A Storm to Remember:
Hurricane Harvey and the Texas Economy

Appendix 2: Detailed Methodology of Net Economic Impact Analysis

Economists use various modeling techniques to estimate the effect of economic trends and government policy options over time. For this report, the Comptroller’s office developed an approach to analyze the cost of a disruptive event on the Texas economy as well as the subsequent economic activity generated as a result of the shock. Our economic impact analysis thus estimates the net effect of Hurricane Harvey on the Texas economy.

To estimate the cost of the storm on Texas, productivity loss is approximated by discounting the expected economic forecast for three years by the amount of time businesses were closed or out of production, varying in length by industry.

To estimate the gain from rebuilding, reported and anticipated expenditures are introduced that offset the negative effects of production loss.

The forecast employs a 70-sector, 24 Council of Government (COG) region version of Regional Economic Models Inc. (REMI) Policy Insight+ for Texas, Version 2.0, an economic software application that generates realistic annual estimates of the total regional effects of policy or other market changes, based on an approach that combines and builds on input-output, general equilibrium, computable econometric and economic geography modeling techniques. The software calculates differences between the baseline (a regional control forecast) and the shock forecast.⁴⁶

The COG regions affected by the storm are assumed to be those containing counties that received FEMA assistance due to the storm.⁴⁷ All counties in the Houston-Galveston, South East Texas and Golden Crescent COGs were affected by the storm. The Brazos Valley, Coastal Bend, Deep East Texas, Alamo Area and Capital Area COGs were only partly affected and were discounted by the share of population in the affected counties in each COG to the total population of each (a “population discount”). The estimate assumes all of the businesses in affected counties were affected.

The estimate’s timeframe is the initial shock year and two forecast years. Determining the cost share among federal, state and local governments is ongoing, even as more costs are being recorded. Because we do not yet know who will ultimately bear the burden of some Harvey-related costs, the scope of this analysis is limited to a relatively short time period.

The estimate uses nominal dollars (unadjusted for inflation).

Productivity Loss

The productivity-loss component of the estimate assumes business days lost due to the storm, whether from power outages, damaged structures or temporary labor shortages, result in lower output (a “time discount”). The estimate assumes most industries were offline or experienced reduced revenue for one week, from landfall on Friday, Aug. 25, followed by five days of rain and subsequent dam overflows in Houston until it dissipated in Louisiana on Aug. 30.⁴⁸ Industries are discounted differently depending on the amount of time they were estimated to be offline, their level of competition and their place in the supply chain:

• manufacturing and mining were assumed to be offline or experiencing reduced revenue for 15.4 days.⁴⁹
• hospitals were assumed to be offline or experiencing reduced revenue for four days.⁵⁰
• firm-level competition is assumed for all industries except those with a high location quotient (LQ > 4) such as oil and gas extraction, which are considered exogenous.⁵¹
• retail and wholesale trade are further discounted to account only for the markup of cost of goods sold, to avoid double-counting.⁵²
• accommodation is assumed to be unaffected by productivity losses as the decrease in tourism from the storm could be counterbalanced by the increase in hotel occupancy by evacuees.

The standard regional control is reduced by a percentage of sales derived from a combination of the time discount and population discount. The results show the reduction in GSP due to this reduction in output.

A similar reduction in labor productivity was considered; however, it is assumed the negative effects on wages in August would be counterbalanced by increases in the fourth quarter. It is also assumed that salaried employees were largely unaffected by the storm and would either telework or make up lost time in September, while non-salaried employees would experience a dip in productivity and income in August but would have more work opportunities and higher wages in the recovery months following the storm.

Because the model treats labor productivity differently based on regional and industry variation, the effects of the storm on labor productivity could have counterintuitive effects; nevertheless, change in labor productivity is left outside the scope of this estimate.

Gains from Rebuilding

Following economic shocks, institutions begin to respond to the community’s needs, both immediate and ongoing. Gains from the rebuilding component of the estimate account for increases in spending from government, businesses and nonprofits on timely disaster relief, shelter and food for displaced people, debris removal, medical attention and reconstruction.

For this estimate, current and expected expenditures were collected — via either news sources or self-reporting — from federal and state agencies as well as private insurance companies and large nonprofits; these are non-exhaustive. Each organization’s expenditures are categorized by expected use over a three-year period, divided 40/20/20, and assuming the remaining 20 percent will be spent in future years beyond the three-year scope of this analysis.

Funds flowing from and through state agencies are allocated by individual industry and weighted by output in construction, housing, health care and social services.⁵³ The estimate weights funds categorized as government administration or equipment by population. It excludes agency expenditures reallocated from a similar use, such as medical costs expected to be covered by Medicaid funds.

Funds from the National Flood Insurance Program,⁵⁴ Small Business Administration (SBA) loans,⁵⁵ private insurance companies⁵⁶ and nonprofits⁵⁷ are allocated by individual industry in the proportion by which funds were released for SBA loans following Superstorm Sandy: roughly 64 percent on construction for real estate damages, weighted by output; 10 percent on equipment for business content, weighted by population; and 26 percent on relevant consumer spending for home content (such as motor vehicles, furnishings, housewares and health services), weighted by consumer spending.⁵⁸

Limitations

This estimate is intended to depict Texas’ economy as a whole and the net effects of Hurricane Harvey based on currently available data. Figures for government spending may change as agencies report expenditures and more people submit claims.⁵⁹ It is a projection, and does not account for:

• damage to commercial, government or personal property, including real estate, contents, equipment, vehicles, inventory, etc. Instead, it accounts for the funds likely be spent in the next three years to rebuild and replace these items.
• expenditures from smaller nonprofit organizations.
• change in tax burden on Texans at the local and state levels due to increased costs from Harvey recovery or state budgetary actions that may be taken.
• change in government services provided due to resource reallocation.
• income to insurance companies from deductibles or potential changes in insurance premiums.
• productivity loss and gains from agricultural insurance; this study focused on the nonfarm portion of the economy. The REMI model does not include an agriculture sector.
• non-pecuniary losses due to fatalities or decreased desirability of living in an area. The estimate assumes people who do not receive a buyout will rebuild — especially property owners along the coast in hurricane-prone areas, who are likely to understand the risk of property ownership in their location.
• the long-term costs of flooding, including buyout programs, new reservoirs, bayou dredging or seawall construction. These flood mitigation efforts will cost billions of dollars over a number of years and are beyond the scope of this study.

The model available divides Texas into COG regions and depicts dynamic relationships between industries and market forces; future studies may benefit from a more granular model to show county-level damage to housing stock, which would eliminate the need for the population discount.

This estimate intends to depict the order of magnitude of the net effects of Hurricane Harvey on the Texas economy as a whole. The economic losses and gains may not be known in their entirety, but our approach aims to provide a high-level perspective of the possible damage by a severe weather event such as Harvey as well as the strength of the Texas economy to withstand such events. Individual Texans and communities may continue to bear a heavy burden rebuilding their lives in the wake of the storm, but the assistance provided by government, business and nonprofit resources and the diversity of Texas’ economy protect the state from the level of devastation experienced by smaller, less robust economies after an economic shock.