • National I/O Table for 2005 (26x26 sectors) (Step 1)
  • Regional employment statistics per sector of Murcia region (26 sectors) (CREM, 2010)

Calculations:

  • The SLQ (Miller and Blair, 2009), CILQ (Miller and Blair, 2009), FLQ (Flegg and Webber, 1997), and AFLQ (Flegg and Webber, 2000) methods are applied to construct non-survey based regional I/O tables of technical coefficients. Both a Leontief and Ghosh formulation are applied.

Estimations and assumptions:

  • The FLQ formula contains an unknown parameter δ(0 ≤ δ ≤ 1) influencing the degree of convexity of scaling factor λ* (Flegg and Webber, 1997). The appropriate value forδis derived from Step 3.

Outputs:

  • Regional I/O Tables of technical coefficients for 2005 (26 x 26 sectors) for each of the methods applied.
  1. Selecting the most appropriate location-coefficient-based I/O approach

Inputs:

  • National I/O Table for 2005 (73x73 sectors; INE, 2009)
  • Regional I/O Tables for neighbouring autonomous regions: Andalucia (63x63 sectors; IEA, 2010) and Valencia (67x67 sectors; IVE, 2008)
  • Sector definitions as per the CNAE93 accounting system (INE, 2009)

Calculations:

  • Adding together sectors in rows and columns of the I/O table to reduce the table from 73x73 to the number of sectors in the respective regional tables. Calculations are performed on the table with inter-industry relations in base prices. Total inputs (and intermediate consumption categories) and total outputs (and final demand categories) are aggregated as well.
  • The SLQ (Miller and Blair, 2009), CILQ (Miller and Blair, 2009), FLQ (Flegg and Webber, 1997), and AFLQ (Flegg and Webber, 2000) methods are applied to construct non-survey based regional I/O tables of technical coefficients based on employment statistics contained in the I/O tables. For the FLQ and AFLQ methods, simulations were done with the following values of δ: 0.05, 0.10, 0.15 and 0.20. The Leontief inverse matrix is calculated and, finally, regional output multipliers are produced.
  • Calculation of two measures to evaluate the relative success of each LQ method in reproducing the survey-based output multipliers by estimating regional output multipliers from the national I/O table (Equations 11 and 12).

Estimations and assumptions:

  • Initially a larger range of values for δwas tested but found to yield poorer results
  • The following criteria were used to determine the appropriate LQ method for Murcia: a) the need to have a low average absolute deviation of the average regional multiplier; b) a preference for a slight underestimation of the average regional multiplier; c) the trend observed in literature that smaller regions (such as Murcia) have a higher propensity to have a lower optimal value for δ; and d) that such a trend would place an optimal δ for Murcia’s agricultural sector in the FLQ approach below 0.15; as well as e) that the average absolute percent error for the six data rows in Table 3 is lowest for FLQ with δ = 0.10 (see overall rank), we applied FLQ with δ = 0.10 to develop a non-survey based regional input-output table for Murcia.

Outputs:

  • Regional I/O Table of technical coefficients and inverse Leontief matrices for the two test regions for each of the methods applied.
  • A ranking of LQ methods (including an appropriate value of δ) – See Table 3.
  1. Disaggregating the agricultural sector of the regional I/O Table

Inputs:

  • Regional I/O Tables of technical coefficients for Murcia for 2005 (26 x 26 sectors) created by the selected LQ method (FLQ with δ= 10), based on the Steps 2 & 3.
  • Definition of 5 agricultural subsectors: 1) grains and other annual field crops; 2) horticulture and fruit trees; 3) grapes; 4) olives and almonds; and 5) livestock.
  • Value of agricultural output for agricultural enterprises and groups of enterprises for 2005 (CREM, 2010)
  • A list of quantities of the most important intermediate consumption categories for the agricultural sector (CREM, 2010).
  • Regional agricultural statistics and miscellaneous secondary data (CARM, 2005; 2007; Fleskens, 2005)
  • Regional survey-based I/O Table for Valencia (IVE, 2008) in which agricultural sector is disaggregated.

Calculations:

  • First, the technical coefficients for sectors i supplying inputs to the agricultural sector were multiplied with the total value of agricultural output.
  • Second, total output from the newly defined 5 agricultural sectors was calculated from the aggregation of different individual agricultural enterprises and groups of enterprises.
  • Third, based on a list of quantities of the most important intermediate consumption categories (CREM, 2010), 42% of intermediate consumption could directly be attributed to specific subsectors. In other cases, agricultural statistics and secondary data (CARM, 2005; 2007; Fleskens, 2005) were employed to distribute a further 22.5% of intermediate consumption items over relevant subsectors.
  • Fourth, for smaller categories of intermediate consumption for which no further data was available, with a known value of total agricultural output (from step 1), the regional I/O table with a single agricultural sector was (with some assumptions, i.e. proportionate allocation) used to balance remaining expenditure on intermediate consumption in the five subsectors.
  • Fifth, using subsector total output, the quantities of inputs were converted into technical coefficients.
  • Finally, constructing input to non-agricultural sectors from the 5 agricultural subsectors was relatively straightforward as the sum of subsector technical coefficients was required to remain equal to that of the non-disaggregated agricultural sector technical coefficient for each column. The distribution over subsectors for key-sectors with high volumes of agricultural inputs (i.e. agro-food, textile and leather, lumber and cork, and paper industries, and hotels) was informed by a comparison with data for the neighbouring Valencia autonomous region. The sub-matrix of distribution coefficients was used to balance the inter-industry input coefficients.

Estimations and assumptions:

  • Agricultural statistics and secondary data (CARM, 2005; 2007; Fleskens, 2005) were employed to distribute intermediate consumption items such as fertilizer (8.5%), phytosanitary products (7.4%) and energy/lubricants (6.6%) over relevant subsectors. Distribution took into account the share of the subsector in regional land use and rates of application reported in local or similar conditions.
  • For minor categories of intermediate consumption for which no further data was available, proportionate allocation was assumed over sectors to balance remaining expenditure (this accounts for 35.5% of intermediate consumption).
  • The distribution over subsectors for key-sectors with high volumes of agricultural inputs (i.e. agro-food, textile and leather, lumber and cork, and paper industries, and hotels) was informed by a comparison with data for the neighbouring Valencia autonomous region.

Outputs:

  • Regional I/O Table of technical coefficients for Murcia with the agricultural sector disaggregated (31x31 sectors).
  1. Estimating regional final demand and sector output

Inputs:

  • Available regional final demand data for Murcia (CREM, 2010)
  • National I/O Table for 2005 (26 x 26 sectors) (Step 1)
  • Data on regional exports (good quality) and sector output (inconsistent, except for agricultural sector) (CREM, 2010)

Calculations:

  • Most required final demand data for Murcia were obtained from CREM (2010). National sector final demand scaled down using employment data was used to fill regional data gaps. For example, regional household final consumption was found to correlate very well (r2 = 0.996; µ1= 0.8%; µ2= 3.8%) with national data for an aggregated number of consumption goods and services. Therefore, disaggregated household final demand could be obtained from the scaled down national data. One exception is the sector hotels and restaurants where the significantly lower regional household expenditure data was inserted.
  • Similarly, capital formation for industries was derived from the scaled national data, and the entire expenditure structure of national public administration was used in deriving individual sector totals from the regional aggregate total.
  • Importantly, good regional data on exports were available. As expected, the regional and national level data bear little relation, both in overall size (regional exports were 20 times larger than the scaled national data) and structure (r2=0.07).
  • After deciding on the location quotient method to employ (Step 3), the regional total final demand vector (f) was entered in Equation (1) to estimate total regional output. Incomplete sector output data was available from CREM (2010), but appeared to be inconsistent in its definition of sectors and in relation to final demand. Agricultural sector output data was an exception, and these were used in further analyses (Equations 13-19) together with simulated output for industrial and service sectors.

Estimations and assumptions:

  • Final demand for sectors for which no data was available was estimated from the national I/O table scaled down using employment data (Step 2).
  • Capital formation for industries was derived from the scaled national data
  • The entire expenditure structure of national public administration was used in deriving individual sector totals from the regional aggregate total.
  • Output for industrial and service sectors was simulated from scaled national data.

Outputs:

  • Regional I/O Tables of technical coefficients and inverse matrices for Murcia with the agricultural sector disaggregated (31x31 sectors) and including final demand and output.
  1. Creating water I/O table

Inputs:

  • Regional water statistics for agriculture (2005), industries (1999), and piped distribution network (2007) – CREM (2010)
  • Water statistics data from Andalucía (Consejería de Medio Ambiente, 1996) and Spain (INE, 2010)
  • Regional I/O table (Step 5)

Calculations:

  • Some regional water statistics were available as a basis to calculate sectoral water use (CREM, 2010). Note that water statistics for agriculture were available for 2005, but breakdown of industrial water use was only available for 1999, and specified water use of the service sector could not be found at all. To circumvent these lacunas, data for 2007 from the piped water distribution network used in economic sectors yielded some piecemeal information, and the available statistics were used together with equivalent data from Andalucía (Consejería de Medio Ambiente, 1996) and Spain (INE, 2010) to calculate Direct Water Consumption (DWC) and to harmonise sectoral water consumption
  • Data for industrial sectors for 1999 was updated by estimation of the 2005 level output using the input-output model. Total sectoral water use was subsequently updated where sector growth (positive or negative) had been such that DWC calculated with the 1999 water use would become questionable in comparison to national data. The largest water consumers are the agro-food and chemical industries, although DWC is equally high in rubber and plastics and metallurgy. At the national level, DWC’s for industrial sectors are generally lower, although electricity, gas and water stands out as a relatively heavy water user. The very high DWC’s of the paper (including publishing and printing), chemical, and other manufacturing industries reported for Andalucía were not found in Murcia.
  • Water use of the service sectors was redistributed according to the relative importance of water consumption of these sectors in Andalucía, while respecting the total service sector consumption for Murcia. Like with industrial sectors, the DWC’s thus obtained are lower than those in Andalucía. Water consumption is largest in the hotel and restaurants and real estate sectors, with the former having the largest DWC amongst the service sectors.
  • Calculation of the matrix with water inter-industry input coefficients, the inverse matrix and back and forward linkages water multipliers. Backward linkages water multipliers represent how much water is used indirectly in a given sector by considering the water consumption for its intermediate consumption in relation to direct water use. Forward linkages water multipliers represent the ratio of additional water use in purchasing sectors relative to the direct water consumption ‘embedded’ in output from the supplying sector considered.

Estimations and assumptions:

  • Table 1 gives an overview of sector water consumption data and the reference area (Murcia, Andalucia, Spain) for which they were obtained. Harmonized water consumption data are also included.
  • In the case of Murcia, grains and olives and almonds are hardly irrigated. The bulk of water is used in producing high value fruit and vegetable crops. The high DWC in Andalucía may stem from significant water use in low value crops (grains) and relatively wasteful irrigation techniques: 45% of irrigation is by gravity (Dietzenbacher and Velázquez, 2007). In contrast, in Murcia 85% of water is supplied to crops by drip irrigation (CREM, 2010). The exception to relative water use efficiency is the livestock sector which is intensive in Murcia and presumably less so in Andalucía (also note that the latter figures are considerably older).

Outputs:

  • Regional I/O Tables of water inter-industry coefficients and inverse matrices for Murcia with the agricultural sector disaggregated (31x31 sectors) and including final demand and output.
  1. Water scarcity scenarios and farmers’ land use responses in Torrealvilla catchment

Inputs:

  • Farm survey in Torealvilla catchment (99 valid interviews), representing land uses as observed from satellite imagery (2004).
  • Agricultural census data of the Murcia region
  • Description of scenarios A-C and D

Calculations:

  • Different scenarios were presented to farmers who currently have access to water and those who do not. The former group of farmers was asked how the following will affect the future of their current principal land use:

-Scenario A – No access to water for agricultural use (total water depletion – this could occur as a physical lack of water locally, or as water quality deteriorates beyond maximum tolerable salinity levels);

-Scenario B – Government imposes tax on groundwater abstraction resulting in a water price higher than maximum willingness to pay for water (WTP – lowest €0.20 m-3; highest €0.60 m-3; average €0.31 m-3; standard deviation €0.08 m-3) by individual farmers; and

-Scenario C – Government imposes tax on groundwater abstraction resulting in a water price of up to the individual farmer’s maximum WTP.

  • Individual WTP was used as cut-off point to avoid presenting multiple (fixed) price scenarios to each farmer and is justified by the fact that our purpose was not to elicit farmer WTP, but to explore potential land use change along a gradient of physical water scarcity (Scenario A), economic water scarcity (Scenario B) and economic water insecurity (Scenario C). Farmers’ responses were: 1) no change; 2) conversion to other agricultural land uses; and 3) stop farming/abandonment.
  • Farmers who currently do not have access to water were asked how their principal agricultural land use may alter if water became available, e.g. through IBWT. This led to a fourth scenario (D):

-Scenario D1 – Water becomes available to previously non-irrigable areas.

At this stage, we found that grain farmers demonstrated little dynamism as compared to olive and almond farmers. This is counter-intuitive, as conversion costs are considerably lower for the former group. As grain farmers may have been underrepresented in the sample, we therefore also defined an adjusted hypothetical scenario:

-Scenario D2 – as Scenario D1, but for the grain farmers we adopted weights of conversion to irrigated farming as elicited from olive and almond farmers (resulting in increasing propensity of grain farmers to change).

The responses registered in Scenarios D1 and D2 were: 1) no change; 2) increase production (expansion); and 3) conversion to irrigated agriculture.

Estimations and assumptions:

  • Estimation of farm sample size. The final number of respondents was 7 for grains, 24 for almonds and olives, 32 for grapes, 24 for horticulture and fruits and 12 for livestock. If we take agricultural census data of the Murcia region as a basis for estimation, the total number of farmers in the Torrealvilla catchment (which is unknown) could be 1,400. As extensive land uses are over- and intensive land uses underrepresented in the catchment relative to the region (where grains occupy 10% of UAA, almonds and olives 18%, horticulture and fruits 19%, grapes 6%, and livestock 2%) the average farm size is likely larger and the number of farmers smaller. The average farm size of our sample is 16 ha, against 10 ha across the Murcia region. Using this figure, the total number of farms in Torrealvilla would be lower, around 875. Our sample of 99 farmers interviewed thus represents at least 7% and perhaps more than 11% of the total population.
  • The tax on water in scenarios B and C was presented as implying a higher price of water, a situation that could also be brought about without government intervention as farmers may need to pay more to obtain water in sufficient quantity and of sufficient quality.
  • In the context of this paper the maximum WTP refers to a threshold beyond which the maintenance of present farming activity is perceived by individual farmers as no longer viable, making drastic change such as agricultural abandonment is highly likely.
  • Although the incentive to exaggerate may be more pronounced for water price than for land use change effects of scenarios, we cannot rule out that (some) responses are exaggerated; therefore the results presented should be regarded as potentially extreme land use change effects
  • For the purposes of expansion (Scenario D) we assumed scrubland and fallow to be available, but not forest and other land uses. The effective area within the Torrealvilla catchment is thus reduced to the 140 km2 of UAA.

Outputs:

  • Discrete choice scenarios and maximum (total and incremental) WTP estimates elicited for different land uses.
  1. Upscaling local scenario responses to Murcia Region

Inputs: