3.1. Overall Willingness to Engage in Each of the Renter or Buyer Behaviors
For all statistical analyses, we used the software package SPSS Statistics 25.0 (IBM, Armonk, NY, USA). Datasets and syntaxes are available as supplements (see
Supplementary Files S4–S9). Before reporting on the differences by randomized conditions within each scenario, we compared the respondents’ relative willingness to adopt the different risk mitigation behaviors described within each study. To do so, we conducted a one-way repeated measures analysis of variance (ANOVA), controlling for experimental conditions, to compare the ratings of willingness to adopt different responses to the presence of LSLs. We used Greenhouse–Geisser corrections when the assumption of sphericity was violated, and Bonferroni corrections for post hoc comparisons [
26]. We present these within-study comparisons of strategies in
Table 2.
In Study 1, featuring landlord disclosures to home renters, participants showed the greatest intentions to adopt RB1 (
M = 4.46,
SD = 1.52) and RB5 (
M = 4.27,
SD = 1.49), and the least intentions to adopt RB2 (
M = 2.25,
SD = 1.50) and RB4 (
M = 2.67,
SD = 1.64), with intentions to adopt RB3 (
M = 3.89,
SD = 1.69) in the middle. Controlling for the experimental conditions, the results of a one-way repeated measures ANOVA with Greenhouse–Geisser correction showed that the five renter behaviors (RBs) differed in the respondents’ intention levels to adopt each RB,
F(2.93, 1944.71) = 51.46,
p < 0.001,
= 0.058. Post hoc tests with Bonferroni corrections showed that RB1 was significantly higher than RB2, RB3, RB4, and RB5,
ps < 0.05. In addition, RB3 was significantly higher than RB2, and RB4,
ps < 0.001. Also, RB4 was significantly higher than RB2,
p < 0.001. Finally, RB5 was significantly higher than RB2, RB3, and RB4,
ps < 0.01. Hence, asking the landlord to replace the LSL was the highest rated behavior, followed by looking for another home. Paying for replacement oneself was the lowest rated option. The results of the one-way repeated measures ANOVAs for all three studies are shown in
Table 2.
In Study 2, featuring seller disclosures to home buyers, participants showed the most intention to adopt BB2 (M = 4.89, SD = 1.27) and BB3 (M = 4.88, SD = 1.30), and the least intention to adopt BB6 (M = 1.84, SD = 1.32), with intentions to adopt BB1 (M = 3.70, SD = 1.69), BB4 (M = 3.33, SD = 1.77), BB5 (M = 3.09, SD = 1.74), and BB7 (M = 3.87, SD = 1.55) in the middle. Controlling for the experimental conditions, the results of a one-way repeated measure ANOVA with Greenhouse–Geisser correction showed that the seven BBs differed in the respondents’ intention levels to adopt each BB, F(4.46, 2959.69) = 103.75, p < 0.001, = 0.110. Posthoc tests with Bonferroni corrections showed that BB1 was significantly higher than BB4, BB5, and BB6, ps < 0.001. BB2 was significantly higher than BB1, BB4, BB5, BB6, and BB7, ps < 0.001. BB3 was significantly higher than BB1, BB4, BB5, BB6, and BB7, ps < 0.001. BB4 was significantly higher than BB6, p < 0.001. BB5 was significantly higher than BB6, p < 0.001. BB7 was significantly higher than BB4, BB5, and BB6, ps < 0.001. Having the seller replace the LSL, or self-replacing it after deducting the cost from the sale price were the highest rated options. Doing nothing was the lowest rated option.
In Study 3, featuring home inspector disclosures to home buyers, participants showed the greatest intentions to adopt BB2 (M = 4.74, SD = 1.34), and BB3 (M = 4.79, SD = 1.36), and the least intentions to adopt BB5 (M = 2.79, SD = 1.69), and BB6 (M = 1.61, SD = 1.19), with intentions to adopt BB1 (M = 3.54, SD = 1.70), BB4 (M = 3.53, SD = 1.69), and BB7 (M = 3.65, SD = 1.64) in the middle. Controlling for experimental conditions, the results of a one-way repeated measures ANOVA with Greenhouse–Geisser correction showed that the seven buyer behaviors (BBs) differed in their intention levels to adopt each BB, F(4.78, 4140.64) = 142.51, p < 0.001, = 0.114. Posthoc tests with Bonferroni corrections showed that BB1 was significantly higher than BB5 and BB6, ps < 0.001. BB2 was significantly higher than BB1, BB4, BB5, BB6, and BB7, p < 0.001. BB3 was significantly higher than BB1, BB4, BB5, BB6, and BB7, ps < 0.001. BB4 was significantly higher than BB5 and BB6, ps < 0.001. BB5 was significantly higher than BB6, p < 0.001. BB7 was significantly higher than BB5 and BB6, ps < 0.001. As in Study 2, having the seller replace the LSL, or self-replacing it after deducting the cost from the sale price were the highest rated options. Doing nothing was the lowest rated option.
We organized the presentation of the results for the randomized disclosure conditions within these three studies in the sections that follow, this time reporting on all of the outcomes for the different conditions as a set. We used one-way ANOVAs featuring the three (in Studies 1 and 2) or four (in Study 3) randomized disclosure conditions as the independent variables, and the willingness to adopt risk mitigation behaviors, risk perceptions, response efficacy, self-efficacy, and affordability as the dependent variables, respectively. We again used Bonferroni corrections for post hoc comparisons. In addition, since “click on the link” was a binary variable, we used logistic regression featuring the three or four randomized conditions as the independent variables (one condition dummy-coded and omitted as the referent group), and “click on the link” as the dependent variable.
3.2. Study 1: Landlord Disclosures to Home Renters
We continued by comparing all study outcomes by randomized disclosure conditions (“No Water Test Results”, “Above EPA Level”, and “Below EPA Level”) in the landlord/renter study. ANOVAs and subsequent means comparisons (with Bonferroni corrections) revealed a number of statistically significant differences in willingness to take action to mitigate the risks associated with the presence of LSLs (see
Figure 2 for visual display of these findings). First, there were significant differences for the willingness to insist on landlord replacement,
F(2, 664) = 18.47,
p < 0.001,
= 0.053, and look for another home,
F(2, 664) = 14.55,
p < 0.001,
= 0.042. Specifically, exposure to the “No Water Test Results” and the “Above EPA Level” conditions led to a greater willingness to insist on landlord replacement, and to look for another home, than did the “Below EPA Level” condition,
ps < 0.001. These results are consistent with our hypothesis that the “Below EPA level” condition would produce lower levels of willingness to replace the LSL than the other two conditions. In addition, we observed the reverse pattern of results for the willingness to leave LSLs alone,
F(2, 664) = 23.05,
p < 0.001,
= 0.065. In other words, the “No Water Test Results” and the “Above EPA Level” conditions did not differ from one another, and led to a lower level of willingness to leave the lead pipes alone, than did the “Below EPA Level” condition,
ps < 0.001. We also found a significant difference for willingness to install and maintain a filter,
F(2, 664) = 3.07,
p = 0.047,
= 0.009. In particular, the “Below EPA Level” condition produced a stronger willingness than the “Above EPA Level” condition,
p = 0.043.
We observed a similar pattern of results for risk perceptions,
F(2, 664) = 4.72,
p = 0.009,
= 0.014. That is, the “No Water Test Results” and the “Above EPA Level” conditions led to higher risk perceptions than did the “Below EPA Level” condition,
ps < 0.05. In an unexpected finding, the “Below EPA Level” condition led to more clicks on the lead pamphlet than the “Above EPA Level” condition, B = −0.62, SE = 0.27,
p = 0.022. There were no other statistically significant differences by randomized experimental group for the other outcomes, as shown in
Table 3.
Overall, we note from
Table 3 that Study 1 respondents across all conditions reported a high level of willingness to either (a) insist on landlord replacement of the LSL as a condition of renting, or to (b) look for another home to rent (with means for both items above the midpoint of 3.5 across all conditions). Risk perceptions and response efficacy were also very high across all three conditions (with means near or above 5 on a 6-point scale anchored at 5 with “agree”, and at 6 with “strongly agree”). The perceived self-efficacy and affordability were much lower, and less than one in four respondents clicked on the link for additional information about lead in any condition.
3.4. Study 3: Inspector Disclosures to Home Buyers
We continued by comparing all of the study outcomes by randomized disclosure conditions (Recommendations Present/Absent and Why Information Present/Absent) in the inspector/buyer study. In light of the fact that, as noted above, we did not see any significant differences between the three experimental conditions in the home seller/buyer study (Study 2) on any outcome, we chose to combine the three conditions, and to include them, as a whole, as a “comparison group” in the analyses for the home inspector/buyer study (Study 3). While it is true that the stimuli used for the two home buyer experiments were not the same, and they featured different sources of disclosure, we reason that the minimum level of disclosure information that was provided in the home seller/buyer experiment makes it a good standard of comparison to serve as a control group in which the participants knew about the presence of LSLs but nothing more.
Using one-way ANOVAs with Bonferroni corrections and logistic regression for the dichotomous outcome (clicking on the link), we examined whether exposure to any of the four inspector/buyer conditions produced any differences in any of the dependent variables, relative to (a) one another, or (b) the aggregated seller/buyer conditions serving as the control group.
There were a number of statistically significant differences in the willingness to adopt specific risk mitigation strategies (
Figure 4). First, we found a significant difference for the willingness to deduct the cost of LSL replacement from the sale price between the four conditions and the control,
F(4, 1533) = 3.05,
p = 0.016,
= 0.008. Specifically, the willingness was lower when the inspector report contained an explicit recommendation and “why” information, relative to exposure to the seller disclosure control,
p = 0.023. These results run contrary to
a priori hypotheses, in which we expected the explicit recommendations and the “why” information to increase the willingness to have the LSL removed in the home-buying process. On the flipside, however, we observed the same between-conditions pattern for willingness to leave the LSLs alone,
F(4, 1533) = 5.27,
p < 0.001,
= 0.014. That is, the willingness was lower when the inspector report contained an explicit recommendation and “why” information, relative to the exposure to the seller disclosure control group,
p < 0.001.
Second, we found a significant difference for willingness to pay to replace the LSL by participants themselves, F(4, 1533) = 9.08, p < 0.001, = 0.023. In particular, participants were more willing to replace the LSL themselves when there were no explicit recommendations, regardless of the presence or absence of the “why” information, compared to the two conditions where the recommendations were present, or the control condition, ps < 0.01.
Third, there was also a significant difference in willingness to install and maintain a filter, F(4, 1533) = 6.91, p < 0.001, = 0.018. Specifically, compared to when both a recommendation and “why” information was present, the two conditions without the recommendations and the control conditions produced more willingness among the participants, ps < 0.05. In addition, willingness was higher for the control condition than for the condition where recommendations were present, but the “why” information was absent, p = 0.005. Finally, we found a significant difference in willingness to look for another home, F(4, 1533) = 3.20, p = 0.013, = 0.008. Participants in the control condition were more willing to look for another home than those in the condition where recommendations were present, but the “why” information was absent, p = 0.019.
Table 5 shows the mean levels of responses to each dependent variable, by condition in the inspector/buyer study, as well as by the aggregate results from the seller/buyer study. Similar to what we found in the home seller/buyer study, regardless of the randomized experimental condition, participants showed a relatively high level of willingness to deduct the cost of the LSL replacement from the mortgage, or to insist that the seller replace the LSL as a condition of the sale. As for other measured variables, participants rated high on risk perceptions and response efficacy, though these variables did not differ by randomized condition, or between the inspector/buyer study conditions and the aggregate seller/buyer study conditions. Similarly, participants rated well above the scale midpoint on self-efficacy and affordability, but these values also did not differ by condition. Less than one in five respondents clicked on the pdf to learn about lead risks in any of the randomized conditions in the inspector/buyer study, or the aggregate seller/buyer study group.