«THE EFFECT OF SCHOOL FINANCE REFORMS ON THE DISTRIBUTION OF SPENDING, ACADEMIC ACHIEVEMENT, AND ADULT OUTCOMES C. Kirabo Jackson Rucker Johnson ...»
The point estimates tell a similar story. In the 10 years after reforms, the lowest-income districts saw a $413.70 reduction in spending (p-value=0.27). Between years 5 and 10, the reduction for these districts was $547.45 (p-value=0.12). These point estimates suggest a persistent slowdown in spending growth even for the lowest-income districts. Looking at districts in the top 10 percent of income, the patterns are similar. These reforms are associated with a $743.66 reduction in the first 10 years post-reform (p-value=0.04) and a $936.64 reduction (p-value= 0.03) between years 5 and 10. Because the reductions in spending are somewhat larger for the high-income districts than the low-income districts, these reforms likely did reduce spending gaps between the top- and bottom-income districts. Our estimates suggest this was the case; relative to the spending gap in the 10 years prior to reforms, the spending gap between the top 10 and bottom 10 percent income districts was reduced by $329.75 (pvalue=0.05) in the 10 years after reforms. This represents a 27 percent reduction in the spending gap between high- and low-income districts. We conclude from this that legislative changes did have modest effects on spending inequality within states, but also tended to decrease spending overall (Hoxby, 2002).
18 c. Effects by Type of Reform Used While documenting the effects of past court ordered and legislative reforms is important from an historical point of view, it does not address the policy-relevant question of why different kinds of reforms have different effects or what kinds of reforms policy-makers should try to implement in the future. There are numerous ways that reforms can be constructed, and it can be argued that what really matters is the kind of funding formula used in a reform, rather than why or how the reform was implemented. Furthermore, as illustrated in Figure 1, there are many more funding changes that may affect the distribution of school spending that are not tied to specific legislative or court-mandated reforms. This motivates an event-study analysis of the four most commonly introduced types of reforms. Because flat grants were not introduced over time, but rather replaced with new reforms, we do not estimate the effects of introducing flat grants.
Figure 8 shows the event-study graphs for the imposition of spending limits. There is little evidence of any differential pre-existing trend in school spending for districts that imposed tax limits and those that had no change in their tax prices. It is also apparent that spending gaps across income levels were stable prior to reforms. Consistent with theoretical predictions, spending limits reduce per-pupil spending for all districts in the long run, with the most pronounced effect in the more affluent districts of a state. The fact that reductions in spending (relative to the flat trend prior to the change) grow over time is consistent with a spending limit that becomes more likely to bind as the underlying level of spending increases for all districts to the level of the limit.
One would expect the spending limit to bind first for the highest-spending districts. Then, as overall spending increases it would bind for lower-spending districts. This is the pattern observed in Figure 8. For the poorest 10 percent of districts, the spending limit reduces spending by $15.39 (p-value=0.946) in the 10 years after reforms. However, between years 10 and 20 post reforms, these low-income districts experience a $910.63 relative reduction in spending (pvalue=0.01). For higher-income districts, the reductions in spending are much more immediate.
For the most affluent 10 percent of districts, the spending limit is associated with a reduction in spending of $535.91 (p-value0.01) in the 10 years after reforms. The reduction increases to $1,494.96 (p-value0.01) between years 10 and 20. Not surprisingly, spending limits are effective at reducing spending inequality: the spending gap between the high- and low-income districts narrows by about $520.15 (p-value0.01) after five years. This is a non-trivial reduction in the spending gap, but it appears to come at the expense of slower spending growth for all 19 districts. The decreases in spending are consistent with the theoretical prediction that decreases in inverted tax prices will tend to decrease the overall level of school spending.
On the other side of the policy spectrum are policies that promote school spending by encouraging local districts to increase per-pupil spending with matching funds. We refer to these as “reward for effort” policies. Figure 9 provides the event study for this kind of reform. Unlike other kinds of reforms, there is clear evidence of a downward trend in per-pupil spending for those states that implemented local equalizing policies. This is consistent with the notion that the kinds of policies states employ are not random and that one must be careful to consider preexisting trends when analyzing the effects of such policies. Despite the existence of a negative trend, there is clear evidence of a structural break at the time of passage of reforms. While spending is clearly declining in all districts in the pre-reform years (seven out of nine of the changes are negative realizations for the lowest-income districts), there is an upward trend that lasts about five years (four out of five first post-reform realizations are positive for the lowestincome districts). The fact that this negative to positive change is experienced for all districts suggests that this is not merely a statistical artifact. After this five-year period of increased spending, however, spending reverts to the pre-existing downward trend.
Because of the pre-existing negative trend, estimating the effects on spending levels with a DiD model is unwise because the common trends assumption is clearly violated for spending levels. However, the common trends assumption may be valid for spending growth. If so, one can estimate credible effects on spending growth by applying equation , on the one-year change in spending rather than the level of spending. This allows for the estimation of the effect of reward for effort reforms on spending growth because it takes into account differences in spending growth between reform and non-reform districts.
The lower panel of Figure 9 shows the event study for changes in school spending. It is clear that while the common trends assumption was violated for levels, it appears to be satisfied for year-to-year changes in spending. The figure shows that during the first five years after the introduction of a reward for effort reform, all districts experienced increased spending growth relative to the previous 10 or five years; low-income districts experienced an increase in the year-to-year increase in spending of $131.13 (p-value=0.01), and high-income districts experienced an increase in the year–to-year increase in spending of $126.10 (p-value=0.03).
Consistent with a reversion to the pre-reform growth rate after about five years, there is not a statistically significant difference between the growth rates for post-reform years 5 through 10 20 and the pre-reform years (i.e., both yield p-values above 0.1). However, there is evidence suggestive of increased spending growth for the lowest-income districts in the long run such that during post-reform years 10 through 20, average annual spending changes were $175.88 more (p-value=0.08) than during the pre-reform years. This is consistent with the analysis in levels that reveals that reward-for-effort plans reduce the spending gap between low- and high-income districts in the long run by $295.83 (p-value=0.11). Overall, the patterns show an increase in spending and spending growth in the short run (lasting about five years after reforms) for all districts, with a possible permanent increase in spending growth for the poorest districts. Results suggest that these policies increase the growth of spending (particularly for low-income districts) and reduce spending gaps between high- and low-income districts by about 12 percent.
The last two kinds of reforms are foundation plans and equalizing plans (Figure 10). Both kinds of plans generally adjust state spending such that districts with low tax bases (rather than low income) receive additional funds from the state. For both these types of reforms, spending behaviors were erratic more than five years prior. Accordingly, the figures only plot the four years before reforms, and all statistical inferences are relative to the five years prior to reforms (when behaviors were more stable). The figures reveal that for both kinds of plans, low-income and high-income districts were on similar trajectories (and as were districts in other states) for the five years prior to reforms.
After reforms, both kinds of plans increased spending for the low-income districts and had small effects for the high-income districts. Foundation plans increased spending for all districts below the 90th percentile in median income. For the lowest-income districts, equalizing plans increased per-pupil spending (relative to the four years prior to reforms) by $464.03 (pvalue=0.06) in the 10 years post-reform. However, for high-income districts there was a slight decrease of $84.47 (p-value=0.74). The gap in spending associated with these reforms between the low- and high-income districts was reduced by $548.21 (p-value0.001) in the 10 years after reforms. Equalization plans had a very similar effect: there were increases for low-income districts ($529.07) and small decreases for high-income districts ($47.10) such that the gap in spending was reduced by $576.18 (p-value=0.03). In sum, both equalizing plans and foundation plans reduced spending gaps between high- and low-income districts by about one-third, and appear to have done so primarily by increased per-pupil spending for the lowest income districts.
The figures reveal that, by and large, school finance reforms achieve the stated objective of reducing inequalities in school spending between low- and high-income districts and increase the 21 level of per-pupil spending in poor communities. Both equalization plans and foundation plans are effective at reducing spending gaps between low- and high-income areas. The results also indicate that plans that aim to increase equality by reducing spending for the highest-income districts achieve this objective, but with the unintended impact of also reducing spending in lowincome districts in the long run. In contrast, plans that promote greater education spending through matching tend to have a positive effect on the growth of school spending for all districts, with particularly large effects for low-income districts. Having established to what extent and how SFRs change the distribution of school spending, the remaining question is how changes in school spending caused by these reforms affect the educational and adult economic outcomes of children. This is the topic of Part Two.
PART TWO: EFFECTS OF SCHOOL SPENDING ON LONG-RUN OUTCOMES
IV. Description of the Longer-Run Outcome Data The primary micro dataset utilized to analyze the effects of reform-induced changes in school spending on long-run outcomes is the restricted, confidential geocoded version of the PSID (1968–2011) with identifiers at the level of the neighborhood blocks in which children grew up.14 We link our district-level data on school spending and the timing of reforms to the nationally-representative sample of children born between 1955 and 1985 from the PSID.
Following Johnson (2012), we then merge neighborhood and school characteristics, as well as information on other key policy changes (e.g., the timing of school desegregation, hospital desegregation, rollout of “War on Poverty” initiatives, and expansion of safety net programs) from multiple data sources on the conditions that prevailed when these children were growing up, allowing for a rich set of control variables.15
14 The PSID began interviewing a national probability sample of families in 1968. These families were reinterviewed each year through 1997, when interviewing became biennial. All persons in PSID families in 1968 have the PSID “gene,” which means that they are followed in subsequent waves. When children with the “gene” become adults and leave their parents’ homes, they become their own PSID “family unit” and are interviewed in each wave.
The original geographic cluster design of the PSID enables comparisons in adulthood of childhood neighbors who have been followed over the life course.
15 The data we use include measures from 1968–1988 Office of Civil Rights (OCR) data; 1960, 1970, 1980, and 1990 Census data; 1962–1999 Census of Governments (COG) data; Common Core Data (CCD) compiled by the National Center for Education Statistics; Regional Economic Information System (REIS) data; a comprehensive case inventory of court litigation regarding school desegregation over the 1955–1990 period (American Communities Project); and the American Hospital Association’s Annual Survey of Hospitals (1946–1990) and the
The sample consists of PSID sample members born between 1955 and 1985 who have been followed into adulthood; these individuals were between the ages of 26 and 56 in 2011. We include all information on them for each wave, 1968 to 2011.16 We include both the Survey Research Center (SRC) component and the Survey of Economic Opportunity (SEO) component, commonly known as the “poverty sample,” of the PSID sample. Due to the oversampling of African-American and low-income families, 59 percent of the sample members were poor as children (N=15,353 individuals; 9,035 poor children; 6,318 non-poor children). Sixty-six percent of the PSID individuals born between 1955 and 1985 and followed into adulthood grew up in a school district that was subject to a court-mandated school finance reform sometime between 1972 and 2000, with the timing of the court order not necessarily occurring during their schoolage years. Eighty-eight percent of the PSID individuals born between 1955 and 1985 who were poor as children and followed into adulthood grew up in a school district that was subject to a court-mandated school finance reform sometime between 1972 and 2000. Given the patterns in Figure 1, the share of individuals exposed to school finance reforms during childhood increases significantly with birth year over the 1955–1985 birth cohorts analyzed in the PSID sample.