Dams are old and abandoned, are people concerned? Risk perceptions from Michigan, USA

doi:10.21203/rs.3.rs-4022111/v1

Download PDF

Research Article

Dams are old and abandoned, are people concerned? Risk perceptions from Michigan, USA

https://doi.org/10.21203/rs.3.rs-4022111/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Dams across the U.S. are in a precarious state. Many have been in disrepair for decades and the entities that built them have long since disappeared, raising difficult questions as to who will manage this growing problem. Despite the state of the nation’s dams, we know very little about public perceptions of dams. In this analysis, we evaluate public concern for dams using survey data from Michigan, a state with an extensive network of aging dams. Using a survey experiment, we find that risk perceptions increased when respondents were provided information about the age of the dams. Respondents identified as “news junkies” also expressed higher risk perceptions. We also tested for “white male” and partisan effects on risk perception but found little evidence of those. We conclude by discussing implications for policy and the future of dam management.

There are nearly 92,000 dams in the United States. Many are over 50 years old, with some dating back over 100 years, and were built to support industries that either severely contracted or no longer require dams for their core operations. Often, the companies that built the dams have long since ceased to exist, and management and maintenance has fallen to state governments that do not have adequate resources to inspect and repair failing dams. Michigan is emblematic of this problem, as the state is home to 2,600 dams—most constructed to support mining and logging industries that either transitioned away from water-intensive operations or have simply disappeared. Most of the dams are assessed as being at high risk of breaching by state regulators. However, many of the existing dams have become an important source of recreation around the reservoirs, and valued second homes for leisure and vacation, despite the danger of breaching that they represent. This motivates this paper and its focus on risk perception.

There is a decades-long tradition of research into risk perceptions around infrastructure, chiefly focusing on new or proposed projects (Freudenburg and Davidson 2007; McAdam and Boudet 2012; Kahan et al. 2007; Dake 1992), but comparatively little research on redundant, abandoned, idled, and abandoned infrastructure (RAIN; see Fisk, Mayer, and Morris 2023). Most of Michigan’s dams were constructed before the risk perception literature even existed, but this work still provides insights that are useful for the current analysis. Indeed, there is a paucity of risk perception research on dams, perhaps because researchers are more apt to study emerging or more publicly salient technologies, not technologies that have outlived their useful life. Still, we can draw some insights from the risk perception literature.

Despite a few high-profile breaches like the ones occurring in Gladwin and Midland counties in 2020, we suspect that Michigan residents know relatively little about the state’s dam infrastructure and its current state of disrepair. There has been little attention paid in Michigan, in contrast to other states where active dam removal projects are underway in the Northeast and in the Western US. In this paper, we evaluate how providing information cues about Michigan’s dams shapes risk perception. We suspect that Michigan residents are not aware of the sheer number of dams, or of their condition, and being informed about this may heighten risk perceptions. Secondly, we suspect that Michigan residents are unaware of the condition and age of the state’s dams and priming them with this information will heighten concerns. In the next section, we describe the history of dams in Michigan, and then transition to a discussion of the relevant literature on risk.

Starting in the 1840s, metallic mining interests found that Michigan’s Upper Peninsula was ideal for mining because of its resource abundance and free-flowing rivers with steep gradients that made hydropower effective. Dams were constructed at mill sites to power stamp mills to break up metallic ores (Courter, 1992). Michigan’s copper and iron were vital to the economic development and industrialization of the U.S. (Schaetzl, 2004). Roughly 11 billion pounds of copper and over a billion tons of iron ore were mined and shipped out of Michigan, which brought in billions of dollars (Schaetzl, 2004).

Agriculture relied heavily upon dams to grind grains. These dams turned wheat, rye, and corn into flour. The mills had a small dam placed upstream to hold back water, effectively providing a consistent source of water that would move the wheels that drove the grinding stones. Simple grist mills for cereal production gave way to industries such as the breakfast cereal boom, where Kellogg, Post, and General Mills molded the economy in towns like Battle Creek and which made Michigan agriculture important to the national diet.

Michigan’s river systems and dams also facilitated the development of a large forestry industry. The state’s extensive network of rivers and tributaries reaching far into the interior provided an easy and inexpensive mode of transportation of raw materials down river to mills (Fuller, 1916). The first sawmill was constructed in 1832, and by 1873 there were 1600 sawmills in operation (Michigan State University Extension, 2022). Dams were used to power and support the industry from its genesis in the 1840s well into the 1870s, when the industry largely switched to steam power (Schaetzl, 2004). The city of Grand Rapids, MI is an exemplar of this evolution, moving from an economy based on logging and sawmills, to an important urban center which installed the earliest hydropower cinema, and a thriving economy including a big banking center.

Some of those same dams were converted to hydroelectric generation after their initial industrial purposes faded. There are over 100 dams that are retired hydroelectric facilities, leaving around 89 hydroelectric dams producing energy from Michigan’s water (The Public Consultants, 2007). Construction of new hydroelectric facilities ceased, as none have been built since 1973 (NID, 2022). This was also true across the country, as concerns with the social and environmental impacts of dams led to growing regulation. Figure 1 shows the current purposes of Michigan’s dams—note that 11% currently produce power, and most (67.4%) now exist to support recreation.

Figure 1. Current purpose of Michigan’s dams. Source: National Inventory of Dams (https://nid.sec.usace.army.mil/#/)

In May 2020, sustained heavy rains drenched central Michigan (National Weather Service, 2020). In Gladwin and Midland counties, the heavily dammed Tittabawassee River swelled. The Edenville dam ruptured, sending millions of gallons billowing down the Tittabawassee into Wixom Lake, in turn breaching the Sanford Dam. The Village of Sanford and the City of Midland lost over 2500 structures and suffered $200 million in damage (Hayes, 2020). Both dams were constructed in the 1920s and had been listed as “high hazard” (NID, 2020). In 2004, FERC had informed Boyce Hydro—the prior owner of the dams—to fix the spillways, but the company ignored these mandates (Colaianne and Kepler 2022). In 2018, Boyce Hydro’s permit to generate power was revoked and management shifted to the state, to eventually transfer ownership to local stakeholders. Colaianee and Kepler (2022) argue that a similar fate may await many other dams in Michigan unless regulators act. Of course, regulators may not respond unless communities pressure them to do so, implying that studying public perspectives is important.

Michigan is in this situation because of multiple factors. By 2023, 80% of the state’s dams will be over 50 years old with 271 that are over 100 years old, well past their useful life (ASCE, 2018). The state employs only three full-time inspectors—there are roughly 342 dams per inspector, significantly higher than most other midwestern states. The inspectors are given a budget of $339 per dam, below the national average of $695 (Michigan Department of Environment, Great Lakes, and Energy 2021; Matheny, 2020). Compared to other states, Michigan’s budget is almost $100,000 less than the next lowest budget, when the number of dams being managed is somewhat comparable (Wilson, 2020). There are 251 dams located upstream of communities with populations of 5000 or more, with 107 located within a mile of a population center (The Public Sector Consultants, 2007).

Climate change exacerbates the precariousness of Michigan’s dams. Michigan is now significantly wetter. Heavy precipitation is likely to rise in “frequency and intensity”; “heavy precipitation events above the 99th percentile of daily values” increased by 42% in the Upper Midwest, fueled by more moisture in a warming atmosphere (Hayhoe et. al., 2018, p. 88). Older dams in disrepair may not be able to handle this volume of water. Climatic shifts, coupled with the age and poor maintenance of the dams, render the state especially vulnerable to future breaches.

Governor Gretchen Whitmer has emphasized infrastructure and allocated $19 million for a dam emergency fund (Office of Governor Whitmer, 2021). It will be used to update and replace dams and to adapt them to climate change. Although this is a step in the right direction, much more money will have to be invested to ensure dam safety. There is a $50 million need in the immediate future for dam repair and/or rehabilitation for 120 dams and an additional $225 million to address future needs in infrastructure and technology updates (The Public Sector Consultants). The state increased the number of inspectors from 2 to 3 only recently, after the Midland breach (Michigan Department of Environment, Great Lakes, and Energy, 2021) and created a 19-member group to address the numerous failures and precarious state of dam infrastructure in Michigan. This group focused on creating a list of necessary steps to create a safe dam infrastructure. These recommendations focus on 8 areas: Funding for Dam Maintenance, Repair, and Removal; Legislation and Authority; Improving Dam Safety; Compliance and Enforcement; Emergency Response; Program Management, Funding, and Budgeting; Safety and Security at Dams; and Outreach and Awareness.

The 2600 dams statewide listed in the National Inventory of Dams are grouped into 10 categories: recreation, hydroelectricity, fish and wildlife, irrigation, flood control, water supply, navigation, fire production/ stock/ or small fishpond, debris control, tailings, and a miscellaneous category (NID 2022). The dams that failed in Gladwin and Midland counties were used for flood control and hydroelectric power. In Michigan’s lake country, reservoirs increasingly provide recreation opportunities and attract second-home owners. This was true of the dams along the Tittabawassee, which are hosts to a lively and vital outdoor recreation and boating culture. The dam reservoirs are lined with housing, the lakes filled with boats, docks, and floats. However, the reservoirs associated with the breached dams currently have much lower water levels than before, and residents likely hope that they will be restored to previous levels.

Gladwin county is heavily dependent upon tourism. During the summers, 21% of the people present in the county stay in these seasonal homes. Additionally, just under 40% of houses in the county are seasonal homes (Gladwin County Chamber of Commerce 2022). The economic impact is clear. This raises difficult questions about how to balance the cultural and economic benefits of dams with the growing hazard that poorly maintained dams represent to both life and property.

Michigan’s dams will increasingly pose a challenge for the state (Department of Energy, Great Lakes, and Energy, 2022). The vast majority are or will have exceeded the recommended lifespan. This is a very precarious situation, and the state has not provided the necessary resources to effectively manage, inspect, and enforce regulations.

Civil society groups have successfully organized to remove some dams. The federal government has encouraged the removal of hydroelectric dams that are no longer productive (Amos 2014; Uría-Martínez, Johnson, and Shan 2020). Perhaps the most notable example of a removal in Michigan is the Lyons Dam. With the combined efforts of local stakeholders, Michigan State University, state regulators, and consultants, a report was produced to convince the state government to act (Hegarty, 2008, p. 207). This long battle ended with the removal of the 160-year-old dam, restoring connection between upper and lower portions of the river, freeing the village from the threat of a breach and flooding, and allowing for a wider array of riverine recreational activities. The removal of the Lyons Dam provides a template for change before other places experience dam failures. It appears that Michigan residents, at least in some situations, have been concerned about old, poorly maintained dams and their potential impacts and acted upon this knowledge.

Although risk perception around dams have received little study, there are several well documented predictors of risk perceptions related to environmental and public health risks. For instance, several studies have documented a “white male” effect on risk perception, wherein white males tend to have lower risk perceptions than other demographic groups (Flynn, Slovic, and Mertz 1994). White males perceive lower risks from nuclear power and firearms (Kahan et al. 2007; Palmer 2003) heavy industry like chemical plants and refineries (Campbell, Bevc, and Picou 2013; Marshall et al. 2006) and climate change (McCright and Dunlap 2011). Conservative political ideology and affiliation with the Republican party is associated with lower risk perceptions for oil and gas development (Howell et al. 2019; Mayer 2016; 2018) and some of the “white male” effect on risk perception occurs because white males are more likely to identify as “conservative” (McCright and Dunlap 2011). White males may have lower risk perceptions because of their position of relative privilege—they may experience less exposure to environmental harms than minoritized groups (Palmer 2003). Further, because white males are in a position of privilege, they are more likely to support the existing economic order that produces environmental injustices (Goldsmith, Feygina, and Jost 2013; Feygina, Jost, and Goldsmith 2010). Still, it is unclear if similar white male and ideological effects would be present for dam risk perceptions.

In addition to socio-demographic and ideological factors, risk perceptions are influenced by media framing (Kasperson et al. 1988; Blair, Heikkila, and Weible 2016; Kristiansen 2017). Media frames can influence people’s perceptions of issues by “select[ing] some aspects of a perceived reality and mak[ing] them more salient … in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation” (Entman 1993, pg. 52). The social amplification of risk framework centers the role of experts and media in creating awareness of risks and encouraging protective action. For instance, Frewer, Miles, and Marsh (2002) found that perceptions of GMOs in the UK tracked media coverage. Media (and social media) amplified and dis-amplified risks during the early stages of the COVID-19 pandemic (Lee, Choi, and Britt 2023; Zhang and Cozma 2022). One possible mechanism for this relationship between media coverage and perceptions of risk is the theory of agenda-setting, which posits that topics which are highlighted in news media elevate the salience of those topics in audience’s minds and can influence their relevant knowledge and attitudes (Weaver 1991). Thus, extant research implies that media plays a large role in risk perceptions.

Notably, there is relatively little research on media coverage of dams, and we suspect that dams receive relatively infrequent media coverage. The initial breach of the Edenville dams prompted a handful of articles in local and regional outlets, but there is little sustained coverage (e.g. (Schafer 2020; Matheny, n.d.; Murdock 2020). In the peer-reviewed literature, Brondizio et al. (under review) consider 100 years of coverage of hydropower dams in two major newspapers with national circulation– the Los Angeles Times and the New York Times. They report that coverage gradually shifted from stories of hydropower dams’ ability to generate energy, towards concerns about environmental and social impacts. Roy et al. (2022) considered dams in the New England region of the U.S and find that stories about dam removal are not more prevalent in areas with more dams. We are not aware of any research that connects media consumption to risk perceptions in the context of dams.

Relatively few studies have considered risk perceptions from dams in the U.S., although there is a large literature on hydroelectric projects in the Global South (e.g. Moran et al. 2018; Mayer et al. 2021; Siciliano et al. 2019). Indeed, we are aware of no studies like ours that consider risk perceptions related to multiple, multi-use, aging and even abandoned dams, such as those in Michigan. Given the age of the dams, their presence is likely taken for granted, or even unknown. This lack of familiarity may increase uncertainty about preparedness actions, and may increase reliance on outside sources for information (Eiser et al. 2012). Past experiences with hazards can also increase risk perceptions, as the hazard does not seem abstract or unlikely (Blanchard-Boehm and Cook 2004; Becker et al. 2017; Bronfman et al. 2020; Castañeda et al. 2020).

Given the lack of dam-specific research, we explore research questions related to information, media, and white male and ideological effects in this study. We suspect that priming respondents with information about the age and the sheer number of Michigan’s dams will increase risk perceptions (Research Question 1). For research question 2, we investigate possible white male and ideological influences on dam risk perceptions. Finally, given the role of media as posited by the social amplification framework, we ask how patterns of media consumption shape risk perceptions (Research Question 3). In the next section, we describe the data we used to evaluate these research questions.

Data Collection

Data was provided by Qualtrics, a well-known provider of survey data—the survey is part of a larger project to understand the partisanship, natural resources, and development issues in non-metropolitan areas of Michigan for that reason we excluded counties from the ten most population dense counties in the state: Wayne, Macomb, Oakland, Kent, Genessee, Ottawa, Washtenaw, Ingram, Kalamazoo, and Muskegon. Research indicates that online panels provide quality data that tends to produce results that are similar to true probability sample, especially if speedier and straight-liners are excluded (Roulin 2015; Peer et al. 2022; Porter et al. 2019; Cornesse and Blom 2023; Cornesse et al. 2020). We note, of course, that this data is not a true probability sample, even if our core results might be like those that we would attain from a true probability sample. Qualtrics typically recruits respondents via a dashboard and they receive a small cash reward or rewards points. Data was collected between August 17 and September 4, 2023.

We performed a series of procedures to ensure data quality. Eight respondents failed a simple attention check and were excluded from the analysis. We also excluded respondents who completed the survey in an implausibly short time (in this case, we removed 38 respondents who finished the survey in less than 5 minutes. We also evaluated our data for evidence of straight-lining but did not find patterns that suggested that straight-lining was an issue. After correcting these problems, we had 1,212 valid cases to use in our analyses below.

Outcome Variable: Concern for Dams

We used a survey experiment to evaluate how information treatments about the state of Michigan’s dams might shape concern. Treatments were randomized among respondents, such that each respondent would only answer one question. For our control condition, we asked “As you may know, Michigan is home to many dams. How concerned are you that dams will fail and cause flooding?”. For the first experimental conditions, respondents were told that 80% of the states’s dams were over 50 years old. For the final experimental condition, respondents read that the state was home to over 2,600 dams.

Although we suspected that respondents knew relatively little about the state’s dams, Fig. 1 implies that the experimental conditions did not shift levels of concern substantially. For example, compared to the control condition, the “age” treatment increased “very concerned” from 7.2–12% when respondents were told that the dams were over 50 years old, well beyond their engineering lifetimes However, priming respondents about frequency produced very similar results as the control condition, with only 7.1% stating that they were “Very concerned”.

Predictor Variables

Early we discussed the well-documented “white male” effects on risk perceptions. We include binary variables for sexual identity (0 = all others, 1 = male), and race (0 = non-white, 1 = white). Political ideology is measured with 3 categories (0 = conservative, 1 = neither, 2 = liberal) and party affiliation also had 3 categories (0 = Republican, 1 = independent, 2 = Democrat).

For media consumption, we adapted a set of vetted questions used in multiple studies of the U.S. media (e.g. Mourão et al. 2018; Edgerly 2015; Weeks, Ksiazek, and Holbert 2016). Factor analysis (see Appendix B) implied that there were three underlying dimensions to these items: consumption of local and national news, news junkies (that is, those who consume a lot of news content from a wide variety of sources), and consumers of conservative media. We used factor scores for these different types of consumption in our regression models. We reproduced a series of questions about social media consumption from Pew (Anderson 2021) to capture social media usage and conducted a factor analysis of these items, which strongly indicated a single factor solution (Appendix C). Factor scores for media consumption and social media usage serve as predictors in our regression equation. Table 1 provides descriptives statistics for all the variables used in this study.

Table 1

Descriptive Statistics
Income
less than $24,999	341 (28.1%)
$25,000 to $49,999	410 (33.8%)
$50,000 to $74,999	234 (19.3%)
$75,000 to $99,999	110 (9.1%)
$100,000 to $149,999	82 (6.8%)
$150,000 or more	35 (2.9%)
Education
less than HS	57 (4.7%)
High School	392 (32.3%)
Associates	150 (12.4%)
Some College	340 (28.1%)
College Grad	273 (22.5%)
Political Ideology
Conservative	361 (29.8%)
Neither	622 (51.3%)
Liberal	229 (18.9%)
Political Party Affiliation
Republican	460 (38.0%)
Independent	362 (29.9%)
Democrat	390 (32.2%)
Male	0.209 (0.407)
White	0.908 (0.290)
Conservative Media	0.856 (0.763)
Social media	1.768 (0.626)
News Junkie	2.943 (0.838)
Local and National News	1.049 (0.824)

Modelling Strategy

Our dependent variable—dam risk perceptions—is ordinal, so we use ordinal logistic regression to model the effects of the predictors on dam risk perceptions. We first estimate a model with only the informational treatments and the controls for education and income. In the second model, we add variables for white and male status along with political ideology and political affiliation, and in the next two models use interaction terms to test for “white male” and “white conservative” effects. In the fifth and final model, we add the media consumption variables.[1]

Modelling Results

In Table 2, we provide coefficients (flagged for statistical significance), 95% confidence intervals, and AIC and BIC statistics for our ordinal logistic regression models. We first discuss our experimental informational treatments. Priming respondents about the age of Michigan’s dams increased concern, an effect that was consistent across all models, but pointing out that Michigan has over 2,600 dams did not approach statistical significance in any model.

Political ideology and party affiliation do not operate in the expected manner, as we anticipated a stark difference between liberals and conservatives. In models 2, 3, and 5, “neither liberal nor conservative” is associated with lower concern than conservative and in Model 4 the main effect of “neither” is not statistically significant (that is, the effect of “neither liberal or conservative” for the non-male category of sex). Notably, the interaction between “neither” and “male” is statistically significant—we do not find evidence of a “male conservative” effect on dam risk perceptions, but it does appear that sex may partially condition the effect of political ideology. For party affiliation, we find that Democrats tend to perceive more risk than Republicans, although the coefficient loses its statistical significance in Model 5 when the variables for media consumption were entered. Although prior work has documented “white male” effects on risk perception wherein white males perceive lower risk, our models imply that male sex and white race have little influence on risk perceptions, and their interaction was not statistically significant. We return to this somewhat surprising finding in the discussion.

Finally, we consider the media and social media variables entered in Model 5. The News Junkie and Local and National News scales are statistically significant and, notably, the effect of conservative media consumption is not. This model has also produced the lowest AIC and BIC statistics, implying that it fits the data better than the earlier, simpler models.

Table 2

Ordinal Logistic Regression Models
		Model 2	Model 3	Model 4	Model 5
	b/ci95	b/ci95	b/ci95	b/ci95	b/ci95
Informational Treatments
Old	0.649***	0.636***	0.629***	0.652***	0.624***
Old	[0.339,0.958]	[0.325,0.946]	[0.317,0.941]	[0.340,0.964]	[0.308,0.940]
Frequent	0.028	-0.02	-0.023	-0.027	-0.042
Frequent	[-0.306,0.362]	[-0.358,0.318]	[-0.360,0.315]	[-0.365,0.312]	[-0.380,0.296]
Income (reference = less than $24,999)
$25,000 to $49,999	-0.126	-0.12	-0.122	-0.123	-0.176
$25,000 to $49,999	[-0.459,0.207]	[-0.461,0.220]	[-0.466,0.222]	[-0.463,0.216]	[-0.522,0.170]
$50,000 to $74,999	-0.284	-0.267	-0.272	-0.266	-0.304
$50,000 to $74,999	[-0.671,0.104]	[-0.664,0.130]	[-0.668,0.123]	[-0.662,0.130]	[-0.712,0.104]
$75,000 to $99,999	-0.162	-0.151	-0.148	-0.146	-0.159
$75,000 to $99,999	[-0.708,0.385]	[-0.699,0.397]	[-0.694,0.399]	[-0.688,0.397]	[-0.692,0.374]
$100,000 to $149,999	0.036	0.05	0.057	0.084	0.069
$100,000 to $149,999	[-0.598,0.669]	[-0.599,0.698]	[-0.586,0.700]	[-0.569,0.736]	[-0.612,0.750]
$150,000 or more	-0.177	-0.093	-0.116	-0.034	-0.187
$150,000 or more	[-1.259,0.904]	[-1.184,0.998]	[-1.214,0.983]	[-1.127,1.060]	[-1.243,0.870]
Education (reference = less than HS)
High School	-0.168	-0.321	-0.309	-0.369	-0.406
High School	[-0.780,0.443]	[-0.916,0.274]	[-0.908,0.291]	[-0.960,0.221]	[-0.987,0.174]
Associate's Degree	0.23	-0.027	-0.008	-0.075	-0.131
Associate's Degree	[-0.438,0.897]	[-0.687,0.634]	[-0.669,0.654]	[-0.734,0.583]	[-0.768,0.507]
Some college	0.005	-0.189	-0.174	-0.235	-0.329
Some college	[-0.613,0.622]	[-0.789,0.412]	[-0.778,0.430]	[-0.830,0.360]	[-0.918,0.260]
College graduate	0.328	0.05	0.057	0.001	-0.083
College graduate	[-0.353,1.009]	[-0.620,0.721]	[-0.613,0.728]	[-0.662,0.664]	[-0.730,0.565]
Political ideology (reference = Conservative)
Neither liberal nor conservative		-0.432*	-0.429*	-0.097	-0.442*
Neither liberal nor conservative		[-0.796,-0.068]	[-0.795,-0.064]	[-0.435,0.241]	[-0.816,-0.069]
Liberal		-0.183	-0.171	0.008	-0.195
Liberal		[-0.720,0.355]	[-0.710,0.368]	[-0.474,0.491]	[-0.740,0.351]
Party affiliation (reference = Republican)
Independent		0.16	0.151	0.169	0.227
Independent		[-0.211,0.530]	[-0.223,0.525]	[-0.203,0.540]	[-0.146,0.600]
Democrat		0.453*	0.440*	0.444*	0.366
Democrat		[0.037,0.870]	[0.022,0.858]	[0.027,0.861]	[-0.052,0.784]
Male		-0.213	-0.779	0.182	-0.259
Male		[-0.490,0.063]	[-1.561,0.003]	[-0.316,0.680]	[-0.538,0.020]
*Neither liberal nor conservative				-0.661*
				[-1.274,-0.047]
*Liberal				-0.329
				[-1.224,0.566]
White		0.115	-0.273	0.114	0.162
		[-0.359,0.589]	[-0.777,0.231]	[-0.361,0.589]	[-0.316,0.639]
* Male			0.626
			[-0.202,1.454]
Conservative Media					0.041
					[-0.197,0.280]
Social Media					0.019
					[-0.209,0.247]
News Junkie					0.372***
					[0.189,0.555]
Local and National News					0.197*
					[0.028,0.367]
AIC	2904.241	2891.182	2890.227	2888.004	2852.42
BIC	2975.641	2993.183	2997.327	3000.204	2974.82
Note. n = 1,212, * for p < 0.05, ** for p < 0.01, * for p < 0.001

Ordinal logistic regression models have well-documented problems of interpretation, and many methodologists recommend using marginal effects or predicted probabilities to render modelling results more intuitive (Mood 2010; Brambor, Clark, and Golder 2006; Hanmer and Ozan Kalkan 2013; Breen, Karlson, and Holm 2018). In Fig. 2, we provide the probability of being “Very Concerned” about dam failure across the scores of the News Junkie and Local and National news scales by the two different scores of the informational treatment for the dam’s age—these where the variables were among those that were statistically significant in Model 5. Both panels show a consistent gap between the treatment conditions of about 0.05. That is, the informational treatment about the age of the dams increased the probability of “Very Concerned” by about 5%. The probabilities increase in tandem with media consumption. For example, a person that scores low on the “News Junkie” factor score would have a probability of some 0.07, but this same estimate is 0.23 for those who score high on this variable. The effect of the “Local and National News” factor score is not as strong, with a difference of about 0.10 between low and high.

Robustness Checks

We use the konfound method to evaluate the robustness of our results (Xu et al. 2019; Frank et al. 2013). In our application, Konfound estimates the percentage of cases that would have to be measured with error to invalidate the inference—that is, to change a statistically significant coefficient to non-significant. Measurement error has many sources. For instance, respondents may not understand the question, might provide an answer that did not reflect their true attitudes, or the research team could make an error in coding and cleaning the data. Table 3 below implies that the “Old” information treatment is relatively robust, while the only category of political ideology that was statistically significant (“Neither liberal nor conservative”) was comparatively less robust at 15.41%. Among the media consumption variables, Local and National news was less robust than News Junkie (50.69% vs. 13.97%). Although some effects are less robust than others, we do not expect that our data contains a significant degree of misclassification or measurement error that would significantly change our results or alter our take-away conclusions.

Table 3. Konfound Analysis
	% Measurement Error
Old	49.31%
Neither liberal nor conservative	15.41%
New Junkie	50.69%
Local and National News	13.97%
Note: Table displays the percentage of cases that would have to be measured with error to change the coefficient to non-significant

The purpose of this paper is to understand risk perceptions related to Michigan’s rapidly aging dam infrastructure—outside of hydroelectric megaprojects (e.g., the Hoover dam in Nevada), dams have received little attention in the social science and policy literature and redundant, abandoned, and idled infrastructure has received comparatively little scholarship (Fisk, Mayer, and Morris 2023). Like other aging or obsolete infrastructure, dams pose a unique challenge because many are aging yet the builders were often companies that have long since shuttered.

Given the apparent low salience of dams, we evaluated if informational treatments about the age of the dams and the number of dams in the state would heighten risk perceptions (Research Question 1). We found that priming about the age of dams increased perceptions of risk, an effect that was robust across models and against measurement error, but providing information about the number of dams did not. Future research could explore alternative information treatments. Dams are a complex problem given that they have been used for diverse purposes and the original builders of the dams may be companies or other parties that no longer exist. Developing effective communication strategies for the public about the risks posed by dams is an important task for future research and practice. This may help encourage the dismantling of dams, or at least direct resources towards their repair, as the public may be unaware of the extent of the problem.

We tested for a “white male effect” and ideological and gender effects on dam risk perceptions (e.g. Campbell, Bevc, and Picou 2013; Marshall et al. 2006), but found no compelling evidence that white men, or those who identify as conservative, reported less concern about dams as a risk (Research Question 2). We can only speculate on this finding. It is possible that white males, particularly conservative white males, might be sorted into agricultural and related professions that could be significantly disrupted by dam breeches. Also, some white male and white male conservative effects have been observed for relatively high salience issues like climate change that have had significant media discourse and cues from elites (McCright and Dunlap 2011; Carmichael and Brulle 2017). There has been limited news coverage of Michigan’s dams, and few cues from partisan elites, hence a lack of polarization.

This lack of polarization is promising and implies that policy efforts to address the challenges posed by Michigan’s dams might not immediately face significant partisan barriers, at least from the public. To the best of our knowledge, there has been no significant resistance to Governor Whitmer’s efforts to study and mitigate the risks posed by dams. However, we acknowledge that there are difficult trade-offs in the future that may engender political conflict. For instance, housing developments along reservoirs have been constructed in multiple areas of the state. The dams that create those reservoirs are often in dis-repair, and it is unclear who will pay for maintaining the dams that have facilitated the lakefront developments. Further, if risks of dam failure became widely reported in news media, the issue might attract the attention of partisan elites, leading to ideological sorting and elite cues that encourage polarization around the issue. This pattern has been observed in the politicization of formerly less-controversial shifts to sustainable energy infrastructure such as wind and solar power. When right-wing influencers adopted a vocal stand against sustainability efforts that mirrored their rejection of climate change, communities in Michigan began to systematically resist such efforts (Mayer and Parks 2024; Crawford, Bessette, and Mills 2022).

Earlier we described the social amplification of risk framework, which highlights how media and discourse can amplify the significance of an event and intensify risk perceptions, or alternatively downplay hazards (or events related to hazards) and reduce risk perceptions (Research Question 3). We found mixed evidence for the role of media. News Junkies—that is, persons who consume a high volume of news from a variety of sources—and those who consume mainstream Local and National news had increased risk perceptions. This could be attributed to the likelihood that people who spend the most time-consuming news are more likely to encounter disaster-oriented stories around dams from any part of the world, which would make the potential failure of Michigan dams more salient when primed through the survey experiment. Predominant news values tend to encourage journalists to emphasize negative events (Parks 2019) and therefore news stories related to dams that news junkies are likely to encounter could very well skew toward failures, breaches, etc. Importantly, conservative news, which has often downplayed risks like climate change, was not associated with dam risk perceptions – again, quite likely because the issue simply has not been high on conservative media agendas and therefore not made salient to their audiences.

In the case of Michigan’s dams, the Edenville rupture may have acted as a focusing event (Kingdon 1984), temporarily increasing the salience of dams among those who consume local news, given that the ruptures were not high-profile national issues. Beyond local news, it is possible that media coverage of the rupture was limited, and many of our respondents were simply not aware that it had occurred. We are aware of no studies that have tracked media coverage of a dam rupture in the U.S., or issues surrounding dams more generally. We suggest that future research should evaluate how coverage of dams evolves over time, especially before and after a dam failure.

This is a key moment to reduce the risk to the population of Michigan from dams that are highly prone to failure. Risk reduction could include raising awareness of the issue in Michigan (and across other states where numerous aging dams also exist). This way they can play their role in civil society to pressure legislatures to invest in either fixing or removing these aging dams—civil society groups have had some success in capping abandoned oil and gas wells, which are somewhat similar to idled dams (Mallinson et al. 2023). Research on other dangerous infrastructure (such as nuclear plants) suggests that people close to these plants tend to downplay the dangers, and that the dangers disproportionately affect the less powerful in society. Alternatives for citizens in areas with dams that have benefitted economically from this infrastructure will need to be developed. This is an opportunity to build new green infrastructure that creates jobs and income in these areas to replace the losses from tourism should dams be removed. Educational efforts can be made to show people with information from areas where dams have been removed that new opportunities arise with a return to free-flowing rivers.

A first step would be to encourage regulators and environmental advocates to more frequently raise the issue of dam safety in encounters with journalists, with the aim of helping to set a media agenda that would raise the issue’s salience among news audiences. These communicators should encourage media frames that emphasize the citizens’ agency to intervene with officials to initiate policy changes and appropriate dam removals, with the Lyons Dam as an exemplary case.

Conflict of Interest

The authors declare no conflicts of interest.

Amos, Adell L. 2014. “Dam Removal and Hydropower Production in the United States-Ushering in a New Era.” J. Envtl. L. & Litig. 29: 1.
Anderson, Brooke Auxier and Monica. 2021. “Social Media Use in 2021.” Pew Research Center: Internet, Science & Tech (blog). April 7, 2021. https://www.pewresearch.org/internet/2021/04/07/social-media-use-in-2021/.
Becker, Julia S., Douglas Paton, David M. Johnston, Kevin R. Ronan, and John McClure. 2017. “The Role of Prior Experience in Informing and Motivating Earthquake Preparedness.” International Journal of Disaster Risk Reduction 22: 179–93.
Blair, Benjamin, Tanya Heikkila, and Christopher M. Weible. 2016. “National Media Coverage of Hydraulic Fracturing in the United States: Evaluation Using Human and Automated Coding Techniques.” Risk, Hazards & Crisis in Public Policy 7 (3): 114–28.
Blanchard-Boehm, R. Denise, and M. Jeffrey Cook. 2004. “Risk Communication and Public Education in Edmonton, Alberta, Canada on the 10th Anniversary of the ‘Black Friday’ Tornado.” International Research in Geographical and Environmental Education 13 (1): 38–54. https://doi.org/10.1080/10382040408668791.
Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14 (1): 63–82.
Breen, Richard, Kristian Bernt Karlson, and Anders Holm. 2018. “Interpreting and Understanding Logits, Probits, and Other Nonlinear Probability Models.” Annual Review of Sociology 44 (1): 39–54. https://doi.org/10.1146/annurev-soc-073117-041429.
Bronfman, Nicolás C., Pamela C. Cisternas, Paula B. Repetto, Javiera V. Castañeda, and Eliana Guic. 2020. “Understanding the Relationship Between Direct Experience and Risk Perception of Natural Hazards.” Risk Analysis 40 (10): 2057–70. https://doi.org/10.1111/risa.13526.
Campbell, Nnenia M., Christine A. Bevc, and J. Steven Picou. 2013. “Perceptions of Toxic Exposure: Considering ‘White Male’ and ‘Black Female’ Effects.” Sociological Spectrum 33 (4): 313–28. https://doi.org/10.1080/02732173.2013.732882.
Carmichael, Jason T., and Robert J. Brulle. 2017. “Elite Cues, Media Coverage, and Public Concern: An Integrated Path Analysis of Public Opinion on Climate Change, 2001–2013.” Environmental Politics 26 (2): 232–52.
Castañeda, Javiera V., Nicolás C. Bronfman, Pamela C. Cisternas, and Paula B. Repetto. 2020. “Understanding the Culture of Natural Disaster Preparedness: Exploring the Effect of Experience and Sociodemographic Predictors.” Natural Hazards 103: 1881–1904.
Colaianne, Jospeh W., and David E. Kepler. 2022. “The" Four Lakes Story"-Mid-Michigan’s Dam Failures: What Happened and What We Learned. A Case Study of the Four Lakes Task Force and the Restoration of the Four Lakes System.” Wayne L. Rev. 68: 65.
Cornesse, Carina, and Annelies G. Blom. 2023. “Response Quality in Nonprobability and Probability-Based Online Panels.” Sociological Methods & Research 52 (2): 879–908.
Cornesse, Carina, Annelies G. Blom, David Dutwin, Jon A. Krosnick, Edith D. De Leeuw, Stéphane Legleye, Josh Pasek, Darren Pennay, Benjamin Phillips, and Joseph W. Sakshaug. 2020. “A Review of Conceptual Approaches and Empirical Evidence on Probability and Nonprobability Sample Survey Research.” Journal of Survey Statistics and Methodology 8 (1): 4–36.
Crawford, Jessica, Douglas Bessette, and Sarah B. Mills. 2022. “Rallying the Anti-Crowd: Organized Opposition, Democratic Deficit, and a Potential Social Gap in Large-Scale Solar Energy.” Energy Research & Social Science 90: 102597.
Dake, Karl. 1992. “Myths of Nature: Culture and the Social Construction of Risk.” Journal of Social Issues 48 (4): 21–37.
Eiser, J. Richard, Ann Bostrom, Ian Burton, David M. Johnston, John McClure, Douglas Paton, Joop Van Der Pligt, and Mathew P. White. 2012. “Risk Interpretation and Action: A Conceptual Framework for Responses to Natural Hazards.” International Journal of Disaster Risk Reduction 1: 5–16.
Entman, Robert M. 1993. “Framing: Toward Clarification of a Fractured Paradigm.” Journal of Communication 43 (4): 51–58.
Feygina, Irina, John T. Jost, and Rachel E. Goldsmith. 2010. “System Justification, the Denial of Global Warming, and the Possibility of ‘System-Sanctioned Change.’” Personality and Social Psychology Bulletin 36 (3): 326–38. https://doi.org/10.1177/0146167209351435.
Fisk, Jonathan M., Martin K. Mayer, and John C. Morris. 2023. “Introduction to the Symposium: Redundant, Abandoned, Idled, and Neglected Infrastructure: Problems and Solutions for the 21 ^st Century.” Public Works Management & Policy 28 (1): 4–10. https://doi.org/10.1177/1087724X221128816.
Flynn, James, Paul Slovic, and C. K. Mertz. 1994. “Gender, Race, and Perception of Environmental Health Risks.” Risk Analysis 14 (6): 1101–8. https://doi.org/10.1111/j.1539-6924.1994.tb00082.x.
Frank, Kenneth A., Spiro J. Maroulis, Minh Q. Duong, and Benjamin M. Kelcey. 2013. “What Would It Take to Change an Inference? Using Rubin’s Causal Model to Interpret the Robustness of Causal Inferences.” Educational Evaluation and Policy Analysis 35 (4): 437–60.
Freudenburg, William R., and Debra J. Davidson. 2007. “Nuclear Families and Nuclear Risks: The Effects of Gender, Geography, and Progeny on Attitudes toward a Nuclear Waste Facility.” Rural Sociology 72 (2): 215–43.
Frewer, Lynn J., Susan Miles, and Roy Marsh. 2002. “The Media and Genetically Modified Foods: Evidence in Support of Social Amplification of Risk.” Risk Analysis 22 (4): 701–11. https://doi.org/10.1111/0272-4332.00062.
Goldsmith, Rachel E., Irina Feygina, and John T. Jost. 2013. “The Gender Gap in Environmental Attitudes: A System Justification Perspective.” In Research, Action and Policy: Addressing the Gendered Impacts of Climate Change, edited by Margaret Alston and Kerri Whittenbury, 159–71. Dordrecht: Springer Netherlands. https://doi.org/10.1007/978-94-007-5518-5_12.
Hanmer, Michael J., and Kerem Ozan Kalkan. 2013. “Behind the Curve: Clarifying the Best Approach to Calculating Predicted Probabilities and Marginal Effects from Limited Dependent Variable Models.” American Journal of Political Science 57 (1): 263–77.
Howell, Emily L., Christopher D. Wirz, Dominique Brossard, Dietram A. Scheufele, and Michael A. Xenos. 2019. “Seeing through Risk-Colored Glasses: Risk and Benefit Perceptions, Knowledge, and the Politics of Fracking in the United States.” Energy Research & Social Science 55: 168–78.
Kahan, Dan M., Donald Braman, John Gastil, Paul Slovic, and C. K. Mertz. 2007. “Culture and Identity-Protective Cognition: Explaining the White-Male Effect in Risk Perception.” Journal of Empirical Legal Studies 4 (3): 465–505.
Kasperson, Roger E., Ortwin Renn, Paul Slovic, Halina S. Brown, Jacque Emel, Robert Goble, Jeanne X. Kasperson, and Samuel Ratick. 1988. “The Social Amplification of Risk: A Conceptual Framework.” Risk Analysis 8 (2): 177–87.
Kingdon, John W., and Eric Stano. 1984. Agendas, Alternatives, and Public Policies. Vol. 45. Little, Brown Boston.
Kristiansen, Silje. 2017. “Characteristics of the Mass Media’s Coverage of Nuclear Energy and Its Risk: A Literature Review.” Sociology Compass 11 (7): e12490.
Lee, Jiyoung, Jihyang Choi, and Rebecca K. Britt. 2023. “Social Media as Risk-Attenuation and Misinformation-Amplification Station: How Social Media Interaction Affects Misperceptions about COVID-19.” Health Communication 38 (6): 1232–42. https://doi.org/10.1080/10410236.2021.1996920.
Mallinson, Daniel J., Aqsa Ali, Jingyu Guo, and Pedro Robles. 2023. “The Scourge of Orphaned and Abandoned Wells: Leveraging Public-Private-Citizen Collaboration to Solve a Big Problem.” Public Works Management & Policy 28 (1): 33–52. https://doi.org/10.1177/1087724X221112958.
Marshall, Brent K., J. Steven Picou, Cecilia Formichella, and Keith Nicholls. 2006. “Environmental Risk Perceptions and the White Male Effect: Pollution Concerns among Deep-South Coastal Residents.” Journal of Applied Sociology os-23 (2): 31–49. https://doi.org/10.1177/19367244062300203.
Matheny, Keith. n.d. “Michigan Puts $175M Price Tag on Flooding Damage in Midland County.” Detroit Free Press. Accessed February 19, 2024. https://www.freep.com/story/news/local/michigan/2020/06/08/midland-flood-damage-major-disaster-whitmer/5321673002/.
Mayer, Adam. 2016. “Risk and Benefits in a Fracking Boom: Evidence from Colorado.” The Extractive Industries and Society 3 (3): 744–53.
———. 2018. “A Just Transition for Coal Miners? Accountability Frames, Community Economic Identity, and Just Transition Policy Support among Local Policy Actors.” Environmental Innovation and Societal Transitions.
Mayer, Adam, Laura Castro-Diaz, Maria Claudia Lopez, Guillaume Leturcq, and Emilio F. Moran. 2021. “Is Hydropower Worth It? Exploring Amazonian Resettlement, Human Development and Environmental Costs with the Belo Monte Project in Brazil.” Energy Research & Social Science 78: 102129.
Mayer, Adam, and Perry Parks. 2024. “Media and Partisanship in Energy Transition: Towards a New Synthesis.” Energy Research & Social Science 108: 103368.
McAdam, Doug, and Hilary Boudet. 2012. Putting Social Movements in Their Place: Explaining Opposition to Energy Projects in the United States, 2000–2005. Cambridge University Press. https://books.google.com/books?hl=en&lr=&id=CzwgAwAAQBAJ&oi=fnd&pg=PR9&dq=mCadam+and+boudet&ots=vWh-oebSOX&sig=iq8gXs-sGidZms24zRy8e0-wo00.
McCright, Aaron M., and Riley E. Dunlap. 2011. “Cool Dudes: The Denial of Climate Change among Conservative White Males in the United States.” Global Environmental Change 21 (4): 1163–72.
Mood, Carina. 2010. “Logistic Regression: Why We Cannot Do What We Think We Can Do, and What We Can Do about It.” European Sociological Review 26 (1): 67–82.
Moran, Emilio F., Maria Claudia Lopez, Nathan Moore, Norbert Müller, and David W. Hyndman. 2018. “Sustainable Hydropower in the 21st Century.” Proceedings of the National Academy of Sciences 115 (47): 11891–98.
Palmer, Christina. 2003. “Risk Perception: Another Look at the’white Male’effect.” Health, Risk & Society 5 (1): 71–83.
Parks, Perry. 2019. “Naturalizing Negativity: How Journalism Textbooks Justify Crime, Conflict, and ‘Bad’ News.” Critical Studies in Media Communication 36 (1): 75–91. https://doi.org/10.1080/15295036.2018.1533990.
Peer, Eyal, David Rothschild, Andrew Gordon, Zak Evernden, and Ekaterina Damer. 2022. “Data Quality of Platforms and Panels for Online Behavioral Research.” Behavior Research Methods 54 (4): 1643–62.
Porter, Christopher OLH, Ryan Outlaw, Jake P. Gale, and Thomas S. Cho. 2019. “The Use of Online Panel Data in Management Research: A Review and Recommendations.” Journal of Management 45 (1): 319–44.
[email protected], Riley Murdock |. 2020. “2,300 Homes Damaged and 9 More Reasons Michigan Wants Flooding Declared a Federal Disaster.” Mlive. June 16, 2020. https://www.mlive.com/news/saginaw-bay-city/2020/06/2300-homes-damaged-and-9-more-reasons-michigan-wants-flooding-declared-a-federal-disaster.html.
Roulin, Nicolas. 2015. “Don’t Throw the Baby out with the Bathwater: Comparing Data Quality of Crowdsourcing, Online Panels, and Student Samples.” Industrial and Organizational Psychology 8 (2): 190–96.
Schafer, Ashley. 2020. “Troubled Waters: Edenville Dam Just One Part in a System.” Midland Daily News. May 30, 2020. https://www.ourmidland.com/news/article/Troubled-Waters-nbsp-Edenville-Dam-just-one-part-15304173.php.
Siciliano, Giuseppina, Daniela Del Bene, Arnim Scheidel, Juan Liu, and Frauke Urban. 2019. “Environmental Justice and Chinese Dam-Building in the Global South.” Current Opinion in Environmental Sustainability 37: 20–27.
Uría-Martínez, Rocío, Megan M. Johnson, and Rui Shan. 2020. “Challenges in the Post-Licensing Stage of US Hydropower Development: Current Status and Some Potential Solutions.” The Electricity Journal 33 (4): 106729.
Weaver, David. 1991. “Issue Salience and Public Opinion: Are There Consequences of Agenda-Setting?” International Journal of Public Opinion Research 3 (1): 53–68.
Xu, Ran, Kenneth A. Frank, Spiro J. Maroulis, and Joshua M. Rosenberg. 2019. “Konfound: Command to Quantify Robustness of Causal Inferences.” The Stata Journal 19 (3): 523–50.
Zhang, Xiaochen Angela, and Raluca Cozma. 2022. “Risk Sharing on Twitter: Social Amplification and Attenuation of Risk in the Early Stages of the COVID-19 Pandemic.” Computers in Human Behavior 126: 106983.

We tested all of our models for the parallel lines assumption, also known as the proportional odds assumption, using a brant test. Briefly, the parallel lines assumption refers to a constraint placed on predictor variables in ordinal logistic regression models wherein their effect is fixed across categories of the outcome variable. Alternative modelling strategies can relax this assumption, but it produces a far less parsimonious model. We found that in some model specifications the assumption was violated for some categories of income and education, but not for the explanatory variables of primary interest such as political ideology, political party, media consumption, white race and male sex. Accordingly, we opt to retain the parsimonious ordinal logistic regression approach.

Appendices.docx

Download PDF

Reviewers agreed at journal
07 Mar, 2024
Reviewers invited by journal
07 Mar, 2024
Editor assigned by journal
06 Mar, 2024
First submitted to journal
06 Mar, 2024

You are reading this latest preprint version

Dams are old and abandoned, are people concerned? Risk perceptions from Michigan, USA

Status:

Version 1

Abstract

Figures

Introduction

The Context of Dams in Michigan

Risk Perceptions

Data Collection

Outcome Variable: Concern for Dams

Predictor Variables

Modelling Strategy

Modelling Results

Robustness Checks

Discussion

Declarations

References

Footnotes

Supplementary Files

Status:

Version 1