Abstract
A growing body of empirical research seeks to quantify the causal effects of social policies on health by exploiting variation in the timing of policy changes across places. However, multiple social policies are often adopted simultaneously or in close succession in the same locations, creating clustering which must be handled analytically for valid inferences. Although this is a substantial methodological challenge for studies aiming to isolate social policy effects, yet systematic assessments of available analytic solutions and tradeoffs among approaches are lacking. We designated eight analytic solutions prior researchers have adopted, including efforts to disentangle individual policy effects and efforts to estimate the joint effects of clustered policies. We leveraged an existing systematic review of social policies and health to evaluate how often policy clustering is identified as a threat to validity and how often each analytic solution is applied in practice. Of the 55 studies, only 17 (31%) reported checking for any clustered policies, and 36 (67%) used at least one approach that helps address clustered policies. The most common approaches were adjusting for clustered policies, defining the outcome on subpopulations likely to be affected by the policy of interest but not other clustered policies, and selecting a less-correlated measure of policy exposure were the most common approaches. Systematically assessing policy clustering and applying analytic solutions when necessary would strengthen future studies on the health effects of social policies. Adequate reporting on these analytic decisions would facilitate evaluating validity and interpreting study findings.
INTRODUCTION
Social policies are promising mechanisms to improve population health and reduce health disparities. Analyses of the health effects of social policies routinely leverage policy changes occurring in different places at different times, with differences-in-differences or similar study designs (1). In epidemiology and related fields, empirical health research using these methods has proliferated rapidly and yielded important findings (2–4). However, this approach is challenged when multiple related policies are adopted simultaneously or close succession in the same jurisdiction. Bundles of related policies with similar potential health effects are often changed together, creating effective “clustering” that must be addressed analytically for valid inference. Specifically, analyses that do not account for clustered policies are likely to be confounded, while analyses that incorporate measures of clustered policies can encounter imprecise or unstable estimates and bias resulting from data sparsity (5,6).
For many policy domains, adopting groups of policies as a set is common (7). In these cases, researchers can implement a variety of study designs or statistical strategies to address potential bias or imprecision resulting from policy clustering. Among these approaches, an overarching distinction is whether the approach aims to disentangle the effects of individual policies, or conceptualizes the clustered policies as a group and evaluates their joint effects. For a given study, either approach may be policy-relevant, depending on whether the goal is to deliver actionable evidence on the effects of a single policy or on the effects of a set of policies that would likely be adopted together. To date, most approaches to handling policy clustering have been ad hoc. The existing literature lacks systematic assessments of the analytic solutions that are available, how often these solutions are used, and the tradeoffs to consider in selecting an approach.
This is the second paper in a series on the policy clustering problem. The first paper demonstrated that social policy clustering is pervasiveness and that adequate adjustment for clustered policies is likely to substantially reduce the precision of estimated effects (7). Given this, delineating methods appropriate for this context is a high priority for the next generation of research on the health effects of social policies. Here we review analytic strategies prior researchers have adopted address policy clustering with the goal of attaining valid inferences. We categorized these approaches based on the type of causal question they answer (e.g. the effect of an individual policy on a population subgroup versus the effect of a bundle of policies on the overall population). Using the systematic review of social policy evaluations developed in Part 1 of this series (7), we evaluated the proportion of studies in which authors reported assessed policy clustering and the proportion applying each analytic solution. We discuss the advantages and disadvantages of each approach and provide guidance on selecting among them.
METHODS
Identification of social policy studies
We leveraged a systematic sample of studies on the health effects of social policies to review common strategies for addressing policy clustering. The details of this review are described in Part 1 of this paper series (7). Briefly, we selected a multidisciplinary set of journals that publish health-related social policy research and are leading journals in their respective fields (American Journal of Public Health, American Journal of Epidemiology, Journal of the American Medical Association, New England Journal of Medicine, The Lancet, American Journal of Preventive Medicine, Social Science and Medicine, Health Affairs, Demography, and American Economic Review). Then, we screened all 6,794 articles published in 2019 in these journals and included all empirical studies estimating the causal effects of one or more social policies on health-related outcomes (N=55). “Social policies” were defined as non-medical, population-based or targeted policies adopted at a community or higher level, and hypothesized to affect social determinants of health or health inequalities across social dimensions.
Categorization of analytic approaches
A priori, we designated eight categories of analytic approaches that researchers have adopted in the face of policy clustering, based on the causal question they answer (Table 1). We identified these approaches by reviewing multidisciplinary scientific literature on the study of correlated exposures, consulting with experts, and drawing on methods used in our own fields of research. We focused on methods that apply to study designs leveraging policy changes occurring in different places and different times, including aggregate or multi-level differences-in-differences, panel fixed effects, and interrupted time series. We defined “clustered” policies as policies whose adoption or implementation was correlated in space and time with index policy and that likely affected the health outcome under study. This clustering could be at multiple jurisdictional levels (e.g. cities within states), although single-jurisdictional-level studies were the norm.
Our ultimate framework (Table 1) applies to research questions about the health effects of one particular policy (the “index” policy) or the combined effects of multiple policies, in a defined target population, and to investigations in which any relevant clustered policies have already been identified, and policies that do not meet criteria for “clustered” have already been ruled out (e.g. based on theory, expert opinion, existing evidence, preliminary analysis, or variable selection algorithms). Each approach is described briefly below.
Approach 1: Adjust for clustered policies
If clustering of related policies with the index policy is not severe (see below), the researcher can adjust for measures of the other policies—for example, by controlling for correlated policy measures in a regression. The resulting research question corresponds to the effect of the index policy on the health outcome. This approach will often rely on some degree of model-based extrapolation, because not all possible combinations of policies actually occur. It is incumbent on the investigator to confirm that any extrapolation is either well-founded in theory or evidence, or otherwise avoided.
In the case of severe policy clustering—i.e., if a clustered policy aligns perfectly in place and time with the policy of interest; or if after adjusting for clustered policies, there is insufficient independent variation in the index policy left to study; or if after adjustment, estimates are extremely imprecise—the only analytic solution is to modify the research question. Approaches 2-8 involve such alternative research questions and corresponding analytic approaches.
Approach 2: Restrict the study sample to the region of common support
The issue of policy clustering can be conceptualized as a form of strong confounding of the index policy by the other clustered policies. This confounding and resulting data sparsity result in a lack of common support in the data, also known as a violation of the positivity assumption (8). Positivity violations occur when some confounder strata do not have variation in the exposure—for example, because the confounding policy and index policy are always adopted as a set. This situation can be resolved by restricting the analysis to the confounder strata for which there is variation in the index policy, i.e., the data region of “common support”. This approach changes the target population, so the corresponding causal question refers to the effect of the index policy on the health outcome in the restricted sample, and results are only generalizable to the population represented by the restricted sample.
When there are many confounder strata, one of the most accessible ways to assess positivity and identify the region of common support is using propensity scores. In the context of assessing policy impacts, the propensity score (9) is the probability of adopting the index policy, given the confounding policies. Units that are “on-support” are those with propensity scores within the range of observed propensity scores both for units adopting the index policy and for units not adopting the index policy. A wide variety of matching and weighting methods involve using propensity scores to identify and restrict to the region of common support (10,11). Numerous variations on this restriction have also been proposed, including restricting to units with propensity scores within a prespecified range (e.g. 0.1 to 0.9) or dynamic optimization procedures for selecting propensity score cutoffs (8,12,13).
An alternative approach to using propensity scores is to directly restrict the sample based on the distribution of the clustered policies themselves. Several approaches have been proposed, including restricting to units inside the convex hull of the covariate spaced defined by the secondary policies (14,15), restricting to a sufficiently data-dense, rectangular region of the covariate space defined by the clustered policies (16–18), or tree-based methods (see for example (19)). These approaches are less common, but most can be readily implemented using existing software.
In all cases, assessing the region of common support helps ensure that estimates are not relying on extrapolation to policy combinations which are never observed. The restricted study population should be well-defined, so that the investigator can transparently describe the places and times to which the results apply (16). In some cases, the area of common support may be small or non-existent, in which case alternative approaches such as must be considered (e.g. evaluating the joint impacts of a bundle of policies).
Chang and colleagues applied this approach to study the impacts of prescription drug monitoring programs (PMDPs) and pill mill laws using a comparative interrupted time series analysis (20). Rather than using all states in the analysis, some of which implemented other opioid policy changes in concert with PMDPs or pill mill laws, the authors restricted their analysis to Florida, which adopted PDMP and pill mill laws, and Georgia, which had a similar policy profile to Florida but did not implement the index policies during the study period. The authors determined that the combination of PDMPs and pill mill laws were associated with reductions in high-risk opioid prescribing for the Florida population.
Approach 3: Restrict the study sample so clustered policies do not vary
If policies that might confound the association of interest do not vary within a particular study population, then they cannot cause confounding. Thus, one strategy to address policy clustering is to restrict the study sample to a subpopulation for whom exposure to the non-index clustered policies does not vary. For example, many households are eligible for multiple social welfare programs including the Supplemental Nutrition Assistance Program (SNAP) and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). Liu and colleagues addressed this correlation by studying outcomes for dual beneficiaries of SNAP and WIC versus WIC alone; WIC receipt cannot confound because all study participants are WIC beneficiaries (21). Like Approach 2, this approach changes the target population, so the corresponding causal question refers to the effect of the index policy (e.g. SNAP) on the health outcome in the restricted sample (e.g. WIC beneficiaries) and results generalize to the population represented by the restricted sample.
Approach 4: Define the outcome on subpopulations likely to be affected by the index policy but not other clustered policies
Identifying health effects that are specific to the index policy can be achieved by changing the outcome measure to one that is closely aligned with the index policy but unlikely to be affected by other clustered policies. In particular, if the outcome is focused on a particular population subgroup (e.g. defined by age, gender, place, or time) that is likely to be most- or least-affected by the index policy, study results can provide pointers to the impacts of the individual policy.
For example, changes in state Earned Income Tax Credit (EITC) policies have often co-occurred with other changes in other social welfare policies such as SNAP(22). Rehkopf and colleagues took advantage of the fact that EITC cash benefits are typically delivered in February, March, and April, while other benefits do not have the same seasonal dispersal pattern, to examine the association of EITC policies with health(23). They used a differences-in-differences approach to compared health outcomes that can change on a monthly basis (e.g. health behaviors, cardiovascular and metabolic biomarkers) for EITC-eligible versus non-eligible individuals in months of income supplementation versus non-supplementation. The authors were thus able to measure some potential short-term health impacts of EITC independent of other social welfare policies that do not have this seasonality.
This approach can be further strengthened by incorporating falsification tests or negative control analyses. Rehkopf and colleagues strengthened their findings by confirming that treating non-EITC transfer months as “treated” months produced null findings, and by confirming null associations for outcomes that do not change on a monthly basis.
Approach 5: Select a less-correlated measure of policy exposure
Studies that use binary (0 or 1) characterizations of policy adoption are widespread in studies of the health impacts of social policies. However, more detailed characterizations of individual policies—for example, the amount of funding allocated, benefit generosity, participation rate, or population reach of a program; the size of a tax; or the number of years a policy has been in place—can deliver policy measures that are less correlated with other related policies, or opportunities to examine dose-response effects among jurisdictions adopting a policy. For example, adoption of various unemployment benefit policies is highly correlated across state-years, but researchers have successfully assessed their health impacts by studying varying levels of unemployment benefit generosity (24,25). Similar approaches have been taken to studying the effects of alcohol taxes (26,27), tobacco taxes (28), EITC benefit generosity (29,30), and many others.
One useful modification to this approach is to study factors that may specifically mediate the relationship between the index policy and the health outcomes. For example, Matthay and colleagues generated evidence on the impacts of policies regulating gun shows by examining the impacts of gun show events themselves on firearm-related injuries in differing policy environments (31). Similarly, the impacts of medical cannabis provisions allowing supply through dispensaries (as opposed to home-cultivation) can be quantified by studying the direct effects of dispensaries on health (32,33).
Studying mediators may also offer the opportunity to identify policy effects via the Front Door Criterion (34), a rarely-used alternative to confounder-control or instrument-based methods (4). If all the pathways by which the index policy affects the outcome can be measured, and there are no unmeasured confounders of the index policy-mediator relationship, then the effect of the index policy on the outcome can be identified even without measuring the clustered policies.
Approach 6: Use Bayesian methods
Several approaches to addressing policy clustering involve incorporating prior knowledge about the policies, determinants of the outcome, or hypothesized mechanisms of effect—for example, by making judgements about which policies are likely to affect the outcome of interest or to modify one another, and thus deciding which policies need to be controlled and how. One formal, statistical way to incorporate prior knowledge is by using Bayesian methods to specify prior distributions on the effects of the index and clustered policies and to combine these with empirical information in the data. Bayesian methods can help address estimation issues and recover precision when highly correlated policies lead to convergence problems or imprecision. In particular, Bayesian approaches can stabilize estimates by constraining the effect sizes or interaction effects among policies and “shrinking” coefficients towards the specified prior distributions (35). This approach is common in the environmental epidemiology literature as way to study multiple correlated exposures such as air pollutants (6,36). Harper used a Bayesian approach to estimate the effects of state seat belt laws on motor vehicle crash deaths; this method enhanced the stability and precision of the effect estimates while incorporating existing evidence on these policies’ effects (37).
Approach 7: Identify and evaluate the impacts of policy clusters
If a set of policies are typically adopted as a group, the effect of the combined set of policies may be the most pertinent parameter to estimate. By conceptualizing policy clusters as the exposure of interest, the investigator can preserve the original target population, outcome measure, and exposure measure,. The index policy can be bundled with related clustered policies to study their combined effects. This approach can be taken in ad hoc fashion, by observing that two or more policies are highly correlated and deciding to estimate their combined impact (i.e., comparing health outcomes if all both policies are adopted versus if neither policy is adopted) (38). Policy clusters can also be defined based on substantive or policymaking considerations—for example, if decisionmakers are specifically interested in understanding the combined effects of a set of policies they are considering for adoption. Furthermore, numerous data-driven algorithms exist for defining clusters. Clusters can be defined based on how frequently policies occur together, using method such as hierarchical cluster analysis, latent class analysis (LCA), and principal components analysis (PCA) (6,39,40). Clusters can also be defined based on the strength of the relationship with the outcome—for example using supervised PCA (6,36). No one algorithm is considered optimal for all settings (6,36).
Among data-driven algorithms, it is common to distinguish between “variable-centered” methods that group similar policy variables (e.g. PCA) and “person-centered” methods that group similar observations (e.g. LCA). The underlying mechanics of variable-centered and person-centered approaches are distinct, but both ultimately result in a small set of variables that summarize the policies to which each observation is exposed. This smaller set of variables is then used to assess health impacts. Erickson and colleagues used LCA to classify states based on their position on 18 alcohol control policies (41). They found that each state fell into one of four categories of stringency—weak except serving policies, average, strong for underage use, and strong policies overall—and that policy category was associated with levels of past-month alcohol consumption.
Approach 8: Use an overall policy stringency or generosity score
If the investigator is interested the effects of the overall policy environment on health, a straightforward approach is to use a summary score of the stringency or generosity of a set of policies. The simplest scores sum the number of policies in the set that apply, for each place and time. Policies must be coded in the same direction so that more policies indicates greater restrictiveness, or vice versa. This method is easy to operationalize, but it implies that all policies carry equal weight and are interchangeable in achieving health effects.
A more sophisticated approach is to weight some policies more heavily than others, based the strength of the relationship with the outcome, existing evidence, or expert opinion that they are more important for the health outcomes of interest in terms of efficacy, restrictiveness, implementation, enforcement, enforceability, reach, or other metrics. Investigators have applied this approach in literature on firearm policy (42,43), alcohol policy (44–47), and cannabis policy (48,49). Although there are an infinite number of ways a set of policies can be ranked or weighted, use of systematic methods such as the Delphi technique can enhance rigor and replicability (50,51). Assigned weights are typically outcome-specific—for example, weighting state alcohol policies with different levels effectiveness for binge drinking versus impaired driving and for adults versus youth (44,45). Investigators can also explore different methods of weighting in sensitivity analyses (45).
Data extraction and analysis
For each social policy study, we assessed: (a) the overall analytic approach (e.g. differences-in-differences); (b) whether the authors reported assessing any co-occurring policies related to the health outcome of interest (“clustered policies”) that might pose a threat to validity; (c) whether the authors determined that any clustered policies did, in fact, threaten validity; (d) if the authors did not report assessing potential threats to validity posed by clustered policies, whether there was any other indication that clustered policies exist for the study’s application (e.g. a co-occurring policy mentioned in the limitations); (e) what analytic strategy the authors used to address policy clustering, if it was identified as a threat; and (f) any other aspects of the analytic strategy that may help address clustered policies, whether they were identified as a threat or not. We also documented whether studies utilized any approaches to address policy clustering not identified a priori. We then tabulated these characteristics.
RESULTS
We assessed 55 studies of social policies encompassing diverse topics, countries, and jurisdictional levels (52–107). Studies included, for example, a comparative interrupted time series evaluation of the impacts of lowering the blood alcohol concentration limit for drivers on road traffic accidents in Scotland (55) and a differences-in-differences analysis of the effects of state paid family leave policies on breastfeeding (80). Studies employed a range of study designs, the most common being differences-in-differences, before-after (e.g. t-tests or interrupted time series), and regression (Table 2).
Figure 1 presents a flowchart of the included studies, broken down by whether the authors evaluated policy clustering or used techniques to address policy clustering. Of the 55 studies, 4 involved methods for which assessing policy clustering was not relevant: One involved a national policy with randomized rollout across village clusters, and for three others, the primary research question was about the overall policy environment and authors employed policy stringency scores. Of the remaining 51 studies, authors infrequently reported evaluating policy clustering: only 17 reported checking for at least one clustered policy. Of these, 10 reported identifying at least one clustered policy. For example, in a study of state texting-while-driving bans and traffic injuries, the authors acknowledged that administrative license suspension, speed limits, seatbelt requirements, and graduated driver licensing laws had also changed over the study period and might affect traffic injuries; they therefore controlled for measures of these policies in their differences-in-differences analysis (94). In contrast, a study of the effects of tuition-free primary education on access to family planning and health decision-making evaluated potentially co-occurring paid family leave policies (68); they determined that these policies were not substantially correlated with tuition-free primary education but acknowledged that there may be other correlated policies that were not measured. Of the 34 studies that did not report checking for at least one clustered policy, 5 had some other indication that policy clustering may be a threat.
Overall, 36 of the 55 studies (65%) incorporated at least one approach that addresses policy clustering (Table 2). All of the approaches enumerated a priori to address policy clustering were used in at least one study, except restricting the study sample to the region of common support and using Bayesian methods. Studies using differences-in-differences, panel fixed effects, or regression were more likely to use at least one approach to address policy clustering than studies using before-after designs. Of the 10 studies that explicitly reported identifying one or more clustered policies, 9 used at least one of the eight techniques to address policy clustering. Of the 41 studies that did not check for or identify policy clustering, 23 nonetheless used at least one technique that helps address policy clustering. Several studies used multiple approaches in the same analysis. For example, a study of the effects of losing SNAP benefits controlled for participation in WIC and Temporary Assistance for Needy Families (TANF) program (Approach 1), excluded those with potential concurrent benefit changes in Supplemental Security Income (SSI) (Approach 3), and compared outcomes among SNAP beneficiaries, for those cutoff from benefits versus similar individuals with continuous benefits (Approach 4) (77).
DISCUSSION
Policy clustering is common and can threaten the validity of studies of the health effects of social policies. In this paper, we designated eight analytic approaches that can be used to address policy clustering, either by seeking to disentangle the effects of individual policies or estimating the combined effects of clusters of policies. Using a review of contemporary studies on the health effects of social policies, we found that potential policy clustering is frequently unidentified and unaddressed: only 33% of studies reported checking for policy clustering as a potential threat to validity and only 65% incorporated any approach that helps to address policy clustering.
Among studies that did utilize at least one approach to address policy clustering, the most common approaches were adjusting for clustered policies, defining the outcome on subpopulations likely to be affected by the index policy but not other clustered policies, and selecting a less-correlated measure of policy exposure. No study restricted the study sample to the region of common support or used Bayesian methods; these methods may be under-utilized and researchers may benefit from considering these approaches in future work. Several studies that estimated the individual effect of the index policy used multiple approaches to address clustering in the same analysis; this may further enhance validity, although it is not guaranteed.
In future studies, systematically evaluating and reporting on policy clustering would facilitate the evaluation of validity and interpretation of findings. Many studies (35%) did not incorporate any approach to address policy clustering. In these cases, authors may have failed to consider it as a potential concern. Alternatively, authors may have checked for policy clustering but not reported it, particularly if it was not found to be a concern. However, because failure to address policy clustering (if it exists) poses a major threat to validity, consumers of research need to understand if the authors believe that no policy clustering exists or if they believe it has been addressed, and if so, through what analytic strategies. We assessed whether the studies in our sample checked for any clustered policies, but researchers should ideally evaluate all policies and related social, economic, and political phenomena that may co-occur in space and time with the index policy and affect the health outcome of interest. This is a formidable task, especially when databases measuring relevant policies do not exist, or the factors determining the outcome are not fully understood. However, striving for this goal would enhance the rigor and validity of studies of the health effects of social policies.
Selecting among approaches: Advantages and disadvantages
This review addresses how approaches to address policy clustering have been used practice; a logical next question is which approaches are best-suited for different circumstances. Future simulation-based investigations could help to answer this question. In general, if policy clustering poses a threat, the preferred approach to address it should be driven by the target causal question (108). Investigators should select the approach that best-answers their causal question, while achieving sufficient accuracy and precision (e.g. based on likely sources of bias and evidence on the precision of different estimators). If an approach fits these criteria while simply adjusting for measures of the clustered policies (Approach 1), this option will often be preferable. However, if the target causal question is deemed unanswerable due to severe policy clustering, the choice to shift to a different causal question and corresponding analytic approach should be explicit and intentional (108).
Table 3 summarizes key tradeoffs in selecting among approaches to address policy clustering. Overall, approaches that preserve estimates of the independent effect of the index policy may be particularly useful for decision-makers considering alternative policy options. However, these approaches generally sacrifice some aspect of generalizability by restricting the analysis to certain populations, subgroups, outcomes, or time periods for which policy effects can be estimated. Results may therefore serve as markers of policy impacts rather than measures of overall impact. Approaches that involve estimating the combined effects of clusters of policies sacrifice estimates of the independent effects of the index policy, instead estimating a combined effect of multiple policies, but preserve generalizability to the original target population, outcomes, and time period under study. The preferred approach depends both on what options are viable (i.e. unconfounded, sufficient precision) and which causal question is of greatest interest. For example, if certain combinations of policies are always adopted together, then their independent effects may be neither estimable nor of interest.
Among the individual approaches, key considerations include the circumstances in which the approach is feasible (e.g. controlling for clustered policies is not possible if policy clustering is severe), the availability of evidence to support making analytic decisions (e.g. on how to use propensity scores, select weighting schemes for policy scores, or choose a clustering method), the extent to which the approach provides evidence that is relevant to the original causal question, ease of implementation, available data and measures, and interpretability of the results (see Table 3 for details). All of the approaches discussed here can also be used to evaluate whether policy clustering is a concern: if results of analyses that do not account for policy clustering are substantially different from those that do, policy clustering may be a concern. While none of these approaches will answer the exact same research question, they can serve as robustness checks.
Limitations
The approaches discussed here are not guaranteed to solve the policy clustering problem. If used appropriately, they may help, but the potential for residual confounding, other sources of bias, and issues of precision should be evaluated and addressed as for all epidemiologic studies. Additionally, the approaches presented here are not an exhaustive list and many sub-options exist. However, we did not encounter any other method that addresses policy clustering in our review.
Conclusions
Most research on the health effects of social policies faces the challenge of clustered policies. In combination with Part 1 of this series which illustrated how to assess the pervasiveness and consequences of policy clustering (7), this review serves as a guide for researchers seeking to address this challenge in their own work. While randomization of policy rollouts remains a preferred design for assessing the causal effects of social policies, careful selection of the research question and analytic approach, guided by deep substantive knowledge and creativity, can go a long way to overcome policy clustering and deliver stronger evidence on the health effects of social policies.
Data Availability
All data used in this study arise from published studies that are publicly available.
Footnotes
Funding This work was supported by the Evidence for Action program of the Robert Wood Johnson Foundation.
Conflicts of interest None declared.
References
- 1.↵
- 2.↵
- 3.
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.
- 54.
- 55.↵
- 56.
- 57.
- 58.
- 59.
- 60.
- 61.
- 62.
- 63.
- 64.
- 65.
- 66.
- 67.
- 68.↵
- 69.
- 70.
- 71.
- 72.
- 73.
- 74.
- 75.
- 76.
- 77.↵
- 78.
- 79.
- 80.↵
- 81.
- 82.
- 83.
- 84.
- 85.
- 86.
- 87.
- 88.
- 89.
- 90.
- 91.
- 92.
- 93.
- 94.↵
- 95.
- 96.
- 97.
- 98.
- 99.
- 100.
- 101.
- 102.
- 103.
- 104.
- 105.
- 106.
- 107.↵
- 108.↵