Friction in governing institutions
To test H2 concerning the effect of institutions on the dynamic representation of public priorities it is necessary to first assess the degree of friction present in each of the policy agendas and to compare processes along the policy cycle within each country. It is expected that those government processes subject to higher decision costs also tend to be associated with higher levels of friction (e.g., Baumgartner et al. 2009). To ascertain the degree to which decision making within each of these governing institutions is subject to friction against change we use stochastic process methods (following Breunig & Jones 2011; Jones & Baumgartner 2005a, 2005b; Jones et al. 2003). This tests the normality of the distribution of values of the dependent variables for this analysis (i.e., the policy content of executive speeches, legislation and budgetary expenditure). With the exception of budgets, the ‘percentage-percentage’ calculation method is used to analyse the overall distribution of (percentage) change of the difference between (percentage) agenda share in one year and the next (see Baumgartner et al. 2009: 610). Budgetary spending is not treated as bounded in the same way (i.e., as a percentage) because the numbers involved are much larger and because the values are reported in real prices, which removes variation due to inflation. Note that for the executive and legislative agendas, cases in which attention to a particular topic remains stable at zero are treated as missing to avoid over-inflation of the kurtosis scores.
The distribution of year-on-year percentage changes is presented in Figure 1, with the kurtosis scores super-imposed on each histogram. If the value of the kurtosis statistic is greater than three, the distribution exhibits positive kurtosis and can be said to be leptokurtic, consistent with the presence of a high degree of cognitive and/or institutional friction. In addition, the Shapiro-Wilk test is presented, which considers whether the sample is drawn from a normal distribution.
Figure 1. Change distributions in executive, legislative and budgeting agendas in the United States and the United Kingdom, 1951–2003.
Download figure to PowerPoint
Overall, the results are consistent with theoretical expectations regarding the level of friction in each of the governing institutions: the highest level of kurtosis in both countries is observed for budgetary expenditure, consistent with a large number of studies that show that public budgeting exhibits incrementalism, interspersed with occasional extreme disturbances (e.g., Jones & Baumgartner 2005a, 2005b; Jones et al. 2003; Baumgartner et al. 2009). This is evidenced in the tall slender peak and the fat tails of the distributions plotted in Figure 1. Indeed, all of the institutions exhibit some evidence of disproportionate information-processing in the leptokurtic distributions of attention change. For the United Kingdom, the pattern of institutional differences is clear, with the lowest kurtosis observed for the executive agenda, the next lowest for the legislative agenda and the highest for budgets. This is unsurprising in light of the incrementalist tendencies of decision makers and the numerous veto points in budgeting processes. In the United States, however, the level of kurtosis is in fact higher for the executive agenda than for legislative outputs. This reflects a greater degree of variation in how different presidents use the State of the Union Address. For example, the shortest State of the Union message, delivered by President Nixon in 1973, contained just 36 policy-related statements, while the longest, presented by President Carter in 1981, contained 1,336. Part of this variation is no doubt due to the unique nature of these particular messages, which were written instead of being delivered orally and due to the uniqueness of individual presidential character (Neustadt 1960). While the unique nature of the State of the Union Address puts the findings in Figure 1 somewhat at odds with expectations concerning the ordering of levels of institutional friction, the results nevertheless confirm the presence of punctuations. It is further expected that these patterns of institutional friction structure the interaction of the priorities of the public and the policy content of governing agendas.
Error-correction models of dynamic agenda representation
To test the dynamic representation of the issue priorities of the public in governing agendas, comparing both across institutional levels and across countries, time-series error-correction models are now estimated according to policy topic. The use of an error-correction model (ECM) enables diagnosis of both the short- and long-run effects of the issue priorities of the public on public policy.8 The error-correction framework is selected in light of past studies which demonstrate that, in both theory and practice, agenda-opinion dynamics ‘coexist in a long-run equilibrium state that is subject to short-run corrections’ (Jennings & John 2009: 838). In other words, dynamic agenda representation can arise from long-term trends in public priorities and from short-run variation and shocks – that is, events such as the global financial crisis. An ECM framework is appropriate when testing for both contemporaneous and lagged effects. The model can be represented in the form:
where short-run changes in the policy agenda relating to a particular issue (ΔAGENDAt) are a function of short-run changes in the public's prioritisation of that same topic (ΔOPINIONt), the long-run changes (OPINIONt−1) and where the lagged value of the dependent variable (AGENDAt−1) measures the speed of re-equilibration () in response to shocks to the long-run agenda-opinion equilibrium. Consistent both with our theoretical expectations and other models of dynamic representation (e.g., Wlezien 2004; Jennings & John 2009), this model includes a variable (PARTYt) to capture the contemporaneous effects of indirect representation through partisan control of government. This controls for difference in the governing agendas of political parties, and is coded 1 for the Conservative Party in the British case and the Republican Party in the American case and is coded 0 for the Labour Party in the United Kingdom and for the Democratic Party in the United States.9
Within the ECM framework, changes in the policy content of governing agendas are estimated as a function of contemporaneous changes in the issue priorities of the public and the degree to which these are outside the long-run agenda/opinion equilibrium. This suggests that if the governing agenda deviates from its long-run equilibrium, as the institution commits either ‘too much’ or ‘too little’ attention to a particular issue, responsiveness is equal to the degree of equilibration that restores correspondence between the agenda and public opinion to its previous status quo.
To first assess the general pattern of dynamic agenda representation, a time-series cross-sectional ECM is estimated for the seven policy topics that are common to each institution. This pooled model specification has the advantage of measuring the degree to which the policy agenda of each governing institution is representative, overall, of public priorities. The results that are reported in Table 3 show rather limited evidence of dynamic representation: the long-run (lagged) effect of the issue priorities of the public is positive and significant, at the 95 per cent confidence level, for the executive agenda in the United States and for the legislative outputs in the both the United States and the United Kingdom, while the short-run effect of public priorities is significant for legislative outputs in the United Kingdom. Furthermore, there is no evidence of representation in the link between public priorities and changes in budgetary expenditures in either country. In addition, the negative and significant error-correction parameter (AGENDAt−1) for the executive and legislative agendas indicates that shocks to the long-run equilibrium of the policy agenda are corrected over time. The absence of a similar effect for budgetary expenditure (indeed, the parameter is positive and significant for the United Kingdom) suggests that errors tend to accumulate. This is consistent with the high level of institutional friction in budgeting compared to the other venues, with the build-up of errors contributing to punctuations in policy change. The absence of a significant link between public priorities and the executive agenda in the United Kingdom, in contrast, might be attributable to the lack of a general pattern of representation, rather than suggesting that the policy agenda is unresponsive on all topics.
Table 3. Time-series cross-sectional model of public priorities and government agendas
|Δ Opiniont||0.233 (0.321)||−0.053 (0.033)||0.077 (0.087)||0.072** (0.027)||258.254 (202.967)||39.629 (31.735)|
|Opiniont−1||0.890*** (0.171)||0.020 (0.015)||0.228*** (0.049)||0.048*** (0.013)||96.024 (105.212)||−14.862 (14.928)|
|Agendat−1||−0.864*** (0.053)||−0.518*** (0.047)||−0.622*** (0.050)||−0.994*** (0.053)||0.013 (0.010)||0.020** (0.007)|
|Partyt||−2.186 (3.472)||−0.468 (0.331)||−6.013*** (1.287)||0.275 (0.276)||6064.818† (3140.560)||−336.452 (324.371)|
|Constant||14.375*** (3.307)||3.414** (0.446)||7.390*** (0.887)||3.556*** (0.333)||803.301 (1971.571)||739.139† (389.697)|
While the use of a time-series cross-sectional framework enables comparison of the structure of representation across institutions and across countries, it is possible that unresponsiveness on some issues might cancel out responsiveness on others, obscuring important features of the underlying categories – that is, the policy agenda might be responsive for a number of the most salient topics (such as the economy and defence) but this would not necessarily be reflected in the pooled analysis. Indeed, variation in rates of responsiveness for individual issues is to be expected due to differences in the intrinsic importance of certain issues to citizens (Page & Shapiro 1983; Jones 1994; see Burstein 2003 for a review) and because government agendas possess policy ‘tools’ or instruments that are optimised to solve certain types of problem (Hood 1983). Therefore, an issue-level analysis is needed to fully test the dynamic inter-relationship between public priorities and policy agendas.
Individual topic analyses
To assess issue-specific patterns of dynamic agenda representation in each of the institutional venues, 7810 individual ECMs are next estimated. Table 4 summarises the findings on the short-run and long-run effects of the issue priorities of the public on the policy content for each of the institutional agendas, in each of the countries, by presenting the coefficient estimates for each topic (full results are reported in the Online Appendix, Tables A1 to A6, see the Supporting Information section at the end of the article). The topics are reported in the first column and the responsiveness coefficients for each of the governing agendas are presented in turn across the columns of Table 4, enabling comparison across the two countries and across institutions. For each topic, the direction, size and significance of the responsiveness of governing agendas is measured with the short- and long-run effects of public issue priorities. The final three rows of Table 4 further indicate the general pattern of dynamic agenda representation, summarising the total number of topics for which the issue priorities of the public have a positive short-run, long-run or either effect on the policy content of each institutional agenda.
Table 4. Summary of the effects of public priorities on government agendas
| 1 – Economy||Short||−0.064||0.056||0.129*||0.080†||0.441||76.661|
| 2 – Civil||Short||−0.156||−0.333*||−0.119†||0.088|| || |
|Long||−0.140||−0.399*||−0.101†||0.038|| || |
| 3 – Health||Short||0.515||−0.030||−0.558*||0.072||72.673||−12.042|
| 4 – Agriculture||Short||4.435*||−0.090||1.929†||−0.206|| || |
|Long||2.274||−0.169||0.740||−0.322|| || |
| 5 – Labour||Short||−1.001||0.040||0.253||0.007|| || |
|Long||1.826||0.059†||−0.247||0.055†|| || |
| 6 – Education||Short||6.476**||−0.031||0.487||0.110||−1675.423*||49.732|
| 7 – Environment||Short||3.407**||0.376*||1.846†||−0.396†|| || |
|Long||0.249||0.527***||2.293**||0.490*|| || |
| 8 – Energy||Short||0.628||0.020||0.215||0.058|| || |
|Long||2.759***||0.236||0.446*||−0.006|| || |
|10 – Transport||Short||14.760||0.500†||14.725||0.259|| || |
|Long||47.080||0.121||36.067||0.517|| || |
|12 – Law||Short||1.312**||0.032||0.085||−0.178||−5.905||−82.274|
|13 – Social||Short||1.860†||0.064||0.160||−0.113||1151.909||−106.974|
|14 – Housing||Short||3.028||−0.174||1.321||−0.132||−6772.070||−216.407|
|18 – Trade||Short||−1.080||0.070||−3.594*||0.030|| || |
|Long||0.600||0.123||−5.615**||0.097|| || |
|20 – Government||Short||−0.617||0.359||1.370||−0.502|| || |
|Long||0.247||0.134||−2.049†||−0.695|| || |
|21 – Lands||Short||−34.261||0.123||−383.446||0.073|| || |
|Long||15.206||0.019||−269.913||0.445**|| || |
|16/19 – Foreign||Short||−1.191||−0.367**||0.446*||0.122*||403.245||12.608|
These results show that the policy agendas of governing institutions in both countries exhibit a substantial degree of responsiveness to the issue priorities of the public (H1). Specifically, statistically significant short- or long-run effects of public priorities are observed in four or more of the 16 topics for the executive agenda and legislative outputs. In the United States, the executive agenda is responsive to public priorities for seven out of the 16 topics (health, agriculture, education, environment, energy, law and order, and social welfare), while in the United Kingdom it is responsive for six (macroeconomic issues, health, education, environment, law and order, and housing). In the United States, legislative outputs are responsive for five of the 16 (macroeconomic issues, environment, energy, social welfare, and defence and foreign affairs) and in the United Kingdom for four (macroeconomic issues, environment, public lands and territorial issues, and defence and foreign affairs). It is interesting to note that there is little overlap in responsiveness within each country from institution to institution. This suggests these institutional agendas have different competencies as we discussed earlier. Furthermore, our example concerning education holds, with executive agendas in both countries demonstrating significant responsiveness to public priorities, with no responsiveness in either legislative agenda.
In contrast to these findings, there are no significant effects for budgetary expenditure in any of the seven topics tested here. This is consistent with the theoretical expectation that a decision-making venue like budgeting which lacks the same pressure for serial-processing and the prioritisation of issues is less likely to be responsive to public priorities. Further, in a few cases across the executive, legislative and budgetary agendas the coefficient estimates are negative and significant suggesting that change in the policy agenda leads public concern about that issue. For example, a short-run increase in American spending on education precedes a decrease in the public's prioritisation of the same issue. It is therefore possible for agendas to shape public priorities.
Overall, these results provide strong evidence of a link between the issue priorities of the public and the policy agenda of governing institutions in both countries (H1). Of all institutional venues, public priorities have the strongest effect, on average, on the executive agenda in terms of the frequency of statistical significance of short- and long-run effects. The results do not enable direct comparison of the level of responsiveness either between countries or institutions because the measure of the policy agenda in each institution uses a different unit of analysis (i.e., speeches, legislation and expenditure) and the mean level varies between countries. Further, the volume of statutes passed in the United States far exceeds the volume of acts passed in the United Kingdom.11
The higher levels of agenda representation observed for executive speeches followed by legislation, in comparison to budgetary spending, indicate variation between institutional settings, consistent with theoretical expectations (H2) and the pattern of friction reported earlier. The responsiveness of policy agendas to the issue priorities of the public is highest in institutional settings where attention is most scarce (executive speeches), which requires that decision makers prioritise issues, and is lowest in the institutional venue subject to the highest degree of friction (budgetary expenditure).
While there is variation at the topic level, there are many similarities between the United States and the United Kingdom in the responsiveness of governing institutions to the issue priorities of the public. For example, there is no opinion-responsiveness in the executive agenda or budgetary spending for defence and foreign affairs in either country while there is evidence of responsiveness in legislative outputs in both countries in the short- and the long-run (with the long-run effects in the United Kingdom being significant at the 90 per cent confidence level). For law and order, on the other hand, the executive agenda is responsive to public priorities in the long-run in both countries (as well as being responsive in the short-run in the United States), while public priorities do not have a significant effect on legislative outputs for the same topic in either country. There is, nevertheless, variation in responsiveness across institutions between the countries that is likely driven by differences in the level of importance of issues to citizens (Page & Shapiro 1983; Jones 1994; also see Burstein 2003) and due to the ability of particular institutional agendas to better address certain issues (Hood 1983). Institutional variation therefore structures the overall pattern of representation (H2), where governing agendas in which the attention of decision makers is most scarce or subject to lower levels of friction tend to be more responsive, but also leads to variation in dynamic representation across issues.
The non-responsiveness of budgetary expenditure is an interesting finding that further confirms previous evidence that spending is not responsive to public concern about the ‘most important problem’ in contrast to relative preferences (Wlezien 2005). This is perhaps because whereas budgets have directional implications and the public can prefer either more or less spending in a particular policy domain, changes in the issue priorities of the public do not signal the desired direction of change (Jennings & Wlezien 2012). For example, the issue of health care might be highly salient to the public either because the government is spending too much or too little on it. The issue priorities of the public are, however, a useful information signal for valence issues – that is, those issues on which there is broad consensus over ends, such as lower crime or economic growth (Stokes 1963).