Local Response to Water Crisis: Explaining Variation in Usage Restrictions during a Texas Drought Megan Mullin Duke University megan.mullin@duke.edu Meghan E. Rubado Temple University mrubado@temple.edu April 10, 2015 Abstract: What explains local policy response to extreme weather events? This question takes on growing importance as climate change increases the frequency of droughts, floods, heat waves, and severe storms. Local governments considering how to respond to these events may face political opposition to policies that restrict resource use or otherwise limit personal activities. Using data on the adoption of local water usage restrictions during the 2010-13 Texas drought, we examine how features of a water system and its customer base as well as severity of the drought influence the timing and stringency of policy response. We find that problem conditions and institutional capacity of water systems outweigh political interests in shaping drought emergency response. As global climate change increases the frequency of extreme weather events such as droughts, floods, heat waves, and severe storms, local governments come under growing pressure to implement effective emergency response. Although responsibility for crisis management often is shared among multiple levels of government, natural hazards typically are experienced at the local level, and local policy makers are most familiar with local conditions and citizen preferences. In responding to these events, localities have a duty to protect the well-being of their residents, but they also want to avoid overreacting and causing further damage to the local economy. Moreover, they may have limited capacity to take action, constrained by shortfalls in knowledge about the problem, organizational weakness, or financial or human resources. This paper reports findings from research analyzing local government decision making in response to a climate-change related extreme event. Existing research offers little evidence comparing the reactions of different localities to a common crisis in order to unravel how objective conditions related to the crisis itself, political demands from the community, and previous policy context influence emergency response. Our work is situated in Texas, which experienced severe drought following an exceptionally devastating La Niña weather pattern in the winter of 2010-11. By April 2011, the entire state was experiencing drought conditions, with 68% of the land area designated in one of the highest two categories of extreme or exceptional drought. Continued warm weather and lack of rainfall over subsequent months produced the hottest and driest year ever recorded in Texas. Another La Niña the following winter exacerbated these conditions, producing a long-term drought from which much of the state still has not recovered. 1 The drought has caused significant economic disruption and losses. In 2011, more than 31,000 fires occurred, affecting more than four million acres of land and destroying almost three thousand homes. Through 2013, the drought has cost the state’s farmers and ranchers an estimated $8 billion and caused $3.4 billion of losses to the timber industry (House Committee on Natural Resources 2013, 26). Unsurprisingly, it also has taken a toll on the state’s water resources. Groundwater-level recorders in the state’s nine major aquifers showed declines in water levels between 2010 and 2012, with the most dramatic drops observed between 2010 and 2011. The median decline in water level between 2010 and 2011 for 125 recorders in the nine major aquifers was 4.8 feet, and the median decline between 2011 and 2012 was 0.9 feet, with the Ogallala Aquifer wells in Northwest Texas showing the greatest declines through 2012 (Texas Water Development Board 2013). Water levels in Texas reservoirs also showed steep drops during the drought of 2011 and have failed to rebound since that year. While the median percent of capacity for Texas reservoirs between 1990 and 2013 hovered around 80 percent, the levels in 2013 and early 2014 remained at near-record lows of 60 to 65 percent (Texas Water Development Board 2014a). Two reservoirs dried up completely between 2012 and 2013 (Texas Water Development Board 2014b). Faced with these shortages in supply, public water systems have needed to find ways to stretch their limited water resources further. Many are pursuing new infrastructure projects to tap additional supply sources, but these are long-term strategies that at best will help communities prepare for future drought events. In the short term, utilities facing water shortfalls have two choices: they can find ways to immediately boost supply by drilling a new well, extending surface water intakes, or interconnecting 2 with neighboring systems; alternatively, they can encourage customers to change consumption patterns to live within resource limits. Living within limits long has been a challenge for Texas water utilities, however, which often have played an important role in enabling rapid growth that has contributed to groundwater depletion (Perrenod 1984; Porter, Lin, and Peiser 1987; Thomas and Murray 1991). Texas water systems pursuing aggressive growth policies sometimes have been blind to resource constraints, ultimately running out of the water they promised to deliver (Mullin 2009, 107-110). Water officials seeking to satisfy customers’ demands for plentiful water may postpone setting limits on water usage as long as possible, jeopardizing reliable delivery of water for essential needs in order to avoid setting restrictions on non-necessary water usage. We examine how conditions related to the event as well as characteristics of local water utilities and the communities they serve influenced short-term crisis management decision making during the Texas drought. Extreme weather events place considerable stress on local officials who may be uncertain about the extent and duration of the problem and may face political resistance to any policies that restrict resource use or otherwise limit personal activities. Research on local policy response to water shortages and other weather-related emergencies largely has focused on individual cases, because data are rarely available documenting the actions of a large number of communities to a common crisis. To fill this gap, we take advantage of data collected by the Texas Commission on Environmental Quality (TCEQ) on local adoptions of voluntary and mandatory water usage restrictions over the period 2010 to 2013. Using Geographic Information Systems (GIS) techniques to combine these data with information about demographic characteristics of water system customers and the spatial spread of the 3 drought, we test how problem conditions and internal water system characteristics and constituencies contributed to systems’ willingness to adopt usage restrictions during the drought. Our results indicate that problem conditions and institutional capacity of water systems outweigh political interests in shaping drought emergency response. Local Decision Making in the Context of Climate Change Climate change is a global problem with effects that are most visible at a local scale. The nature and severity of its impacts vary across communities in ways that are attributable to regional patterns of risk as well as to the condition of a community’s built infrastructure. Local governments are responsible for building and maintaining much of that infrastructure that can help minimize—or amplify—climate change risks. Local governments seem to recognize their critical role in planning for climate change. Although any individual locality acting to mitigate climate change will have negligible effect on the problem, thousands of cities worldwide nonetheless have pledged to reduce their greenhouse gas emissions (Zahran et al. 2008; Krause 2010; 2011; Sharp et al. 2011). These commitments often are part of a broader effort to enhance local sustainability (Portney 2003; Lubell et al. 2009). Local governments also are taking action to adapt to the consequences of climate change, but these efforts have received less attention in the literature (Betsill and Bulkeley 2007; Bulkeley 2010). One reason for the relative lack of attention is a measurement problem: because local governments provide much of the critical infrastructure that can help protect communities or put them at risk from climate-related events, nearly all decision making about how to manage that infrastructure becomes a form of adaptation planning. Whether or not local officials 4 perceive their actions as related to climate change, their decisions about infrastructure investment and land use planning affect the vulnerability of local populations to intensifying hazards. Relative to mitigation efforts, then, adaptation planning produces more direct benefits for the immediate community. Response planning for extreme events is a form of climate change adaptation. Managing impacts from natural hazards such as drought entails both long-term strategies to reduce risk by integrating hazards into overall comprehensive planning processes and short-term strategies to respond to crisis (Wilhite 2000). Our focus in this paper is on crisis response. The literatures in planning and climate adaption advocate for increased emphasis on risk management and resilience, but localities thus far have failed to heed that call. A recent study of comprehensive planning processes in 81 of the fastest growing U.S. counties found that they generally do not address drought risk (Fu and Tang 2013). The literature on disaster planning more generally indicates wide variability in local governments’ preparation for hazards. Although some localities respond to crisis events such as the terrorist attacks on September 11, 2001 by making plans to prepare for potential future events, many do not (Krane 2002). Cities’ planning processes may overlook important sources of risk, as when the 1999 New Orleans comprehensive plan ignored the extreme flood hazard facing the city (Burby 2006). In the absence of advance preparation, how a community responds to emergency conditions takes on greater importance. Our emphasis here is on decisions by local water utilities to enact restrictions on water usage by their customers. Usage restrictions are a blunt instrument for promoting water conservation, but in the case of extreme water shortages they could make the difference about whether a community runs out of water. 5 A study of the effects of various usage restrictions imposed by eight water providers during a 2002 drought in Colorado found that mandatory restrictions were effective at reducing water use, while voluntary restrictions were not (Kenney et al. 2004; Kenney et al. 2008). Under emergency conditions, other demand management strategies such as pricing and incentives are unlikely to produce the immediate reductions in usage that may be needed to conserve dwindling water supplies. Understanding disaster response also may provide insight about long-term drought planning, because localities that are slow to respond to crisis conditions may be less likely to enact policies to manage water demand in the long term as well. Texas state law requires all wholesale and retail public water suppliers to prepare drought contingency plans that outline a set of temporary supply and demand management responses to be introduced during water supply shortages. Revised plans must be submitted to the TCEQ every five years. Drought plans are distinct from the water conservation plans that most Texas water systems also must submit, in that drought plans outline short-term response measures for use during emergency conditions, while water conservation plans focus on producing lasting improvements in water use efficiency. In practice, achieving long-term reductions in water usage can make drought planning more difficult because less water is dedicated to nonessential uses and therefore available for short-term reductions. Drought contingency plans prepared by local water suppliers identify best management practices for reducing water use in a series of successive stages according to the severity of water shortage conditions. Each plan must include a set of triggering criteria specifying when a response stage should begin or end, targets for water use 6 reductions within each stage, and a list of supply and demand management measures designed to produce the target reductions and better manage available supply. To the extent possible, triggering criteria are supposed to be expressed in quantitative terms, and they typically take the form of expectations about how long existing water supplies will last given operators’ assumptions about demand conditions and replenishment of supplies. Water systems choose how supply levels relate to drought stage triggers, however, so that a six-month supply would qualify as a shortage condition in one community but not another. Moreover, the state recommends that local plans provide discretion to system operators in deciding if and when to initiate or terminate a response stage, advising the operator to “weigh the risks of delay against the potential public relations problems caused by ‘false alarms’” (TCEQ 2005, 9). The decision to implement usage restrictions is likely to invite public resistance and political controversy. Rules that limit outdoor watering to certain days or ban activities such as car washing or hosing down pavement often seem arbitrary and overly restrictive to a public that has short time horizons and little knowledge about water supply conditions, even during a period of drought. If neighboring communities have not enacted similar restrictions, residents may perceive that costs of the drought are not being shared equally. Water-intensive industry and commercial businesses such as golf courses, car washes, and landscaping companies could suffer significant economic losses from the enactment of usage restrictions. Thus local water officials may face competing pressures in deciding whether and when to implement usage restrictions. On the one hand, they want to ensure reliable water supply for essential uses throughout the uncertain duration of the drought event, and they may be under pressure from other water agencies in the 7 region to reduce consumption from a shared supply source. On the other hand, to limit the water use of their own customers counters the goals of most water utility professionals, who traditionally have aimed to satisfy customers’ demands for water even in the presence of variability in supply (Lach, Rayner, and Ingram 2005). Many local water agencies also have institutional features such as systems of political incentives and rules for participation that can interfere with strong policy action (Brown 2004; Hughes and Mullin n.d.). Prior research has shown that even where a local agency has clear authority to address sustainability, it can be difficult to integrate sustainability considerations with other responsibilities that are perceived as more central to the agency’s mission—especially if environmental goals are perceived to threaten economic goals (Bulkeley and Betsill 2003; 2005). Facing resistance to usage restrictions among their customers, water managers may anticipate that the state or neighboring localities would bail them out if they were to drain their own water supplies. We assess how water system features, demographic characteristics of the system’s customer base, and local drought severity influence how local officials balance these competing pressures. In developing our hypotheses, we draw on the policy diffusion literature, which has emphasized the importance of accounting for internal determinants of policy adoption as well as external diffusion mechanisms (Berry and Berry 1990; 1999; Mintrom 1997; Daley 2007). While there have been studies of environmental policy adoption that account for internal and external conditions at the state level (Lowry 1992; 2005; Daley 2007), diffusion studies of local environmental policy adoption have been lacking. In part, this may be due to the difficulty of controlling for internal determinants at this level of analysis. The data leveraged in this study provide a unique 8 opportunity to account for factors internal to water systems and their political jurisdictions that may influence policy makers’ approach to usage restrictions as well as addressing problem conditions that are shared across jurisdictions yet differentially affect each of them. Because we have data not only on whether water systems enacted usage restrictions but also when, we are able to evaluate how these factors relate to the speed with which a system acts to request or require its customers to limit water usage in order to preserve existing supplies. Hypotheses Our hypotheses capture different potential sources of influence on water system decision making: objective problem conditions, institutional capacity, and political interests. With respect to problem conditions, we expect as drought conditions become more severe, the amount of time a system can resist imposing usage restrictions will be reduced. A water system typically draws its supply from a river or groundwater aquifer located within or near its service area, so more severe drought in its area is likely to indicate scarcity of supply. Local drought severity also may increase public awareness and willingness to accept limitations on water use. H1: Water systems will adopt usage restrictions more quickly as drought conditions become more severe. Our conception of institutional capacity relates to the organizational resources available to water managers that may provide the information and discretion needed to implement usage restrictions, as well as features of the water system itself that influence its ability to withstand severe drought without restricting use. Previous case study-based research indicates that management capacity is an important predictor of local 9 governments’ ability to plan for and adapt to climate change-induced water shortages (Ivey et al. 2004; Pirie et al. 2004). We use system size as a proxy measure for this capacity. Larger water systems tend to have more technical expertise and human resources that are needed to assess supply vulnerability and to enforce usage restrictions. Recent work on local management of terrorism risk found a significant relationship between level of preparedness and city size (Gerber et al. 2005). Although smaller water systems may be more vulnerable to supply shortages, all else equal we expect them to be slower to enact a major policy initiative such as usage restrictions. H2: Large systems, as measured by the population served, will adopt usage restrictions earlier than systems delivering water to fewer customers. We also expect that basic features of a water system that define its capacity and resiliency will allow the system to endure water shortages for longer before enacting usage restrictions. H3: Water systems that have less storage and fewer interconnections will adopt usage restrictions earlier than systems with more storage and interconnections. Another element of capacity is a water system’s governing structure. Drinking water may be distributed by a city department in a municipality, with ultimate decision making authority lying with the mayor and city council, or by a municipal utility district or other form of limited-purpose special district. More rarely, a private nonprofit or forprofit water supply company may manage a community’s drinking water. All else held constant, we expect water systems operated by cities to hesitate least in enacting restrictions. Requiring or even requesting changes in behavior is a departure from the usual types of policies that drinking water systems enact. City councils will have more experience and more legitimacy than other types of water governing boards in issuing 10 mandates. As more visible members of their communities, they also likely have more persuasive power in requesting voluntary usage restrictions. Special districts and private water companies are not perceived to have the same level of authority as a city and do not have access to the same enforcement tools, so should be more hesitant to enact restrictions that they may be unable to implement. H4: Systems operated as city departments will adopt usage restrictions earlier than special districts or private water systems. Our final measure of capacity is the relationship of water systems with the endusers of their water. Wholesale suppliers are required by law to prepare and implement drought contingency plans, but because these agencies do not have direct relationships with the households and businesses that consume the water, they are less well positioned to design appropriate restrictions and enforce them effectively. Implementation of a drought plan requires coordination with retail water systems, creating significant transaction costs and raising questions about equal treatment among the water agencies served by a wholesale provider. Thus, although wholesalers are likely to encourage water usage reductions during drought emergencies, we hypothesize that they will hesitate to enact mandatory use restrictions. H5: Water systems that serve a larger percentage of retail customers will adopt usage restrictions earlier than predominantly wholesale systems. In addition to assessing the impacts of problem status and system capacity, we test how the customer base influences a water system’s approach to enacting use restrictions. Constituent interests have an important influence on local land use regulation outcomes (Gerber and Phillips 2004; Lubell et al. 2005; Ramirez de la Cruz 2009; Lubell et al. 11 2009), and evidence suggests that local government efforts to reduce earthquake risk may be related more to local political demands than to objective earthquake risk (May and Birkland 1994). Whether they are elected or appointed to their positions, water officials—like politicians making land use decisions—have an incentive to make decisions that allow them to keep their jobs (Mullin 2009). As evidenced in the guidance quoted above, even state regulators advise that local officials be sensitive to the potential for use restrictions to incite political outcry. We expect this type of outcry will be most likely in communities with high per capita water usage, where the built infrastructure of large lawns and swimming pools or the water-intensity of commercial activity heighten the real or perceived costs of water restrictions. Because water usage is strongly correlated with income, we also include a measure of poverty. This is an especially hard test of political interests, because high levels of water usage prior to drought indicates more flexibility in water demand, allowing usage restrictions to produce bigger conservation gains (Kenney 2014). H6: Systems with low rates of consumption per connection and higher poverty rates will adopt usage restrictions earlier than systems with high consumption rates and lower poverty. We also hypothesize that customer bases that are more Republican will be more sensitive to the potential impact of restrictions on personal liberty and business activity, so will resist at least mandatory restrictions most actively. H7: Systems that serve customer bases with higher proportions of Democratic voters will adopt usage restrictions earlier than systems with more fewer Democratic voters. Following work showing that land use regulation is associated with the socioeconomic composition of communities, we include variables measuring racial and 12 educational composition as controls, although we do not have strong predictions about the direction of their effects. We also include the percentage of houses built after 1980 Data and Model Part of a public water system’s drought contingency obligations is to report to the TCEQ within five days of the implementation of any mandatory provisions of a drought contingency plan. We use these reports to measure the implementation and timing of usage restrictions. The TCEQ’s reporting form asks water systems to report the level of usage restriction by the following code: mandatory outdoor watering schedule; no outside watering, limited hand-held hose use only; and no outside water use. Because these categories do not match the type or threshold of restrictions designated in many local contingency plans, to avoid measurement error we bundle the categories into a single dichotomous variable indicating whether a mandatory restriction is in place. Many local water systems also report the enactment of voluntary usage restrictions, and in a separate model we include voluntary as well as mandatory limitations. We obtained data on all reported usage restrictions between January 1, 2010 and November 22, 2013 from a Freedom of Information Act (FOIA) request to the TCEQ. In early 2010 when our data begin, only a small portion of east Texas was experiencing drought conditions, unrelated to the statewide event starting in summer of that year. Data on system characteristics also come from the TCEQ on FOIA request. We requested data on all public water systems serving more than 3,300 connections, which are the only systems required to submit drought plans to the state. In addition to variables outlined in the hypotheses above, we included control variables for type of source water 13 (surface or groundwater), the percentage of houses built after 1980, and a water system’s declaration—submitted to the TCEQ at the time of reporting usage restrictions—about its “level of concern,” or the number of days’ water supply the system has remaining. The estimate only sometimes coincides with triggering criteria in a water system’s drought contingency plan. The variable is a five-point scale with values at emergency (could be out of water in 45 days or less), priority (90 days or less), concern (180 days or less), watch (greater than 180 days remaining), and resolved (all drought-related issues have been resolved). We report models both with and without level of concern included. Where concern level is included in the models, we are analyzing how system and demand characteristics as well as drought severity influence variation in usage restrictions, holding constant the estimated longevity of a system’s current water supply. To measure drought severity, we used weekly U.S. Drought Monitor maps produced by the National Drought Mitigation Center at the University of NebraskaLincoln in cooperation with the U.S. Department of Agriculture and the National Oceanic and Atmospheric Administration. The U.S. Drought Monitor uses a five-category scale to characterize drought intensity, ranging from abnormally dry conditions to exceptional drought. The variable thus is a six-point measure including areas unaffected by drought. We assigned scores to water systems by using GIS to match boundaries of drought classified areas for each week of our study period to the boundaries of water system service areas, as indicated on spatial maps provided by the TCEQ. Water systems received the drought score covering the plurality of the spatial area of their jurisdiction, so long as the majority of the spatial area fell into some category of drought. If the 14 majority of area fell outside the bounds of area classified as being in drought, the water system received a 0 score for drought severity in that week. To calculate variables measuring characteristics of the population served by the water system, we aggregated data collected at the Census block group levels up to boundaries of water system service areas, assigning blocks according to where the plurality of their spatial area lies. The included variables—the percentage of a water system’s customer base in poverty, percentage black, percentage Hispanic, percentage with college degrees, and the percentage of houses built after 1980—are characteristics that might affect water demand as well as preferences about government regulation. We used the same process to aggregate precinct-level vote returns up to water service areas. Our measure of Democratic composition is the average share of the vote for the Democratic party candidate across all contested gubernatorial and presidential elections between 2002 and 2010 (Ansolabehere, Palmer, and Lee 2014). Summary statistics for all our independent variables appear in Table 1. Because our hypotheses address relative timing of the implementation of usage restrictions, not simply to whether or not they get implemented, we employ event history, or survival, analysis. This technique allows us to model the hazard rate of implementation, or the likelihood that a water system will implement usage restrictions at a particular point in time, given that the restrictions are not in place already. We employ the Cox proportional hazards model, which makes no assumptions about the form of the baseline hazard rate, but assumes that covariates have a constant effect on the hazard rate over time. We will subject the assumption to testing in future versions of the paper. Figure 1 shows smoothed hazard functions for the adoption of restrictions: the left side 15 shows adoption of either voluntary or mandatory restrictions, and the right side mandatory restrictions only. We received data from the TCEQ about 307 public water systems, and after merging with geographic shapefiles gathered from the state and the U.S. Census Bureau, we retained 280 systems for analysis, 62 of which never adopted either voluntary or mandatory use restrictions during our time of analysis. Numerous water systems experienced multiple spells of restrictions, adopting and removing restrictions only to readopt them on a later date, producing 456 implementations of voluntary or mandatory usage restrictions and 271 implementations of mandatory restrictions only. Water systems are observed every time the value of a time-varying variable (i.e., voluntary or mandatory restrictions, level of concern, or drought severity) changes. Results Table 2 shows results from our analyses as hazard ratios, which can be interpreted relative to the baseline hazard rate of restrictions implementation when covariates in the model are scored zero. The first two columns show models predicting adoption of voluntary or mandatory restrictions, and the second two columns show models for mandatory restrictions only. For each type, we separately estimate models omitting and including a variable measuring a water system’s self-reported level of concern with respect to its remaining water supply. Unsurprisingly, level of concern has a large and highly significant hazard ratio, indicating that water systems facing more severe supply constraints are quicker to implement use restrictions. A one-unit increase in the five-point scale indicating level of concern is associated with 3.4 times as much risk that the system 16 will adopt some type of usage restrictions, and 3.7 times the risk of mandatory restrictions. Among our hypotheses, the most noticeable result is the consistently significant positive coefficient on drought severity. A one-unit increase in the Drought Monitoring index is associated with a 7% increase in adoption of voluntary restrictions during the time period, and a 9% increase for mandatory restrictions, hazard rates that shrink little when controlling for a water system’s level of concern about its remaining supply. Regardless of whether a system is experiencing diminished supply, therefore, systems where drought levels are more severe adopt usage restrictions more quickly.1 Figure 2 shows the difference in hazard functions between water systems not experiencing any drought and those in most extreme drought, based on the models in Table 2 that control for level of concern about remaining water supply. The data suggest that objective problem conditions are strongly associated with the timing and severity of usage restrictions. Findings for our hypotheses about water systems’ institutional and technical capacity are mixed. System size, as expected, is associated with faster adoption of restrictions, likely because large systems have the ability to forecast future shortages and enforce enacted restrictions. We find some support for our hypothesis that city water systems adopt restrictions faster than other types of systems. City water systems adopted mandatory restrictions at over 1.4 times the rate of special districts and private systems, after controlling for a system’s level of concern about its remaining supply. We take this 1 The relationship between drought and water supplies is highly variable across water systems. In a model not presented here, we find that a unit increase in drought severity more than doubles risk of a system having concern about its water supply (i.e., estimating that it has fewer than 180 days of supply remaining), but that result is only weakly significant (p<0.061). 17 is as suggestive evidence that cities may be more comfortable with regulatory authority and therefore able to enact policies that limit resource use even in the absence of immediate resource shortages. We do not find any differences between cities and other governance types when examining adoption of any type of restrictions. In light of Mullin’s (2009) finding that the effect of governing system may be conditional on problem severity, in models not shown here we also included an interaction between city governance and drought severity, but found null results. We find little evidence that a system’s supply and resiliency, as measured by storage and interconnections, slows the adoption of usage restrictions. Similarly, serving a larger percentage of wholesale rather than retail customers appears to have little relationship with a water system’s policy response to drought. Overall, these findings indicate that it is the capacity to make and enforce decisions that matters more than the physical capacity of the water system. Politics appears to have surprisingly little role in influencing the type of emergency decisions we are studying. Partisan composition of a water system has no significant effect on timing of voluntary or mandatory restrictions. With respect to water usage, results indicate that high consumption actually speeds the enactment of voluntary use restrictions, even after controlling for concern levels, but that result does not persist for mandatory limitations only. The results may reflect the difference in potential gains from restrictions, which tend to be higher in high-consumption water systems. We find several significant results among our variables measuring the demographic composition of a water system’s customer base, but the patterns are difficult to decipher. The most consistent result is that water systems serving newer communities, with a larger percentage of houses built after 1980, adopt both voluntary and mandatory 18 restrictions more rapidly. Several explanations could account for this finding: new housing may be associated with a larger percentage of discretionary water usage that is easier to scale back during drought emergencies; occupants of new housing may be more supportive of usage restrictions; or new housing communities may face supply constraints that are unmeasured by variables in the models. Unfortunately, our data do not allow us to evaluate these alternative explanations. With respect to population, a larger percentage of white residents and a smaller percentage of college graduates both are associated with a higher rate of adoption. Discussion Research on the diffusion of public policies across states or localities rarely addresses contexts in which exogenous conditions lend urgency to finding a policy solution. In the case of drought response, decision makers must balance competing pressures to respond quickly with usage restrictions that might protect water supplies against worsening conditions and waiting longer in order to protect users against potentially needless restrictions. In this paper, we examine how problem conditions, institutional capacity, and political interests influence a water system’s decisions to implement usage restrictions. Our findings suggest that system decision-making in drought response in better predicted by problem conditions and institutional capacity than by political concerns. The severity of drought conditions is the most consistent determinant of adoption of usage restrictions. Water systems were far quicker to adopt restrictions when they anticipated shortages in the water supplies, but even holding constant a system’s level of concern about its remaining water, more severe drought in a 19 community prompted water officials to adopt both voluntary and mandatory restrictions more rapidly. Institutional capacity, by several measures, predicts water system decisionmaking. City systems adopted mandatory restrictions faster than other types of systems, likely reflecting higher levels of authority and access to enforcement tools that are unique to general-purpose local governments and central to seeking a behavioral change from system users. Large water systems that have more financial, technical, and managerial capacity to evaluate drought threats responded more quickly with both types of policies, offering further evidence that water systems appear to respond more to objective conditions than to political constraints in managing drought emergencies. The specificity of the data leveraged in this study allows us to evaluate not only whether water systems attempt to enforce conservation, but when they choose to act. In future work, we plan to exploit this feature of the data in order to address spatial diffusion of restriction adoptions across communities. Using data about a community’s supply source will allow us to analyze how competition over a shared resource influences policy diffusion to neighboring jurisdictions and test differing theories of policy diffusion against one another. 20 References Ansolabehere, Stephen, Maxwell Palmer, and Amanda Lee. 2014. “Precinct-Level Election Data.” http://hdl.handle.net/1902.1/21919 UNF:5:5C9UfGjdLy2ONVPtgr45qA== Harvard Election Data Archive [Distributor] V1 [Version] Berry, Frances S. and William D. Berry. 1990. “State Lottery Adoptions as Policy Innovations: An Event History Analysis.” American Political Science Review 84: 395-415. Berry, F.S. and W.D. Berry. 1999. “Innovation and Diffusion Models in Policy Research.” In Theories of the Policy Process, ed. P. Sabatier. Boulder, CO: Westview Press, 169–200. Betsill, Michele M. and Harriet Bulkeley. 2007. “Looking Back and Thinking Ahead: A Decade of Cities and Climate Change Research.” Local Environment: The International Journal of Justice and Sustainability 12(5): 447-456. Brown, Rebekah R. 2005. “Impediments to Integrated Urban Stormwater Management: The Need for Institutional Reform.” Environmental Management 36(3): 455-68. Bulkeley, H. 2010. “Cities and the Governing of Climate Change.” Annual Review of Environment and Resources 35: 229-253. Bulkeley, H. and M.M. Betsill. 2005. “Rethinking Sustainable Cities: Multilevel Governance and the 'Urban' Politics of Climate Change.” Environmental Politics 14(1): 42-63. -- 2003. Cities and Climate Change: Urban Sustainability and Global Environmental Governance. New York: Routledge. Burby, Raymond J. 2006. “Hurricane Katrina and the Paradoxes of Government Disaster Policy: Bringing about Wise Governmental Decisions for Hazardous Areas.” Annals of the American Academy of Political and Social Science, 604: 171-191. Daley, Dorothy M. 2007. “Voluntary Approaches to Environmental Problems: Exploring the Rise of Nontraditional Public Policy.” Policy Studies Journal 35: 165-180. Feiock, Richard C. 2004. "Politics, Institutions and Local Land-use Regulation." Urban Studies 41(2): 363-375. Fu, Xinyu, and Zhenghong Tang. 2013. “Planning for drought-resilient communities: An evaluation of local comprehensive plans in the fastest growing counties in the US.” Cities 32: 60-69. 21 Fu, X., et al. 2013. "Drought planning research in the United States: An overview and outlook." International Journal of Disaster Risk Science 4(2): 51-58. Gerber, Brian J., David B. Cohen, Brian Cannon, Dennis Patterson and Kendra Stewart. 2005. “On the Front Line: American Cities and the Challenge of Homeland Security Preparedness.” Urban Affairs Review 41: 182-210. Gerber, Elisabeth. R., and Justin H. Phillips. 2004. “Direct democracy and land use policy: Exchanging public goods for development rights.” Urban Studies 41: 463479. Hughes, Sara, and Megan Mullin. ___. Local Water Politics. In Oxford Handbook on Water Politics and Policy, eds. Ken Conca and Erika Weinthal. Oxford: Oxford University Press. Ivey, J. L., J. Smithers, R.C. De Loe, & R.D. Kreutzwiser. 2004. “Community capacity for adaptation to climate-induced water shortages: Linking institutional complexity and local actors.” Environmental Management 33: 36-47. Kenney, Douglas S. 2014. “Understanding utility disincentives to water conservation as a means of adapting to climate change pressures.” Journal of American Water Works Association 106(1): 36-46. Kenney, D.S., Christopher Goemans, Roberta Klein, Jessica Lowrey, and Kevin Reidy. 2008. “Residential Water Demand Management: Lessons from Aurora, Colorado.” Journal of the American Water Resources Association 44(1): 192-207. Kenney, D.S., Roberta A. Klein, and Martyn P. Clark. (2004). “Use and Effectiveness of Municipal Water Restrictions During Drought in Colorado.” Journal of the American Water Resources Association 40(1): 77-87. Knutson, C. L. 2008. "The role of water conservation in drought planning." Journal of Soil and Water Conservation 63(5): 7-154A,155A,156A,157A,158A,159A,160A. Krane, Dale A. 2002. “The State of American Federalism, 2001-2002: Resilience in Response to Crisis.” Publius: The Journal of Federalism 32(4): 1-28. Krause, Rachel M. 2011. “An assessment of the greenhouse gas reducing activities being implemented in US cities.” Local Environment 16(2): 193-211. Krause, Rachel M. 2010. “Policy Innovation, Intergovernmental Relations and the Municipal Adoption of Climate Protection Initiatives.” Journal of Urban Affairs 33(1): 45-60. 22 Lach, Denise H., Steve Rayner, and Helen Ingram. 2005. “Taming the Waters: Strategies to Domesticate the Wicked Problems of Water Resource Management.” International Journal of Water 3: 1–17. Lowry, William R. 1992. The Dimensions of Federalism: State Governments and Pollution Control Policies. Durham, NC: Duke University Press. Lowry, W.R. 2005. “Policy Reversal and Changing Politics: State Governments and Dam Removals.” State Politics & Policy Quarterly 5 (4): 394-419. Lubell Mark, Richard C. Feiock, & Edgar Ramirez. 2005. “Political Institutions and Conservation by Local Governments.” Urban Affairs Review 40 (6): 706-729. Lubell, Mark, Richard C. Feiock and Edgar Ramirez 2009. “Local Institutions and the Politics of Urban Growth,” American Journal of Political Science 53(3): 649-65. May, Peter J., and Thomas A. Birkland. 1994. "Earthquake Risk Reduction: An Examination of Local Regulatory Efforts." Environmental Management 18: 923-939. May, P. J. and J. Handmer 1992. “Regulatory Policy Design: Cooperative Versus Deterrent Mandates.” Australian Journal of Public Administration 51(1): 43-53. Measham, T., et al. 2011. “Adapting to climate change through local municipal planning: barriers and challenges.” Mitigation and Adaptation Strategies for Global Change 16(8): 889-909. Mintrom, Michael. 1997. “The State-Local Nexus in Policy Innovation Diffusion: The Case of School Choice.” Publius-The Journal of Federalism 27 (3): 41–59. Mullin, Megan. 2009. Governing the Tap: Special District Governance and the New Local Politics of Water. Cambridge, MA: MIT Press. Perrenod, Virginia Marion. 1984. Special Districts, Special Purposes: Fringe Governments and Urban Problems in the Houston Area. College Station: Texas A&M University Press. Pirie, R.L., Loe, R.C. and Kreutzwiser, R. 2004. “Drought Planning and Water Allocation: An Assessment of Local Capacity in Minnesota.” Journal of Environmental Management 73: 25–38. Porter, Douglas R., Ben C. Lin, and Richard B. Peiser. 1987. Special Districts: A Useful Technique for Financing Infrastructure. Washington, DC: Urban Land Institute. Portney, Kent E. 2003. Taking Sustainable Cities Seriously. Cambridge: MIT Press. 23 Ramirez de la Cruz, Edgar E. 2009. “Local Political Institutions and Smart Growth: An Empirical Study of the Politics of Compact Development,” Urban Affairs Review 45 (2): 218-246. Sharp, Elizabeth B., Dorothy M. Daley, and Michael S. Lynch. 2011. “Understanding Local Adoption and Implementation of Climate Change Mitigation Policy.” Urban Affairs Review, 47(3): 433-457. Texas Commission on Environmental Quality. 2005. “Handbook for Drought Contingency Planning for Retail Public Water Suppliers.” RG-424. Retrieved from http://www.tceq.texas.gov/assets/public/comm_exec/pubs/archive/rg424.pdf. Texas House Committee on Natural Resources. 2013. “Interim Report to the 83rd Texas Legislature.” Retrieved from http://ftp.weat.org/govt/2013HouseNaturalResourcesInterim-Report.pdf Texas Water Development Board. 2013. “Summary of Groundwater Conditions in Texas: Recent (2011-2012) and Historical Water- Level Changes in the TWDB Recorder Network.” Retrieved from http://www.twdb.state.tx.us/publications/reports/technical_notes/doc/TechnicalNote_ 13-02_GW_Recorder2012.pdf. Texas Water Development Board. 2014a. “Water Data for Texas.” Retrieved from http://waterdatafortexas.org/reservoirs/statewide. Texas Water Development Board. 2014b. “Texas Water Conditions Report: February 2014.” Retrieved from http://www.twdb.texas.gov/publications/reports/waterconditions/twc_pdf_archives/20 14/twcFeb2014.pdf. Thomas, Robert D. and Richard W. Murray. 1991. Progrowth Politics: Change and Governance in Houston. Berkeley, CA: IGS Press. Tiebout, Charles. 1956. “A Pure Theory of Local Expenditures.” Journal of Political Economy 44: 416-24. Wilhite, D. A., et al. 2000. “Planning for Drought: Moving from Crisis to Risk Management.” Journal of the American Water Resources Association 36(4): 697-710. Zahran, Sammy, Samuel D. Brody, Arnold Vedlitz, Himanshu Grover, and Caitlyn Miller. 2008. “Vulnerability and capacity: explaining local commitment to climatechange policy.” Environment and Planning C: Government and Policy 26: 544-562. 24 Figure 1. Hazard Rates of Water Use Restriction Adoptions Curves show smoothed hazard functions based on the models that include level of concern (columns 2 and 4) in Table 2 and estimated with values of covariates at their means. 25 Figure 2. Drought Severity and Water Use Restriction Adoptions Curves show smoothed hazard functions based on the models that include level of concern (columns 2 and 4) in Table 2 and estimated with values of covariates at their means. 26 Table 1. Summary Statistics Voluntary or mandatory restrictions Mandatory restrictions Drought score Groundwater Population served (logged) Average daily consumption per 1k people Total storage per 1k people % customers served wholesale Number of interconnections % poverty % houses built after 1980 % black % Hispanic % with 4-year college degree City water system Level of concern Mean 0.430 0.267 2.166 0.125 9.912 0.137 Std. Dev. 0.495 0.442 1.431 0.322 1.144 0.060 Min. 0 0 0 0 -1.093 0 Max. 1 1 5 1 15.357 0.481 0.215 0.160 3.241 15.250 57.631 9.583 29.261 25.768 0.688 0.476 0.121 0.295 5.121 9.571 23.990 10.963 23.228 15.628 0.463 0.536 0 0 0 1.504 6.186 0 0.733 4.983 0 0 0.773 1 53 49.096 99.027 68.053 98.927 81.934 1 4 27 Table 2. Predicting Water Use Restrictions Voluntary/Mandatory Mandatory Only 1.072** (3.32) 1.051** (2.76) 1.091** (2.72) 1.078** (2.53) Population served (logged) 1.023 (0.87) 1.108** (3.71) 1.113** (2.02) 1.179** (3.44) Total storage per 1k people 0.676 (-1.10) 0.626* (-1.71) 1.633 (0.84) 1.780 (1.04) Number of interconnections 1.000 (0.05) 1.001 (0.20) 1.000 (-0.02) 1.001 (0.10) City water system 1.030 (0.26) 1.100 (1.09) 1.309* (1.67) 1.419** (2.55) % Customers served wholesale 1.103 (0.57) 0.867 (-0.93) 0.811 (-0.64) 0.663 (-1.29) 11.52** (3.15) 4.036** (2.11) 3.373 (1.08) 0.881 (-0.13) % Democrat 1.006 (0.94) 0.998 (-0.31) 1.014 (1.42) 1.004 (0.47) Groundwater 1.066 (0.50) 0.942 (-0.43) 1.074 (0.32) 0.901 (-0.50) % Poverty 0.991 (-1.17) 1.001 (0.18) 0.970** (-2.11) 0.980 (-1.56) % Houses built after 1980 1.004* (1.67) 1.006** (3.03) 1.008** (1.99) 1.010** (2.52) % Black 0.998 (-0.32) 1.000 (0.04) 0.977** (-2.22) 0.982** (-2.10) % Hispanic 0.992* (-1.90) 0.990** (-2.93) 0.986** (-2.10) 0.985** (-2.63) % 4-year college degree 0.998 (-0.65) 0.993** (-2.29) 0.986* (-1.88) 0.981** (-2.51) Drought score Average daily consumption per 1k people 3.412** (9.93) Level of concern N N water systems 3.710** (10.12) 10817 10817 10817 10817 280 280 280 280 Cells show hazard ratios from Cox proportional hazard models with robust standard errors in parentheses: * p<0.10, ** p<0.05. 28