State Revenue Forecasting Practices: Accuracy, Transparency, and Political Acceptance

Transcription

1 State Revenue Forecasting Practices: Accuracy, Transparency, and Political Acceptance A Volcker Alliance Project Paper March 18, 2017 ASPA 2017 Annual Conference Emily Franklin Public Finance Fellow Center for State and Local Finance Andrew Young School of Policy Studies Georgia State University efranklin8@student.gsu.edu Carolyn Bourdeaux Associate Professor Director, Center for State and Local Finance Andrew Young School of Policy Studies Georgia State University Alex Hathaway Public Finance Fellow Center for State and Local Finance Andrew Young School of Policy Studies Georgia State University Special thanks to the Volcker Alliance for their research support and permission to use their data in this analysis. DRAFT: Please do not cite without permission of the authors.

2 ABSTRACT This paper will discuss the budget practices around revenue forecasting. Recent research has discussed how revenue forecasting is not just a matter of accuracy but also is important for transparency and political acceptance of revenue forecasts as a guide to the budget process. At the same time, recent research has pointed out there is significant variation in consensus forecasting practices. This paper will look at the diversity of revenue forecasting processes across the 15 states and assess the extent to which they have proven to be accurate, transparent, and politically accepted between FY15 and FY16.

3 INTRODUCTION The state revenue forecast sets the tenor for budget deliberations in the U.S. states. Because most states operate under a balanced budget constraint, the revenue forecast establishes a foundation for fiscal discipline since it sets a cap on state spending. Almost every state has evolved to have a unique process around revenue estimation and scholars have long debated whether there are different processes or methodologies that are more accurate, more likely to lead to broad political acceptance and generally provide a firmer foundation for fiscal discipline. The academic literature on these different processes is vast and by no means conclusive regarding best practices. This paper briefly reviews the literature on revenue forecasting and then draws on a dataset produced as part of the Volcker Alliance Truth and Integrity in Government Finance project to examine in detail the revenue forecasting processes in 15 southeastern states. In particular the paper assesses the process and actors involved in forecasting, the methodological approach, the accuracy of the forecast, and ultimately, the political acceptance of the forecast. The analysis draws on the rich detail the researchers have available for each state to explore some of the more interesting state practices around forecasting. LITERATURE REVIEW Forecasting Processes One of the most vibrant debates in the literature on state revenue forecasting centers around the process associated with revenue forecasting. Revenue forecasting is roughly divided into three types: executive, consensus and separate (where the legislative and executive branch and even the different parties within the legislative branch develop separate forecasts).

4 1 Perhaps the most widely recommended process is a consensus forecast According to the National Association of State Budget Officers (NASBO), a consensus revenue forecast is a revenue projection developed in agreement through an official forecasting group representing both the executive and legislative branches. 6 However, these categories mask wide variations, particularly around consensus forecasting, which can involve a number of different combinations of legislative and executive policy-makers, including involvement by legislative and executive staff, elected officials or even elected officials explicitly from opposing political parties as well as involvement by non-partisan external parties such as academics. A common question is the extent to which these process differences are linked to accuracy, transparency or more recently, political acceptance of the forecast. Accuracy Because revenue collections are affected by state, national and global economic conditions, they are difficult to predict. Researchers have looked at both the association of accuracy with consensus forecasting process as well as whether different techniques of forecasting produce more accurate results. There is extensive literature showing that combining independent forecasts increases accuracy. 7 Updating revenue forecasts as close to the start of the fiscal year as possible has also been shown to increase accuracy. 8 Some studies have found that consensus revenue forecasting improves the accuracy of forecasts, though it is not clear why it improves accuracy. Consensus forecasts may lower the percentage of forecast misses. One possible reason is the process often involves a number of experts on different parts of the economy, who can provide a wealth of economic information, as well as ensure the forecasting method is less politically influenced. In addition, if different forecasts are combined during the process of forming the official forecast, then accuracy may increase. Voorhees (2004), however,

5 found less of an effect on accuracy after controlling for frequency of state s forecasts, independent expert input and other factors. 13 Boyd and Dadayan (2014) agree that consensus revenue forecasting can help insulate the forecasting process from undue political influences, but their study did not show that it contributed to the accuracy of the forecast. 14 A further dimension to the revenue forecasting debate is the extent to which the forecast is subject to political influence. Several studies have found forecasting errors increased due to politically opportunistic behavior when governors have a balanced-budget requirement, when the forecasting group is linked with the government, or when elected officials have longer term limits or long-term unified government. The creation of independent groups with knowledge of the forecasting domain, however, reduces forecasting error because they reduce error due to human bias or political expediency In contrast, Mikesell and Ross (2014) found that the Indiana revenue forecast process (which has not only executive and legislative consensus, but also Republican and Democrat party representation) outperformed various naïve revenue forecasting processes and had the further benefit of building political consensus around the forecast. Certain fiscal policies can also predict more transparency around the revenue estimation process. For example, Rose & Smith (2011) found that states that created budget stabilization funds (BSFs), with policies governing how much expected surplus would be put into the account and how much could be taken out and used for other purposes, also decreased their revenue forecast bias. 19 Transparency The level of transparency surrounding the estimation process is another relevant aspect of revenue forecasting. The Government Finance Officers Association (GFOA) recommends governments are upfront about their forecasting philosophies. For instance, some governments

6 prefer to make conservative forecasts, which may make balancing the budget more difficult, but reduce the risk of a shortfall. Other governments try to make an objective forecast, which seeks to be as close to actual revenue collections as possible. These forecasts make it easier to balance the budget, but also increase the risk of a shortfall. The GFOA also recommends that governments clearly present underlying assumptions and methodologies in the final budget document. 20 The Organisation for Economic Co-operation and Development (OECD) and the International Monetary Fund (IMF) recommend that key economic assumptions be disclosed in budget documents, as well. Such assumptions can include gross domestic product (GDP) growth, the employment and unemployment rates, and inflation. Furthermore, an analysis of the impact of these macroeconomic trends on the budget should be publicly available However, McNichol, Lav and Leachman (2015) found that most states do not clearly tie these underpinning economic assumptions to the revenue forecast. A few states, such as Alabama, do not publish these assumptions at all. 23 Different factors predict the level of transparency around the revenue forecast. Alt, Lassen and Rose (2006) provide some anecdotal accounts of states adopting consensus revenue forecasting in an effort to increase fiscal transparency, such as Delaware and Rhode Island. These transparency efforts were generally adopted in periods of fiscal stress, split governments or a recent party turnover in the governor s office. 24 Political Acceptance Recently, Mikesell and Ross (2014) raised the issue of political contention around the revenue forecast, pointing out that both the executive and legislative branches need to be involved in the process in order to build consensus and reduce conflict over the final estimate. 25 Ironically, other researchers have called for de-politicizing the revenue forecasting process for

7 this very same reason There are a number of studies with anecdotal accounts of states adopting consensus forecasts for political reasons; however, the researchers were unable to locate a study that looks at this issue on a national scale. States seem to have adopted a mixture of both approaches. As of 2014, as many as 24 states have adopted some form of consensus forecast, while 36 include non-government experts in their forecasting process. 18 states have both a consensus forecast and non-governmental experts in their forecasting body. 34 This analysis examines three key aspects of revenue forecasting in 15 southeastern states: the level of accuracy, the level of transparency, and the level of political acceptance. Though there is evidence that combined forecasts and independent expert participation increases accuracy, the verdict is still out on whether or not a consensus forecast increases accuracy. Regarding transparency, the GFOA and others recommend that states publish detailed methodologies with underlying macroeconomic assumptions along with their forecasts. There are anecdotal accounts of consensus forecasting increasing fiscal transparency. An interesting question is: are states with consensus forecasts more likely to publish detailed forecast methodologies and underlying macroeconomic assumptions because they are more likely to involve external experts in making the forecast? Finally, regarding political acceptance and the revenue forecast, there are a number of studies that describe states adoption of consensus forecasting processes for political reasons; however, we were unable to find a study that looked at this issue on a national scale. While recognizing the endogeneity of a consensus forecast to inter-party political conflict, a further question of interest is whether there is any evidence that consensus forecasts are more likely to be accepted as legitimate by policy-makers than a pure executive forecast or a forecast where the legislative and executive branch each act independently.

8 Using data compiled for the Truth and Integrity in Government Finance project by the Volcker Alliance, this paper examines state revenue forecasting budget practices and the extent to which they reflect the best practices described above. While the data is limited, the paper also assesses whether there is any evidence of a relationship between these best practices and the extent to which the forecast is accurate and viewed as legitimate by different political actors. The research then draws on some rich contextual detail and case study material available to the researchers to further examine these issues. METHODS The Volcker Alliance s project involved organizing a group of universities to conduct a survey of budget and fiscal practices of all fifty states for the fiscal years 2015, 2016 and The project was organized around assessing whether and to what extent states are facing structural deficits and several questions focused on revenue forecasting processes, multi-year revenue and expenditure forecasts, revenue growth projection rationales, and midyear budget adjustments. There were 29 questions in total, but, for this paper, the analysis focuses on the data gathered for the following five questions: 1. Does the state disclose consensus revenue forecasts in budget documents? 2. Does the state disclose multi-year revenue forecasts (at least 3 years) in budget documents? 3. Does the state disclose multiyear expenditure forecasts (at least 3 years) in budget documents? 4. Does the state reasonably support revenue growth projections at time of initial budget? 5. Was there a need for a meaningful (i.e., greater than 1%) midyear budget adjustment? Since all states were still in the middle of FY2017 at the time of this analysis, only FY2015 and FY2016 responses were considered. For the purposes of the project, budget documents were considered the Governor s recommended budget, enacted appropriations bills,

9 as well as supporting documents such as legislative analysis of the budget and presentations or revenue forecasts produced in tandem with the budget. Consensus revenue forecasts were defined as forecasts with both executive and legislative participation. Additionally, for five states, Maryland, Virginia, North Carolina, South Carolina, and Georgia, the analysis draws on more detailed case study information on the shifts in the revenue estimate during budget development, adoption and implementation. Because questions of accuracy and transparency cannot be entirely addressed based on the Volcker project questions, the analysis also supplements this information with material collected from the National Association of State Budget Officers (NASBO) Fiscal Survey of the States as well as more in depth reviews of revenue forecasting and budget documents across all of the states to examine the methodology in the forecasting process as well as participation in consensus and executive forecasting processes. RESULTS Forecasting Processes in the Southeastern States This section discusses different aspects of the revenue forecasting processes in the southeastern states, including the number of states with consensus forecasts, types of membership in the states forecasting groups (executive, legislative, non-partisan), and multiyear forecasts. [Table 1: Forecasting Processes in the Southeastern States] Table 1 describes the membership of the states forecasting groups and whether or not they include executive, legislative or non-partisan members. Most southeastern states have a consensus revenue forecast-type process (10 out of 15 states); however, four have executive driven processes. Only Alabama has a forecast developed separately by the executive and legislative branches.

10 As noted earlier, states define consensus in different ways with varying degrees of nonpartisan, staff and elected official participation and even the executive dominated processes have different external and staff advisement arrangements. In some states, such as Maryland, only non-partisan staff are involved in the revenue forecasting process, although the process is still considered consensus because the staff are from the executive as well as the legislative research offices. Other consensus, staff driven processes include Florida, Kentucky, North Carolina, and Mississippi. Virginia has an interesting hybrid process with two committees working on the forecast. The first committee is a staff or expert committee, the Joint Advisory Board of Economists (JABE). This committee works on the detailed methodology behind the forecast and produces several options for consideration by elected policy-makers. The Governor s Advisory Council on Revenue Estimates (GACRE) includes legislative leadership as well as the Governor and they ultimately vote on the revenue estimate to use. No southeastern state statutorily requires the participation of both Republican and Democrat politicians. In Delaware, however, some representation from each party is generally included. After an initial review, Delaware appears to be the only state to explicitly include minority party members in its forecasting group, although other states may have multi-party representation if the legislative majorities in either Chamber are from the opposing party. All fifteen states include the governor or some other executive membership, such as an executive budget office, in their forecasting group. Georgia and West Virginia appear to be the only states that exclusively rely on the Governor and his or her direct staff to develop the revenue estimate. Oklahoma is an executive driven process but includes leadership from across the executive branch, including the Lieutenant Governor and Attorney General.

11 The majority also include some sort of legislative membership, whether elected officials or professional staff from legislative research offices. Only a very few include parties from outside government. These include Virginia s JABE and GARCE and the Delaware forecasting group, which have citizen appointees and Louisiana and Alabama which have academic appointees. Accuracy of Southeastern States Revenue Forecasts This section discusses the accuracy of revenue forecasts in the fifteen southeastern U.S. states. Table 2 shows midyear adjustments to the budget, forecasting errors and accuracy grades. [Table 2: Accuracy of Southeastern States Revenue Forecasts] Table 2 shows the response to the Volcker Alliance survey assessment on whether a state made a mid-year adjustment greater than one percent for FY2015 and FY2016. This is paired with a fairly objective assessment of forecast accuracy pairing the general fund forecast with the general fund year-end actual collections from NASBO. The table shows the forecasting error as a raw percentage, which shows whether the state over or under estimated revenues, and absolute value, which helps compare the size of the forecast miss. To more easily compare the forecasting errors, we simply assigned a letter grade with errors between 0 and 1 receiving an A, errors between 1 percent and 2 percent a B, errors between 2 and 3 percent a C, errors between 3 and 4 a D, and errors above 4 percent an F. The grades are the same regardless of whether the error is positive or negative. While an A grade signals a more accurate forecast, an F grade signals a less accurate forecast, Willoughby and Guo (2008) considered less than five percent forecasting error to be a relatively accurate forecast and a Pew-Rockefeller report on revenue forecasts generally found that the mean forecast error was around 3 percent States with D s and F s are outside of the average forecast error.

12 While some important considerations around revenue forecast accuracy will be considered later, just taking these forecasts at face value, in FY2015, the average forecast error (the absolute value of the percentage difference between the forecast general fund revenues and actuals) for the 15 states was two percent. In FY2016, the forecast error was 2.5 percent. This average suggests the southeastern states on average are beating the forecast errors predicted by the Pew-Rockefeller report -- even including states that are intentionally low-balling their forecasts. FY2016 appears to be particularly affected by serious revenue declines in the energy sector dependent states of Oklahoma, Louisiana, and West Virginia, but also by some large misses to the positive in Georgia, South Carolina and Tennessee. The standard deviation for the actual to forecast percentages is 3.72 for 2016 while only 2.58 for Drawing on the more detailed data from various case studies suggests some important considerations when simply taking the difference between forecast and actual revenues at face value. For instance, in the 2015 legislative session Georgia passed transportation legislation that increased tax and fee revenues by $870 million in FY2016. However, the state did not add these new revenues to their budget until the middle of the fiscal year 2016, which makes the error rate seem like 7.38 percent. Additionally, Georgia annually adds a one percent of net revenue increase each year to K-12 education in the mid-year budget out of the revenue shortfall reserve. The state essentially pre-funds this by lowballing the revenue estimate in order to force the state to reserve this amount the year prior. Finally, Georgia s governor publicly committed to building a $2 billion reserve fund before he left office in Georgia s reserve is replenished from any surplus year-end funds, so the Governor has further low-balled the revenue estimate to force sufficient surpluses to rebuild the reserve. None of this is explicit in any budget documents so there is no

13 way to know what the true underlying state forecast actually is. While Georgia does not win points for transparency, the state s overly conservative revenue estimates do not necessarily reflect poor forecasting capacity, but an intentional effort to significantly underestimate the revenues. Of the other southeastern states, South Carolina and Tennessee also appear to be similarly substantially underestimating revenues. Like Georgia, South Carolina has certain designated current year uses for prior year reserves, in their case a reserve for capital outlay. The state actually uses this capital outlay reserve as a first resort rainy day fund prior to tapping their revenue shortfall reserve. However, according to a budget official, South Carolina does not make conservative revenue estimates in order to replenish this fund, but to avoid mid-year budget cuts and year-end deficits. 1 On the other end of the spectrum, six of the fifteen states over-projected their revenues, and four over projected by greater than one percent in FY2015. Not surprisingly, these four states were also forced to make negative mid-year adjustments of greater than one percent. 2 Louisiana, Oklahoma and West Virginia were affected by unanticipated declines in the energy sectors, and in FY2016 face even more dramatic overestimates of revenues; however, Virginia is an anomaly. In 2016, the state gets an A for its revenue forecast, and is the only state to swing from an F to an A. The challenge in Virginia is that the state appears to have intentionally gone into FY2015, the second year in its biennium, knowing that it would overshoot its revenue 1 Interview with Executive Budget Office official, South Carolina Department of Administration 2 Two other states also recorded making mid-year adjustments greater than one percent, even though revenues were tracking to the good. Arkansas s mid-year adjustments are actually pre-planned. As part of the budget process, the state adopts a set of mid-year spending priorities if the state is on track to make budget. If the revenues track below estimate, these second tier priorities are put on hold for the fiscal year. Georgia actually is similar in that it has a pre-planned mid-year adjustment for K-12 growth out of the revenue shortfall reserve, which is required to be one percent of prior year revenues. Additionally, in FY2016 the state added in the tax revenues from the transportation tax. Depending on whether one counts these adjustments, then the state also made a greater than one percent mid-year adjustment like Arkansas. Tennessee continues to be an anomaly.

14 estimate. According to one account, the state had to formally overshoot its revenue estimate in order to access its reserves. 3 The state statute stipulates that the state can only access the revenue shortfall reserve if the general fund revenues appropriated exceed the forecast by more than two percent. Despite passing an adjusted FY2015 budget in June of 2014 and making significant provision for an anticipated shortfall, the legislature did not formally lower the revenue estimate until after an August 2014 revised revenue estimate (which in turn, was legally triggered by problems with the FY2014 revenue estimate). The legislature then formally inserted the reserves into the revised budget in December of In detail, the state ended the regular Session without a budget due to a budget impasse, and the impasse continued into a Special Session. By May, it was clear that the state would not meet its forecast for FY2014, which, on July 1 st, would legally trigger a new forecast for the FY biennial budget period. Because the state did not want to use a revised forecast in adopting its new budget, the impasse ended, and a budget was adopted based on the old revenue forecast from December Now the forecast would not be revised until after the start of the new fiscal year, allowing the state to reopen the newly adopted budget and use the Revenue Stabilization Fund. Virginia s constitution does not allow the use of the Revenue Stabilization Fund in building a budget, which is why it was imperative to adopt the new budget using the December 2013 revenue estimate. 4 Given these variations in managing the revenue estimate, the metric of forecast error may be something of a misnomer for many states. Rather than reflecting any lack of internal capacity or methodological problems, forecasting error at the state level may be more appropriately characterized as a problem of transparency or strategies for management of legal and institutional arrangements. Revenue forecasts to some degree have a unique role in the 3 Interview with Budget Director, Virginia House of Delegates. 4 Interview with Staff Director, Virginia House Appropriations Committee.

15 budget process that goes beyond accuracy around expected revenues. For instance, in Georgia, the balanced budget requirements are tied to the revenue forecast. Since the executive unilaterally sets the forecast, the forecast is used as a policy lever to bound legislative expenditures rather than as a device to actually communicate anticipated revenues. As noted in the literature, a further confounding factor is that some states explicitly adopt an accuracy approach, while others explicitly try to undershoot the estimate. Given that many of the D and F scoring states either face problems like difficult to predict energy prices or have revenue forecasts that do not accurately reflect their actual revenue expectations, it is difficult to assess the relationship between forecasting process (e.g., consensus, etc.) and outcome. A brief look at the states that appear to produce forecasts that are intended to accurately reflect expected revenues shows no evidence of any clear linkage between a consensus forecast and accuracy of the forecast. The states that received A or B grades in FY2015 and FY2016 include Alabama, Arkansas, Delaware, Florida, Maryland and Mississippi; however, these include a mix of forecast process types. Four out of the six have a consensus process, but the committees working on the process are quite varied. Arkansas has an executive forecast and Alabama has a separated process. Obviously, a more sophisticate quantitative analysis is warranted, but as noted above, such an analysis would need to carefully consider actual revenue expectations relative to the reported revenue forecast. Transparency of Southeastern States Revenue Forecasts This section discusses the extent to which the assumptions and methodology underpinning the state revenue forecast are not explained, explained at a high level, but without a clear connection to the forecast, and explained at a high level and clearly linked to the final forecast.

16 [Table 3: Transparency of Southeastern States Revenue Forecasts] In order to create Table 3, the researchers looked at the official revenue forecasting documents for each of the fifteen southeastern states, searching specifically for information on general macroeconomic trends in the forecasting document. Such trends could include housing starts, employment rates, and US gross domestic product (GDP). The researchers wanted to see if states were following best practice as recommended by GFOA and others by linking these macroeconomic trends to the forecast. For instance, the state might link the income tax forecast to personal income forecasts. If the document did not contain such information, the state received an X in the column labeled No Macroeconomic Information Explaining the Forecast. If the forecasting document did contain information on macroeconomic trends, but did not make clear connections between those trends and the state forecast, the state received an X in the column labeled General Macroeconomic Trends without a Clear Link to the Forecast. Finally, if the state did clearly link the forecast to general macroeconomic trends, then it received an X in the column labeled Detailed Methodology with Direct Links to the Forecast. Four of the 15 states received an X in the first column, No Macroeconomic Information Explaining the Forecast. These states only included a chart with the forecast numbers, with no explanation of how these estimates were reached. Nine of the fifteen states received an X in the second column, General Macroeconomic Trends Without a Clear Link to the Forecast. These states included narrative about macroeconomic trends with a general description of how they may affect the state economy; however, they did not detail their methodology for how these trends brought them to their forecast number. Only Virginia and Florida received an X in the third column, Detailed Methodology with Direct Links to the Forecast. These states clearly linked the forecast numbers with these

17 trends or included detail on their methodology for how the macroeconomic trends brought them to their forecast. For example, Florida s estimate of revenue from the ad valorem tax is directly linked to new construction and other important variables that affect that source of revenue. All of these statistics are available online, so it is fairly easy to see how the ad valorem revenue estimate was arrived at by the revenue estimating group. Virginia includes its revenue source calculations in The Economic Outlook and Revenue Forecast, prepared by the Virginia Department of Taxation for one of its forecasting groups, the Governor s Advisory Council on Revenue Estimates. These calculations show directly how the estimate for each revenue source was developed. 5 While the results of the analysis are interesting from a normative perspective, there is no clear interrelationship between accuracy, consensus forecasts or transparency around the methods that went into the forecast. Both Florida and Virginia have consensus forecasts. Florida received a B accuracy grade in fiscal year 2015 and an A in Virginia received an F in fiscal year 2015, but an A in fiscal year 2016; however, as noted earlier, it does not appears that the revenue forecast in the budget actually reflected revenue expectations. That being said, Virginia did struggle during the FY2015 year to arrive at a solid forecast, significantly downgrading a forecast that ultimately turned out to be close to what the state had initially predicted prior to the start of the fiscal year. But the state also did a significant and public analysis of how the revenue forecast missed and what steps the state would take going forward to avoid such errors. By way of contrast, Alabama does not include the macroeconomic assumptions underlying its forecast, but it received A s in both years studied. Generally, most 5 It is pretty clear from Table 3 that different universities evaluating this question for the Volcker Alliance used different decision criteria when assessing whether states provided a reasonable rationale for their revenue estimates. Because the GFOA and others consider linking macroeconomic trends to the forecast to be best practice, the Volcker Alliance might consider using these decision rules when answering Question 4 ( Does the state reasonably support revenue growth projections at time of initial budget? ).

18 states seem comfortable providing general economic trends and then a forecast that is loosely associated with these trends. This holds for executive states and consensus forecasting states. However, the rigor around the Virginia and Florida processes is certainly appealing, and a more expanded analysis of rigor, transparency and accuracy is warranted. Political Acceptance of Southeastern States Revenue Forecasts A further issue raised in Mikesell and Ross (2014) is political acceptance surrounding the revenue forecasts in the southeastern states. The analysis looked in detail at five states, four of which had some form of consensus forecast, and one of which had an executive forecast. While it is difficult to prove a negative, there was no obvious evidence of the revenue estimate being challenged in any of the states, even those with some intense partisan political conflict such as occurred in Maryland and Virginia certainly not in the open way that is visible at the national level. The case studies of Maryland, Virginia, North Carolina, South Carolina and Georgia all included a review of the revenue estimates as they moved across through the budget approval process and all executive and legislative documents in the years studied mapped directly back to the formal revenue estimate. In the case of Maryland and Virginia, where the legislature and Governor were in different parties, the legislature did reverse a number of the Governor s proposed initiatives and newspaper articles as well as some budget documents reflect significant conflict. 6 Additionally, based on an initial review of news articles from the fiscal years 2015 to 2017, the researchers did not find any evidence of contested forecasts in any of the southeastern states and conversations with researchers on the other southeastern teams found no evidence of a contested estimate, although Alabama may bear further scrutiny. Further research on revenue 6 Add citations from Maryland 90 Day Summary and articles from VA about Medicaid expansion conflict and possibly the session summaries.

19 estimation during highly charged periods such as the Great Recession may shed further light on whether different processes are more effective at building consensus; additionally other regions of the country may have different insights. CONCLUSION The state revenue forecast is an important aspect of the budget-forming process. In a balanced budget environment, the forecast establishes a critical constraint on expenditures. Because forecasts are affected by global, national and state economic conditions, revenue collections are notoriously difficult to forecast. As Dadayan and Boyd (2014) found, states are having a hard time making accurate forecasts even after the end of the recession. 37 While accuracy is important, it is also important that the revenue forecast is transparent and politically acceptable. As Mikesell and Ross (2014) commented, revenue forecast accuracy is irrelevant if the budget process does not respect the forecast as a resource constraint. 38 In other words, a forecasting process that lacks transparency and is likely to be unaccepted by political stakeholders can be harmful to the state budget-making process, regardless of the forecast s accuracy. Using data compiled for the Truth and Integrity in Government Finance project by the Volcker Alliance, this paper examined revenue forecasting practices in the southeast and generally found that on average the forecasts were more accurate than prior research would have led us to anticipate. That being said, the 2015 and 2016 budget years were not particularly volatile. Additionally, some of the case analysis of the circumstances around some forecasts suggest that these cannot always be taken at face value: the forecasts exist in institutional as well as political frameworks and for a variety of reasons, the anticipated revenues are not always the same as the formal forecast. This finding is important when considering embarking on a larger

20 quantitative analysis assessing forecast accuracy or using forecast accuracy as an independent variable. While certainly not determinative, there was little evidence that consensus forecasts were consistently associated with improved accuracy. In terms of transparency, only two states draw a clear line between economic assumptions and the actual revenue forecast. Most states present some generic economic trends and the forecast not explicitly connected to these trends. The rigor of states that clearly explain their models is refreshing but more research would be required to validate their accuracy. Last, when examining political acceptance or the legitimacy of the forecast, there was little evidence that any of the forecasts faced a significant challenge, regardless of the methodology or the process around the forecast. Again, this analysis is by no means determinative but simply adds another observation or data point to broader theory. As observed in the literature review, it is quite possible that the revenue forecasting process is endogenous to political conflict so high conflict situations lead to processes that help resolve the conflict, whether this be consensus forecasts or other strategies.

21

22

23

24 24 P a g e

25 1 McNichol, Elizabeth Improving State Revenue Forecasting: Best Practices for a More Trusted and Reliable Revenue Estimate. Center on Budget and Policy Priorities. 2 Mikesell, John L. State Revenue Forecasting in the State of Indiana In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 3 Klay, William Earle. And Joseph A. Vonasek Consensus Forecasting for Budgeting in Theory and Practice. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 4 Sun, Jinping. Forecast Evaluation: A Case Study In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 5 Tebbs, Jeffrey M Breaking The Stalemate: A Proposal for a Consensus Revenue Forecasting Process. Connecticut Voices for Children. 6 McNichol, Elizabeth Improving State Revenue Forecasting: Best Practices for a More Trusted and Reliable Revenue Estimate. Center on Budget and Policy Priorities. 7 Clemen, Robert T Combining forecasts: A review and annotated bibliography. International Journal of Forecasting. 5(1989): Boyd, Donald J. and Lucy Dadayan State Tax Revenue Forecasting Accuracy. Rockefeller Institute. 9 Willoughby, Katherine G. and Hai Guo. The State of the Art: Revenue Forecasting in U.S. State Governments. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 10 Qiao, Yuhua. Use of Consensus Revenue Forecasting in U.S. State Governments. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 11 Klay, William Earle. And Joseph A. Vonasek Consensus Forecasting for Budgeting in Theory and Practice. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 12 Wong, John D. and Carl D. Ekstrom Consensus Revenue Estimating in State Government: A Case of What Works in Kansas. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 13 Voorhees, William R More Is Better: Consensual Forecasting and State Revenue Forecast Error. International Journal of Public Administration. 27(8-9): Boyd, Donald J. and Lucy Dadayan State Tax Revenue Forecasting Accuracy. Rockefeller Institute. 15 Elaine Deschamps The impact of institutional change on forecast accuracy: A case study of budget forecasting in Washington State. International Journal of Forecasting Buettner, Thiess and Bjoern Kauder Political biases despite external expert participation? An empirical analysis of tax revenue forecasts in Germany. Public Choice (2015): 164: Boylan, Richard T Political Distortions in State Forecasts. Public Choice. 136 (2008): Krause, George. A, David E. Lewis, James W. Douglas Politics Can Limit Policy Opportunism in Fiscal Institutions: Evidence from Official General Fund Revenue Forecasts in the American States. Journal of Policy Analysis and Management. 32(2): (2013) 19 Rose, Shanna and Daniel L. Smith. Budget Slack, Institutions, and Transparency Public Administration Review. 72(2): (2012) 20 Government Finance Officers Association (GFOA). Financial Forecasting in the Budget Preparation Process Accessed on March 7, Organisation for Economic Co-operation and Development (OECD). Best Practices for Budget Transparency OECD Journal on Budgeting. 22 International Monetary Fund (IMF). Code of Good Practices on Fiscal Transparency P a g e

26 23 McNichol, Elizabeth, Iris Lav, and Michael Leachman Better State Budget Planning Can Help Build Healthier Economies. Center on Budget and Policy Priorities Policy Futures. 24 Alt, James E., David Dreyer Lassen and Shanna Rose. The Causes of Fiscal Transparency: Evidence from the U.S. States IMF Staff Papers. 53: (2006) 25 Mikesell, John L. and Justin M. Ross State Revenue Forecasts and Political Acceptance: The Value of Consensus Forecasting in the Budget Process. Public Administration Review. 74(2): Elaine Deschamps The impact of institutional change on forecast accuracy: A case study of budget forecasting in Washington State. International Journal of Forecasting Krause, George A., David E. Lewis and James W. Douglas Political Appointments, Civil Service Systems, and Bureaucratic Competence: Organizational Balancing and Executive Branch Revenue Forecasts in the American States. American Journal of Political Science. 50(3): Buettner, Thiess and Bjoern Kauder Political biases despite external expert participation? An empirical analysis of tax revenue forecasts in Germany. Public Choice (2015): 164: Boylan, Richard T Political Distortions in State Forecasts. Public Choice. 136 (2008): Krause, George. A, David E. Lewis, James W. Douglas Politics Can Limit Policy Opportunism in Fiscal Institutions: Evidence from Official General Fund Revenue Forecasts in the American States. Journal of Policy Analysis and Management. 32(2): (2013) 31 Mikesell, John L. and Justin M. Ross State Revenue Forecasts and Political Acceptance: The Value of Consensus Forecasting in the Budget Process. Public Administration Review. 74(2): Wong, John D. and Carl D. Ekstrom Consensus Revenue Estimating in State Government: A Case of What Works in Kansas. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 33 Alt, James E., David Dreyer Lassen and Shanna Rose. The Causes of Fiscal Transparency: Evidence from the U.S. States IMF Staff Papers. 53: (2006) 34 McNichol, Elizabeth Improving State Revenue Forecasting: Best Practices for a More Trusted and Reliable Revenue Estimate. Center on Budget and Policy Priorities. 35 Willoughby, Katherine G. and Hai Guo. The State of the Art: Revenue Forecasting in U.S. State Governments. In Government Budget Forecasting: Theory and Practice. ed. Jinping Sun and Thomas D. Lynch. 142: Boca Raton, FL: CRC Press. 36 Boyd, Donald J. and Lucy Dadayan State Tax Revenue Forecasting Accuracy. Rockefeller Institute. 37 Boyd, Donald J. and Lucy Dadayan State Tax Revenue Forecasting Accuracy. Rockefeller Institute. 38 Mikesell, John L. and Justin M. Ross State Revenue Forecasts and Political Acceptance: The Value of Consensus Forecasting in the Budget Process. Public Administration Review. 74(2): P a g e