Do Students Respond to Sticker-Price Reductions? Evidence From the North Carolina Promise

Abstract

Reduced or free tuition policies are an increasingly popular policy tool for states looking to increase postsecondary enrollment. However, sticker-price reductions are relatively rare in the four-year sector, so we do not know how they will impact student enrollment decisions. Using IPEDS data and a synthetic control approach, we investigate the impact of the North Carolina Promise on enrollment and persistence. NC Promise is a state-level policy that reduced the price of tuition for students enrolled at three public universities in North Carolina starting in fall 2018. We find that NC Promise did not appear to increase enrollment among first-year students. However, it attracted more transfer students and increased enrollment of Hispanic students at one of the institutions.

Keywords

colleges educational policy higher education policy postsecondary education policy analysis quasi-experimental analysis secondary data analysis transfer students tuition Promise Program college enrollment sticker price North Carolina

The rising cost of college has been prohibitive for many lower-income families, who make up a shrinking proportion of college students in the United States (Bailey & Dynarski, 2011). For this reason, efforts to make college more affordable—from grants to loans to free tuition—are among the most common ways of trying to make college a realistic option for more students. Although these efforts are generally successful at making students more likely to enroll, it is likely that who they affect varies with the form of the cost-reducing policy. For example, students generally do not know the amount of scholarship or grant aid they will receive until after they have been admitted to college, perhaps excluding students who do not apply to colleges because of uncertainty about costs, and loans may exclude certain populations because of differing rates of risk-aversion associated with taking on debt (Boatman et al., 2017). Although similar in concept to policies that make tuition free, less is known about the effects of policies that work to reduce published tuition rates.

The state of North Carolina introduced such a program—the North Carolina Promise Tuition Plan (NC Promise)—in 2018. NC Promise is an over $60-million recurring tuition subsidy that supports all undergraduates who enroll at three of the state’s four-year degree-granting institutions.¹ The policy reduces tuition for North Carolina residents and out-of-state students at all three universities to $500 and $2,500 per semester, respectively. In this respect, NC Promise is relatively unique among cost-reduction programs because it explicitly reduces tuition for all students rather than offering aid to a targeted population. While reducing the costs for all students, the size of the tuition reduction—nearly $2,500 per year at the three participating institutions—effectively made tuition free for many students who qualified for federal financial aid. A key feature of the policy is that it offers clear and reliable pricing to all students well before they have to make college application and enrollment decisions. Because of its design, NC Promise is most likely to be most effective at encouraging enrollment among populations of students who are sensitive to the sticker price of college.

Given the weight students and families appear to put on sticker prices, even when net-price calculators are available (Grodsky & Jones, 2007; Levine, 2014; Levine et al., 2023), this distinction is potentially significant. We are unaware of any prior empirical work outside the small literature in the community college context (Acton, 2021; Denning, 2017) and one study in the private university context (Ward & Corral, 2022) that rigorously examines tuition-reduction policies that affect sticker prices like NC Promise. No prior literature examines the effects of tuition reduction policies at public universities.

Because the lower tuition may specifically draw students to enroll at one of the three Promise institutions, we look at the effects of NC Promise by examining how it changes the composition of the student body enrolled at those universities relative to universities that were not affected by the policy. Specifically, to better understand the enrollment response caused by the NC Promise tuition discount and how this response may vary across student populations, we address the following three research questions:

1. How did NC Promise affect the number of:

a. first-year enrollees at Promise institutions?

b. transfer enrollees at Promise institutions?

2. Do these changes in new student enrollment vary by racial-ethnic or economic background?

Finally, if NC Promise induces students to enroll in universities other than the ones they would choose in the absence of the policy, it is possible these universities might provide a different level of academic or social support for those students, so we ask:

3. Did NC Promise affect the likelihood of students re-enrolling at Promise institutions?

We find that the NC Promise did not appear to have a sustained effect on first-year student enrollment. There was a slight initial increase in first-year enrollment at all three institutions, but these changes did not persist in subsequent years relative to changes at comparison institutions. However, we found that transfer-student enrollment increased at two of the Promise institutions. All institutions saw increases in enrollment by White students, and at least one saw a sizeable increase in Hispanic student enrollment. There was no evidence of changes in enrollment by Pell-eligible students. Finally, overall retention rates remained unchanged, suggesting that any students who enrolled at Promise institutions after NC Promise began who might not have enrolled without the policy were just as likely to succeed as students the schools traditionally enrolled.

The Effects of Tuition Decreases

Although colleges notoriously increase tuition regularly, it is less common for tuition rates to decline. This distinction matters because individuals view cost increases as losses, and—because of loss aversion—respond much more dramatically to losses than to gains (Thaler, 1980). This means that if a tuition increase causes enrollment to go down, a symmetric decrease in tuition may not result in enrollment increasing by a corresponding amount. We know that students are sensitive to tuition increases, applying to colleges at lower rates even if the net cost to those students would not change (Levine et al., 2023).

There is not much evidence about how students respond to tuition decreases. Evidence from the community college context suggests that individual students are more likely to enroll in community colleges after tuition decreases and that students are more likely to ultimately transfer to four-year campuses after tuition decreases (Acton, 2021; Denning, 2017). However, community college students generally come from different backgrounds than four-year college students. The only published evidence of the effects of tuition reductions in the four-year context comes from a set of tuition resets at private four-year colleges. Here, total enrollment increased initially by about 7–12 percent at four-year colleges after tuition decreases, but those increases did not appear sustained for more than the first few years after the tuition reduction (Ward & Corral, 2022). We contribute to this literature by examining how tuition reductions may influence enrollment in the four-year public university context, where the vast majority of four-year college students enroll.

Conceptual Grounding: College Prices and Enrollment Decisions

Cost Perceptions

At its most basic, a tuition reduction makes college enrollment less expensive. Under the human capital model, which says that students weigh the costs and benefits of pursuing additional education, cheaper tuition should make more people want to pursue a college degree (e.g., Becker, 1962, 1993). Thus, more people should want to enroll in the Promise institutions because of the tuition reduction.

However, as Perna’s conceptual framework for college choice highlights, the way students assess the costs and benefits of college enrollment can vary by their social background and context (2006). For example, the change in tuition may not make a difference in the college enrollment choices of students from higher-earning families who may be able to pay for college regardless of the tuition rate. We might then expect that the tuition change would make the most difference in the enrollment choices of students from the lowest-earning families. However, need-based financial aid covers the differences between the amount a student is expected to be able to afford to pay for college and the actual cost. Thus, for some lower-income students, the out-of-pocket expenses for a student attending a Promise institution could remain unchanged because financial aid would cover differences in tuition anyway.

By lowering tuition, NC Promise would have reduced the amount of money students would have needed to borrow to enroll at one of the Promise institutions. This means that students who may be loan- or risk-averse may be more likely to enroll in those institutions because they would have to take out less in loans to attend. Such loan aversion for education seems particularly prominent among Hispanic populations, suggesting Hispanic students could be particularly sensitive to the NC Promise subsidy or, more broadly, there is reason to believe students may make different enrollment choices based on their racial or ethnic background (Boatman et al., 2017). There is evidence that students who enrolled in Promise institutions after NC Promise began took out less in total loans than before the policy, though these changes were primarily among students from households earning over $50,000 a year (Zahran & Klasik, 2025).

Alongside these potential changes to the racial/ethnic makeup of a university, we note that more wide-ranging literature and theory highlights that students tend to enroll in colleges close to where they live, particularly students who attend non-elite colleges like the NC Promise institutions (e.g., Klasik et al., 2025; Skinner, 2019). This means that changes to a college’s demographics resulting from NC Promise may be limited by the diversity of the populations surrounding each campus.

Finally, students who could potentially transfer to an NC Promise institution from a community college may also make different enrollment choices as a result of the policy. Many students start their postsecondary education at a community college because of the opportunity to earn credits toward a bachelor’s degree without the expense of spending four years at a four-year university (Jenkins & Fink, 2015). However, at the time of transfer, many community college students continue to express concerns about the costs of attending a four-year university (Cooper et al., 2020). Thus, lower tuition at the Promise institutions may create a more affordable transfer pathway for community college students aspiring to four-year degrees.

Policy Simplicity

NC Promise was also likely to change students’ enrollment choices because it was simple and well publicized among relevant populations. These features are important because the effects of college affordability programs often vary according to their design (e.g., S. Dynarski & Scott-Clayton, 2013; Page & Scott-Clayton, 2016). In general, financial aid programs are more likely to affect enrollment choices when they have easily understood eligibility rules and application procedures. Simple, uncomplicated programs like NC Promise appear particularly valuable among students who sometimes struggle to find reliable sources of information about the college application and enrollment process (S. Dynarski et al., 2022). When financial aid programs are complex, there is generally no association between aid receipt and college enrollment (Carruthers & Welch, 2019; R. B. Rubin, 2011).

Scholarship programs similar to NC Promise, with simple application procedures and few eligibility requirements, like the DC Tuition Assistance Grant Program (DC TAG), appear to increase college enrollment (Kane, 2007). The DC TAG program subsidized tuition for students from the District of Columbia to in-state tuition levels if they attended any public institution in the United States. As a result, the total number of DC residents enrolling at eligible public four-year colleges nearly doubled (Kane, 2007).

Similar results are reflected in the literature examining promise programs, which typically offer early guarantees of financial support to certain populations or those who attend certain institutions. Promise programs, such as the New Haven Promise and the Kalamazoo Promise, with simple qualifications based on where students live, have been associated with enrollment increases ranging from 3 to 8 percentage points per $1,000 of aid (Bartik et al., 2021; Carruthers & Fox, 2016; Gonzalez et al., 2014).

College Affordability and Retention

While increased college enrollment is a key measure of success for promise programs, students must remain enrolled through graduation to realize the benefits of a college degree. The challenge of helping students persist through earning a degree is particularly acute in North Carolina, which has set a specific state goal of increasing the number of North Carolinians with a college degree. Whether students remain enrolled in college is influenced by the interaction of multiple complex factors, including financial aid (Bettinger, 2004), institutional characteristics and supports (Millea et al., 2018), and first-year experiences (e.g., Noble et al., 2007). College costs are likely to be a particular concern in the decision to remain enrolled for the same set of price-sensitive students who may be likely to enroll in a Promise institution because of the cost savings NC Promise offers.

Several studies point to the positive effects of financial aid on retention (Millea et al., 2018). Financial aid can increase retention through several mechanisms, such as alleviating students’ worries about paying for college, freeing up their time from employment to earn supplemental income to help pay for college, and reducing dropout rates (Aina et al., 2022; Bettinger, 2004).

Policy Background and Participating Campuses

In response to concerns about declining enrollment at several of the regional UNC campuses, the North Carolina Legislature passed the Access to Affordable College Education Act (Senate Bill 873) in 2016. The most prominent feature of Senate Bill (SB) 873 was NC Promise, which originally intended to reduce tuition at five campuses, including three of the state’s five public HBCUs, to $500 per semester for in-state students and $2,500 for out-of-state students. However, the North Carolina NAACP and others expressed concern that the focus on three Historically Black Colleges and Universities sought to “rewrite their identities and deprive them of tuition dollars” (Seltzer, 2016). This was because, although SB 873 proposed to cover the cost difference between the new and the old tuition, opponents were concerned about the potential fiscal cliff if the state discontinued funding (Villemain, 2021). Ultimately, two HBCUs opted out before the bill was finally passed.

In its ultimate form, NC Promise provides discounted tuition for all students attending one of the NC Promise four-year institutions in-person or online, whether they are newly enrolled first-time or transfer students, or those already enrolled, since the fall semester of 2018. It applies only in fall and spring semesters—students who enroll in summer classes pay unsubsidized tuition. There was no change in the availability of other forms of financial aid, other than their calculation based on the subsidized tuition rate. Notably, students do not have to apply for NC Promise or complete additional paperwork like other grants or scholarships.

As part of the rollout of the policy, the UNC system launched an advertising campaign beginning in 2017 to spread awareness of the program. The advertising campaign included statewide radio and television commercials, high school and community college outreach, and informational mailers. Thus, it was highly visible across the state to potential students and their families.

NC Promise was first implemented at Elizabeth City State University (hereafter Elizabeth City or ECSU), the University of North Carolina at Pembroke (hereafter UNC Pembroke or UNCP), and Western Carolina University (hereafter Western Carolina or WCU) in fall 2018. These three universities are geographically spread across the state so that most North Carolina residents live within 150 miles of at least one of them. Elizabeth City, located in northeastern North Carolina, is a small liberal arts HBCU emphasizing natural and aviation sciences. It has moderately selective admissions (77 percent acceptance rate) and frequently partners with early college high schools. As a result of NC Promise, the yearly tuition for in-state students was reduced by $1,856.

UNC Pembroke is located in southeastern North Carolina, the cultural center of North Carolina’s largest American Indian tribe, the Lumbee Indians. It enrolls a high number of Native American students and qualifies as a Native American–Serving Nontribal Institution. UNCP’s proximity to Fort Bragg (formerly Fort Liberty) also makes it among the most military-friendly of the three NC Promise universities. It is also the least selective among them, with an 80.6 percent acceptance rate in 2017. NC Promise reduced tuition for in-state students by $2,602 in 2018.

Western Carolina, located in the mountains of western North Carolina, has the least diverse student body of the three Promise universities but shares a relatively high admissions rate. WCU serves a primarily White population and offers a comprehensive curriculum with large programs in business, nursing, and criminal justice. In 2018, NC Promise reduced tuition for in-state students by $2,971.

Data

Our study utilizes institution-level data from the Integrated Postsecondary Education Data System (IPEDS). In addition to information on basic institutional characteristics, we focus on IPEDS fall enrollment data at four-year colleges and universities from the fall 2010 to fall 2019 school years, providing seven fall semesters of pre-NC Promise data and two fall semesters of data following its implementation in 2018. Although we could extend the panel by two additional years, we limit our main analysis to the two fall semesters after implementation because the COVID-19 pandemic affected college enrollment unexpectedly and unevenly across institutions, states, and regions starting in the spring of 2020.

Outcomes

We focus on three primary outcomes: the total number of first-time, first-year students enrolled at a college or university; the total number of new transfer students at a college or university; and the total number of students who remain enrolled from year to year. For the purposes of our analysis, we focus on enrollments by degree-seeking undergraduate students, including both first-time students and transfer students. We use fall enrollment counts both because most students enroll in the fall term and because the alternative, twelve-month enrollment counts, include summer enrollments that should not be affected by NC Promise. For comparability across institutions, we convert the enrollment totals to a log scale to express the NC Promise effect as a percent change in enrollment from the prior year.

We also consider enrollment outcomes by race and ethnicity and Pell eligibility. Because of the small sample sizes of some of these groups at the Promise institutions, which result in noisy enrollment trends over time, we combine first-year students and transfer students for these analyses. For similar reasons, we also focus on the analysis of the three largest race/ethnicity categories at the Promise institutions: White, Black, and Hispanic students. This results in an interpretation of these outcomes as changes in the demographics of the new student population.

For our retention outcome, we use the IPEDS fall-to-fall full-time retention rate, which is calculated as the percentage of the total number of returning full-time students from one fall semester to the next, excluding graduates and other special cases where students are not expected to re-enroll.

Matching Variables

These variables include characterizations of the student body, including the percentage of campus that are female, adult (over the age of 24), Black, Hispanic, and White and receive Pell, as well as institutional characteristics like total undergraduate enrollment; number of applicants; retention and graduation rate; student-to-faculty ratio; endowment per FTE (full-time equivalent) student; and HBCU status.

Method

Two features of NC Promise drive our analytic approach. First, it changed tuition for any student who enrolled at the Promise institutions regardless of whether they were in-state or out-of-state. Second, it lowered tuition at three specific, regional public university campuses. A student-level analysis could look at changes in the likelihood that any individual student enrolls at one of the Promise institutions, but the population for this analysis is not obvious. For example, a student from Hawaii could benefit from the NC Promise policy, but the low baseline likelihood of their enrollment makes the benefit of including Hawaiian students in the analysis unclear. The regional nature of the Promise institutions also makes a study focused only on North Carolina students similarly fraught. Thus, because the way NC Promise creates an incentive for students to enroll at one of the three Promise institutions means that it makes the most sense to examine how enrollments change at those specific campuses. Institution-level data makes it possible to easily look at different types of enrollment (e.g., first-year, transfer) as well as retention outcomes. This institution-level approach is in line with other studies examining tuition decreases that incentivized enrollment at specific campuses (e.g., Ward & Corral, 2022).

There are two main obstacles to studying aggregate enrollment outcomes at the three different universities targeted by NC Promise. First, the small number of schools affected by the policy leaves very few observations on which to base an analysis. Second, although scholars have historically used control units from neighboring states (e.g., Cornwell et al., 2006) to study the effect of state policies, it is not always clear that this choice is ideal or appropriate (McClelland & Gault, 2017). However, these problems are not uncommon, and we address them by using a synthetic control approach. Synthetic control methods are designed precisely for cases like ours, where there are few “treated” units and comparison units are unclear (Abadie & Gardeazabal, 2003; Abadie et al., 2010).

Synthetic control approaches identify a set of comparison units such that a weighted combination of those units matches, as best as possible, the target-treated unit in terms of both the pretreatment trend in the outcome of interest and relevant covariates. When certain assumptions are met, the difference between the posttreatment outcome of the treated unit, and the weighted outcome of the “synthetic control,” gives a causal estimate of the treatment effect. Such methods have been used in education research to study the effect of Hurricane Katrina and subsequent charter school reforms (Barnes et al., 2022), the effect of the promise of a higher education scholarship on school district enrollments and graduation rates (Bifulco et al., 2017), the effect of universal pre-kindergarten (Bassok et al., 2016; Fitzpatrick, 2008), the effect of statewide SAT and ACT requirements (Klasik, 2013), and direct college admissions policies (Odle & Delaney, 2022), among others.

There are many contextual requirements for a valid synthetic control study, particularly concerning threats to validity (Abadie, 2021). First, synthetic control methods are generally not strong when it comes to determining small treatment-effect sizes. Thus, in the case of NC Promise, we will only detect a statistically significant effect on enrollment if it is relatively large and persists over time. Second, synthetic control methods may struggle to detect a treatment effect if there are responses in the outcome that anticipate the treatment effect. In the case of NC Promise, this may look like students enrolling in one of the Promise institutions prior to the policy beginning to take advantage of the eventual decrease in tuition. Any anticipatory behaviors should be apparent in pretreatment trends, though we expect them to be minimal given the relative ease of transferring into these institutions.

We also do not want other policies systematically affecting the outcomes at treated institutions at the same time as NC Promise. There are no other promise programs that affect the Promise institutions,² and we can find no other state policy that would affect our results. It is more challenging to verify the lack of confounding treatments among our candidate control units, although we are certainly not aware of any wide-reaching policies, and the effect of any smaller-scale policies would be dampened as institutions are combined to form the synthetic controls.

Abadie (2021) also stipulates two data requirements for a valid synthetic control approach. First, there must be aggregate data on predictors and outcomes. While this typically means state-level or regional data, synthetic control can also be used, like in this case, for institution-level data (e.g., Odle, 2022; P. G. Rubin & Canché, 2019). Second, synthetic controls require sufficient data in the pretreatment and posttreatment years. While there is no standard number for this requirement, Abadie (2021) argues that the pretreatment data periods must be sufficient to reveal trends before implementation to increase confidence that a change after implementation is spurious or a true result of the treatment. In our case, we use 7 years of pretreatment data (fall 2010 to fall 2017) and 2 years of posttreatment data (fall 2018 and 2019). Two years of posttreatment data may be short, but it appears sufficient to capture short-term enrollment responses to the NC Promise policy. It is also analytically necessary to only use those years given idiosyncratic enrollment changes that resulted from the COVID-19 pandemic.

Constructing the Synthetic Control Group and Estimating the Effect of NC Promise

Synthetic control methods rest on the idea that the combination of many untreated control units serves as a better comparison case than any single control unit (Abadie, 2021). The main mechanism of this approach generates nonnegative weights for potential control units such that the weights sum to one and they jointly minimize (1) the average difference in a vector of covariates X between the treated and untreated units and (2) the average pretreatment difference in the outcome Y between treated and untreated units. Note that this dual minimization works to more closely match covariates that are more strongly related to the outcome Y (Abadie, 2021). These covariates are not part of the estimation of the treatment effect, which, given the set of weights w_j, is calculated by

\hat{A T E} = \frac{\sum_{t = 1}^{T} (Y_{1 t} - \sum_{j = 2}^{J + 1} w_{j} Y_{j t})}{T}

where t = 1,. . .,T are each of the posttreatment time periods (in this case, years), j = 2,. . .J index the synthetic control units, and j = 1 is the treated unit. Plainly, this gives the average difference over the posttreatment period in the outcome between the treated unit and the weighted average of the control units.

Synthetic control matches are strongest when they are created from a large pool of control institutions that are likely to be similar to the treated institutions in measurable and unmeasurable ways (Abadie, 2021). As with difference-in-differences approaches, it is common to generate the pool of candidate comparison units using geographic regions because nearby colleges may follow similar enrollment and demographic trends within similar political environments (e.g., Odle, 2022). We avoided this construction given the possible spillover effects of NC Promise—the lower tuition at NC Promise institutions may have drawn enrollment away from similar nearby schools (in fact, a secondary goal of the policy was to attract more out-of-state students to the Promise institutions), making schools in the same geographic area poor comparisons. We do, however, consider geographically determined comparison institutions as a robustness check.

Instead, our preferred pool of candidate comparison institutions is drawn from colleges and universities that had a shared Carnegie classification. The Carnegie Foundation categorizes colleges according to several different classification schemes, which serve as useful heuristics for grouping similar institutions in research (e.g., Crisp et al., 2019; Engberg, 2012; Holzman et al., 2020; González Canché, 2018). We focus on the Carnegie Classification based on the undergraduate profile.³ This classification scheme groups colleges according to institutional level, admissions selectivity, the percentage of students who are full-time students, and the percentage of students who transfer to the institution (Undergraduate Profile Classification, n.d.). Using this classification scheme to generate candidate control institutions means that potential control institutions are already well matched with treated institutions in terms of preexisting student enrollment patterns.

Our analytic sample consists of the three NC Promise institutions and a synthetic control group for each institution for each outcome of interest. Our population of potential comparison institutions for the synthetic control analysis is the 558 public, four-year institutions in the United States contained in the IPEDS data. We estimate synthetic control weights separately for each treated unit, given that the three Promise institutions are not comparable in size, mission, or demographics. Under the Carnegie Undergraduate Profile Classification, ECSU and UNCP are defined as four-year, full-time, inclusive, and higher transfer-in institutions. WCU is classified as a four-year, full-time, selective, higher transfer-in institution.⁴ Given this classification, ECSU and UNCP have the same classification and a donor pool of 69 institutions. We do not allow ECSU to serve as a control for UNCP and vice versa. WCU has a donor pool of 135 institutions.⁵

We use the wide variety of covariates described previously to generate the synthetic control weights. We selected these variables to help us compare institutions that are similar to the Promise institutions both in terms of the types of students who apply to and enroll in them, as well as the relative availability of resources.

Table 1 provides the analytic sample description for the primary control group classification and each NC Promise institution. This table uses synthetic control weights to generate the summary statistics for the synthetic institution for our first two outcomes in the prepolicy period, showing how well the algorithm was able to match the control group to each NC Promise institution. For ECSU, the synthetic institution for the first-year outcome has a slightly larger average fall enrollment and is 100% comprised of HBCUs. For UNCP, the synthetic institution for the first-year outcome is also slightly larger in average fall enrollment but is roughly similar for the set of other institutional characteristics. For WCU, the synthetic institution created for the first-year outcome has a larger average fall enrollment and is roughly similar on the other institutional variables. Appendix Table A1 gives an example of the institutions and corresponding weights for synthetic controls for each of the three Promise institutions for the first-year student enrollment outcome.

Table 1.

Descriptive Statistics on NC Promise Institutions and Synthetic Control Matches Prior to the Start of NC Promise (2010–2017)

		Outcome-Specific Synthetic Control			Outcome-Specific Synthetic Control			Outcome-Specific Synthetic Control
	ECSU	First-Year Enrollment	Transfer Enrollment	UNCP	First-Year Enrollment	TransferEnrollment	WCU	First-Year Enrollment	Transfer Enrollment
Enrollment Composition (%)
First-year	17.2%	16.6%	19.5%	19.9%	17.4%	16.3%	20.3%	16.8%	16.7%
Transfer	7.2%	8.4%	10.5%	10.1%	11.0%	11.5%	9.8%	11.3%	10.2%
Full-time	92.6%	81.6%	86.7%	82.6%	85.4%	88.6%	86.2%	86.0%	86.5%
Women	58.7%	67.4%	55.2%	60.6%	57.8%	60.7%	53.8%	50.9%	53.0%
Adults (>24)	18.9%	34.8%	41.8%	27.9%	25.0%	21.9%	19.1%	16.5%	17.3%
Race/Ethnicity % of New First-Time & Transfer Undergraduates
Black or African American	69.4%	81.4%	9.6%	34.4%	28.1%	12.3%	5.8%	9.6%	7.6%
Hispanic	2.0%	1.7%	1.6%	5.3%	28.1%	48.5%	5.5%	21.2%	20.2%
White	15.8%	5.7%	79.9%	39.1%	25.8%	22.9%	79.8%	51.9%	51.5%
Institutional Characteristics
Student-to-faculty ratio	14.62	17.77	14.94	15.25	18.50	20.45	16.00	18.30	19.68
Student retention rate	73.25	57.61	63.0	67.25	69.16	72.89	77	81.97	77.24
Total annual graduates	568	508	308	1,005	1,252	1,052	1,451	1,800	1,813
Endowment per FTE	$2,962	$2,910	$17,220	$3,052	$3,423	$2,927	$6,272	$6,804	$4,524
HBCU	100%	100%	42%	0%	50%	21%	0%	0%	0%
Undergraduates on Pell (%)	73.03%	72.45%	58.27%	56.99%	55.73%	59.66%	39.01%	39.17%	39.66%

Notes. Table presents summary statistics for each NC Promise institution and the outcome-specific synthetic control institution drawn from the pool of colleges that share a Carnegie Undergraduate Classification with the Promise institution. Enrollment counts are of degree-seeking undergraduates reported in the IPEDS fall enrollment total. The synthetic comparison for ECSU and UNCP draws from data on 69 peer institutions. The synthetic comparison for WCU draws from data on 135 peer institutions.

Hypothesis Testing

A challenge with the synthetic control method is that standard errors are not useful because there is a single treated unit, making traditional inference-based hypothesis testing unviable (Abadie et al., 2010). To understand whether changes in enrollment detected by the synthetic control model are significant, we can construct p-values using placebo tests among the control institutions as defined by Abadie (2021). This process involves running a synthetic control analysis on each candidate comparison institution and estimating “treatment” effects corresponding with the start of NC Promise in universities where NC Promise should not have had an impact. For each of these placebo tests, we calculate the root-mean-squared prediction error (RMSPE), defined as the square root of the average squared difference between the placebo unit and synthetic control outcomes trends separately for the periods before and after the start of NC Promise. A small RMSPE before NC Promise indicates a strong synthetic control match, and a small RMSPE after NC Promise began indicates little to no enrollment changes in the placebo case.

To calculate the analog of a traditional p-value, we first eliminate any placebo case with a pre-Promise RMSPE more than five times the RMSPE of the target NC Promise institution. This eliminates cases with particularly poor synthetic matches. We then calculate the ratio of the post- to pre-Promise RMSPE for the treatment and remaining placebo cases. Here, a high ratio indicates considerable variation in enrollment after NC Promise began relative to enrollment variation before NC Promise. Thus, the reported p-values in our study are the percentage of placebo cases whose post-/pre-RMSPE ratio is larger than the one we calculated for the actual treated unit in each analysis. We consider results significant if they are extreme within this distribution (Abadie, 2021). Note, specifically, that this p-value is in part a function of the number of units in the synthetic control—low numbers of control units will mechanically generate p-values greater than conventional significance thresholds even if the treated case is the most extreme among the placebo permutations (for example, a treated case that is more extreme than eight placebo comparisons results in a p-value of p = .125). As a result, we report the rank of our treated estimate among all placebo permutations (a two-sided comparison) as well as its rank among results in the estimated direction (a one-sided comparison) to be clear about how extreme the treatment estimate is relative to the placebo cases.

Power

For the same reasons traditional hypothesis testing does not work in a synthetic control approach, it is not possible to calculate standard measures of statistical power. However, we use our placebo tests to perform a novel post hoc analysis to estimate rough estimates of the minimum detectable effect (MDE) we could find for each Promise institution and outcome. The details of this process, and each MDE estimate, are given in Appendix B. In short, we confirm earlier caveats that synthetic controls are better positioned to detect larger effects (Abadie, 2021). For reasons related to the quality of synthetic matches and outcome variation among the comparison units, our analyses would have difficulty detecting enrollment changes lower than about 8% with 90% confidence. As a result, we are generally confident in the direction of our significant results, particularly when they are consistent across comparison samples, as well as that our consistent results that are far from significant likely represent null effects. We have less confidence in the specific point estimates of any given model.

Limitations

In addition to these concerns related to power, there are several important caveats to this work. First, although synthetic control approaches are ideal for the analytic scenario we face, the estimates are still limited by how well the algorithm can construct synthetic controls from the pool of candidate control units. Consequently, results that do not appear significant may result more from an unideal synthetic match than from an underlying null effect. As a result, we try to focus our conclusions on the significant results we do find, except in the case of first-year student enrollment, which we discuss because the results appear relatively consistent across candidate comparison groups.

We also note that although any changes in enrollment that we describe illustrate how NC Promise affected students’ decisions to enroll in the Promise institutions, we are not able to distinguish whether these enrollment changes result from students enrolling in one of the campuses who might not otherwise have enrolled at all, or whether they are students who would have enrolled at different institutions if the policy had not been in place.

Additionally, our 2 years of posttreatment outcomes do not support an analysis of sustained changes in enrollment that resulted from NC Promise. The inclusion of later years adds a substantial amount of noise to our estimates due to the effects of the COVID-19 pandemic. However, the general pattern of results matches what we report later.⁶ We note, however, that efforts to promote the NC Promise policy declined over time, so while it is important that we document the initial changes the policy caused, the policy effects may have dropped off after the initial promotional campaign ended.

As noted previously, we are also limited in our ability to track outcomes by race and economic background by the first-year or transfer-student entry points. Still, by combining these two categories to examine student subgroups, we are able to offer an analysis of what changes in “new” student enrollment look like according to student background.

Results

In our main results, we find that the NC Promise did not have an overall sustained effect on first-year student enrollment. There was a slight initial increase in first-year enrollment at all three institutions, but these changes did not persist. However, NC Promise did result in an increased number of transfer students, effectively increasing total enrollment at Promise institutions. Alongside these changes, we find no change in retention rates. These results hold through a series of robustness checks, which we describe next.

First-Year Enrollment

Figure 1 shows the synthetic control plots for the NC Promise and the synthetic control enrollment for each of the three Promise institutions. The pretreatment plots show that the synthetic comparison sets are similar to the Promise institutions. In all three cases, these initial gains in enrollment decline by the second year of the policy. Table 2 presents the accompanying point estimates for the average synthetic control treatment effects. The average estimated gains in first-year enrollment ranged from 15 percent at UNC Pembroke to over 25 percent at Elizabeth City. However, as indicated by the p-values and the relative rank placement in their respective placebo tests, none of these increases were notably different from changes in enrollment at the comparison institutions over the same time period. At best, the 15 percent increase in first-year enrollment at UNC Pembroke reached a p-value of p = .20 from its rank as the 14th-most “extreme” among the 70 placebo permutations.

Figure 1.

Synthetic control and placebo graphs for the effect of NC Promise on first-year enrollment.

Table 2.

The Effect of NC Promise on Enrollments Using Carnegie Undergraduate Profile

	First-Year Enrollment					Transfer Enrollment
	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE
Elizabeth City	.255	.743	52/70	35/41	.145	.499	.514	36/70	13/25	.153
UNC Pembroke	.154	.200	14/70	12/41	.000	.411	.129	9/70	2/36	.020
Western Carolina	.225	.816	111/136	54/67	.000	.320	.015	2/136	1/58	.000

Notes. Average treatment effect (ATE) expressed in terms of log enrollment totals. p-Value calculation is described more thoroughly in the main body of the text and roughly represents the proportion of placebo tests that result in effect estimates as or more extreme than the effect estimate at the Promise institution. To aid in the interpretation of these p-values, the table also presents the rank of the post-/pre-RMSPE ratio at the Promise institution among the placebo estimates among all institutions in the synthetic control donor pool (two-sided) and among all ATE estimates that are in the same direction (positive/negative) as the ATE at the Promise institution (one-sided). RMSPE gives the root-mean-squared prediction error of the synthetic match, with smaller values indicating a stronger pretreatment match between the Promise institution and the outcome-specific synthetic control.

To illustrate where the estimated changes in first-year enrollment fall among these placebo tests, Figure 1 plots the root-mean-squared error of the treatment estimate (plotted in orange) relative to each of the placebo tests. In short, it shows the average difference from a given case and its synthetic comparison. Ideally, the values prior to the implementation of NC Promise should all be close to zero, indicating a strong synthetic match. In contrast, if it is significant, the treated condition should deviate from zero in the posttreatment period, while the placebo cases should all continue close to zero. Because of the relatively high MDE of our analysis, it is difficult to rule out that the changes in first-year enrollment that occurred at the three Promise institutions were a result of NC Promise and not akin to changes in enrollment at comparison institutions that would not have been affected by the NC Promise policy.

Additionally, note that UNC Pembroke and Western Carolina each have relatively small root mean prediction errors (RMSPE), indicating a strong synthetic match, but the RMSPE for Elizabeth City is .145, which is relatively large given the size of the estimated effect of NC Promise. Perhaps because of its relative rarity as a small, liberal arts HBCU, the synthetic algorithm struggled to find a strong match for Elizabeth City throughout our analysis.

Transfer Enrollment

Figure 2 shows the time series plots of Promise and synthetic control institution transfer-student enrollments, while Table 2 again shows the average estimated synthetic control treatment effect. All three campuses show large increases in transfer-student enrollment in 2018 and 2019 after NC Promise began. As with the estimate of changes in first-year enrollment, the synthetic match is relatively weak for Elizabeth City and strong for both UNC Pembroke and Western Carolina, as indicated by the RMSPEs. In part, given the quality of these pretreatment synthetic matches, the increases in transfers at UNC Pembroke and Western Carolina appear significant, while the increase at Elizabeth City does not.

Figure 2.

Synthetic control and placebo graphs for the effect of NC Promise on transfer enrollment.

Transfer enrollment increased at UNC Pembroke (p = .12), which was the ninth most extreme change in enrollment relative to the pretreatment period among the 70 comparisons, and the second most extreme among the 35 placebo comparisons that demonstrated increases in transfer enrollment. Across the various models, including the main one, estimates for this increase were mostly in the range of 40 percent. Likewise, the increase in transfer enrollment at Western Carolina (p = .01) was the second most extreme among 136 placebo comparisons, and the most extreme among the 58 comparisons with strict increases in transfer enrollment. Most models estimated the magnitude of this increase to be over 30 percent.

Changes in Student Demographics

Race/Ethnicity

Within these increases in transfer-student enrollment, and even within the lack of statistically significant changes in first-year student enrollment, there may have been shifts in the demographic characteristics of the students who enrolled in each of the Promise institutions, depending on whether they differentially responded to the NC Promise tuition change. Recall that because of the relatively low number of students in some racial-ethnic categories, our analyses of these shifts focus to total changes in enrollment among first-year and transfer students combined (all “new” students). The results of this analysis are visually presented in Figure 3, with the corresponding treatment point estimates presented in Table 3.

Figure 3.

Synthetic control and placebo graphs for the effect of NC Promise on new student enrollment by race/ethnicity.

Table 3.

The Effect of NC Promise on New Enrollments Using Carnegie Undergraduate Profile by Student Subgroups

	White
	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE
Elizabeth City	.456	.100	7/70	3/36	.016
UNC Pembroke	.252	.129	9/70	4/36	.000
Western Carolina	.143	.015	2/136	2/73	.003
	Black
	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE
Elizabeth City	.088	.929	65/70	37/38	.215
UNC Pembroke	.230	.071	5/70	4/38	.000
Western Carolina	.120	.324	44/136	2/64	.019
	Hispanic
	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE
Elizabeth City	.840	.700	49/70	12/25	.511
UNC Pembroke	.639	.214	15/70	4/26	.000
Western Carolina	.102	.941	128/136	70/76	.089

Notes. Average treatment effect (ATE) expressed in terms of log enrollment totals. p-Value calculation described more thoroughly in the main body of the text and roughly represents the proportion of placebo tests that result in effect estimates as or more extreme than the effect estimate at the Promise institution. To aid in the interpretation of these p-values, the table also presents the rank of the post-/pre-RMSPE ratio at the Promise institution among the placebo estimates among all institutions in the synthetic control donor pool (two-sided) and among all ATE estimates that are in the same direction (positive/negative) as the ATE at the Promise institution (one-sided). RMSPE gives the root-mean-squared prediction error of the synthetic match, with smaller values indicating a stronger pretreatment match between the Promise institution and the outcome-specific synthetic control.

White student enrollment increased at Elizabeth City after NC Promise was implemented. Here, we see White enrollment increased by 45 percent (p = .1), a recovery after declines in prior years. Although this effect seems large, recall that ECSU is an HBCU and so had a relatively small White student population to start. This change is illustrated in Panel A of Figure 3, which shows a close match in the preperiod between Elizabeth City and its control and an increase at Elizabeth City after NC Promise began, while White student enrollment at the control institutions declined. Based on the placebo test graph, ECSU has the most extreme increase in White student enrollment after the policy, and the seventh most extreme change of any of the 70 placebo cases.

There is evidence of increases across all three demographic groups at UNC Pembroke. Panel B of Figure 3 shows these increases. Specifically, White student enrollment increased by 25 percent (p = .13), Black student enrollment increased by 23 percent (p = .07), and Hispanic student enrollment increased by 64 percent (p = .21). The increase in Hispanic student enrollment has a relatively low p-value, but it is the fourth most extreme increase among the 26 placebo tests that also indicated increases in enrollment.

Finally, Panel C of Figure 3 illustrates a notable increase in White students (14 percent; p = .015) at Western Carolina, but no apparent changes in Black or Hispanic student enrollment. The change in White students was the second most extreme change of any of the 136 placebo tests.

It is worth noting that none of our estimates for changes in Black or Hispanic students indicate a decrease in enrollment, regardless of the significance of the result. Thus, we see no evidence that enrollment changes induced by NC Promise led to any crowding out of those traditionally underrepresented populations.

Pell-Eligible Students

Changes in new Pell-eligible student enrollment are found presented in Table 4 and Figure 4. Although all estimates of the change in new Pell-eligible student enrollment are positive, none of the estimates appear notable relative to the placebo comparisons. We largely discount the ECSU results because of the low apparent quality of the synthetic control fit.

Table 4.

The Effect of NC Promise on Log New Pell Enrollments Using Carnegie Undergraduate Profile by Student Subgroups

	Log Pell Students
	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE
Elizabeth City	.055	.957	67/70	30/36	.127
UNC Pembroke	.072	.500	38/70	18/36	.031
Western Carolina	.057	.397	54/136	23/73	.000

Notes. Average treatment effect (ATE) expressed in terms of log enrollment totals. p-Value calculation described more thoroughly in the main body of the text and roughly represents the proportion of placebo tests that result in effect estimates as or more extreme than the effect estimate at the Promise institution. To aid in the interpretation of these p-Values, the table also presents the rank of the post-/pre-RMSPE ratio at the Promise institution among the placebo estimates among all institutions in the synthetic control donor pool (two-sided) and among all ATE estimates that are in the same direction (positive/negative) as the ATE at the Promise institution (one-sided). RMSPE gives the root-mean-squared prediction error of the synthetic match, with smaller values indicating a stronger pretreatment match between the Promise institution and the outcome-specific synthetic control.

Figure 4.

Synthetic control and placebo graphs for the effect of NC Promise on first-year enrollment.

Retention

Because NC Promise lowered the cost of attendance for both new and continuing students, it may affect retention, the year-to-year re-enrollment of already enrolled students, particularly for students who struggle to afford to continue their studies, perhaps making it easier for them to continue. As shown in Table 5 and graphically in Figure 5, we find no evidence that NC Promise affected retention rates at any of the three promise institutions. All estimates for the change in retention rates were less than +/− four percentage points, and none appeared significant in comparison to placebo cases. These results indicate, on the one hand, that the tuition changes resulting from NC Promise did not seem to induce any more students to stay enrolled at the Promise institutions than otherwise would have. On the other hand, it indicates that the new students who enrolled at the Promise institutions as a result of NC Promise remained enrolled, presumably as successfully, as the students who had enrolled at the Promise institutions prior to the NC Promise policy.

Table 5.

The Effect of NC Promise on Percent of Students Re-Enrolling in Following Year

	Retention Rate
	ATE	p-Value	Rank(two-sided)	Rank(one-sided)	RMSPE
Elizabeth City	−.216	.971	68/70	33/33	2.405
UNC Pembroke	3.840	.486	34/70	13/32	1.407
Western Carolina	.266	.440	44/100	23/53	.227

Notes. Average treatment effect (ATE) expressed in terms of log enrollment totals. p-Value calculation described more thoroughly in the main body of the text and roughly represents the proportion of placebo tests that result in effect estimates as or more extreme than the effect estimate at the Promise institution. To aid in the interpretation of these p-Values, the table also presents the rank of the post-/pre-RMSPE ratio at the Promise institution among the placebo estimates among all institutions in the synthetic control donor pool (two-sided) and among all ATE estimates that are in the same direction (positive/negative) as the ATE at the Promise institution (one-sided). RMSPE gives the root-mean-squared prediction error of the synthetic match, with smaller values indicating a stronger pretreatment match between the Promise institution and the outcome-specific synthetic control.

Figure 5.

Synthetic control and placebo graphs for the effect of NC Promise on student retention.

Robustness Checks

To check that our results are not dependent either on choices we made as researchers or on outlying, but influential, characteristics of the data, we run several robustness checks to bolster the conclusions from our results. We take three approaches to assessing the robustness of our results. First, we use a “leave-one-out” process where we drop each of the top contributors to our synthetic controls one at a time from the donor pool and repeat the analysis to ensure no single comparison unit drives the results (Abadie et al., 2015). Second, we check that our choice of time period and any idiosyncratic trends do not affect our results by repeating the analysis using smoothed time trends that average multiple years together rather than including every year in the analysis (McClelland & Gault, 2017). Finally, we make sure that our results are robust to our choice to use institutions with the same Carnegie Undergraduate Profile Classification as our primary donor pool by first repeating the analyses using different Carnegie Classification schemes and, second, using comparison groups based on geography. In short, none of these checks change our conclusions in substantive ways. Related tables, figures, and a more complete discussion of these checks are available in Appendix C.

Discussion

NC Promise represents an important test of how a universal and highly visible decrease in public university sticker-price tuition affects students’ college enrollment choices. Although it did not make tuition completely free, the program has more in common with “free college” proposals than many of the promise policies its name references, particularly for students who qualify for need-based financial aid. Our synthetic control study provides the first glimpse into the effect of lowering the sticker price in the four-year public university context.

Four things are apparent from our findings. First, first-year student enrollment appears relatively unaffected by NC Promise implementation. Second, there were notable increases in transfer-student enrollment at the non-HBCU campuses affected by the policy, and the HBCU showed consistently positive results in this category. Third, although the policy did not generally affect the demographic composition of the affected universities, it did increase Hispanic student enrollment at the campus that serves one of the more racially diverse areas of North Carolina. Finally, despite the changes in enrollment that we see, there were no discernible changes in student retention at the three Promise institutions, which we take as a positive finding because the policy could have easily induced less academically prepared students to enroll. We discuss these findings in turn.

Despite what appeared to be an initial boost in the first year after the policy began, first-year student enrollment remained largely unchanged at each of the three Promise institutions. This result lies in contrast to recent results from Michigan that show, for example, that clear communication about free tuition and fees can have notable effects on college enrollment (S. Dynarski et al., 2021). One explanation for the difference may be that although NC Promise was well publicized, the Michigan HAIL scholarship sent direct communication to the specific students who would benefit from the scholarship. Additionally, HAIL offered free tuition and fees, while NC Promise only reduced tuition. Ultimately, the tuition change caused by NC Promise was small relative to the total cost of attendance—including fees and expected food and housing costs—particularly for the in-state students that comprise the vast majority of student enrollments at the Promise institutions. The size of a financial aid package matters for increasing enrollment, and similarly, the size of the decrease in cost of attendance at a Promise institution matters (S. M. Dynarski, 2003; Page & Scott-Clayton, 2016). Thus, the decrease in tuition may not be substantial enough to alter the preferences of price-sensitive students, or there may still have been gaps in how families who heard about the NC Promise tuition reduction were able to act on that information. This reasoning also suggests an explanation for why we did not find much evidence to show that enrollment patterns changed among Pell-eligible students.

Although the policy does not appear to have been significant enough to affect the enrollment patterns of first-year students, the fact that it resulted in notable increases in transfer-student enrollment suggests an important way such policies may increase educational opportunity for the nontraditional, and generally more racially and ethnically diverse, students who tend to enroll in community colleges. The increase in transfer-student enrollment at the Promise institutions may have resulted in part because the changes in tuition at the Promise institutions put the total cost of attendance at levels comparable to the nearby community colleges that regularly send a large number of transfer students to the three Promise campuses. That is, a successful transfer to a Promise institution would have resulted in very little change to what a student was already paying for a community college education, potentially alleviating cost concerns for prospective transfer students during this period (Cooper et al., 2020).

There is no finding directly comparable to the transfer results, but they do appear to mirror the increased transfer likelihood observed after tuition was reduced in the community college context (Acton, 2021; Denning, 2017). For example, Acton (2021) found that community college students were 6.5 percent more likely to transfer to a four-year college for every $1,000 decrease in community college tuition. In a rough analog, tuition declined by nearly $3,000 at Western Carolina, which would seem to imply an over 19 percent increase in the likelihood of community college students transferring to WCU. This would result in the roughly 100 transfer-student increase at WCU if roughly 510 community college students were aware of the decrease, making our estimate of a 30 percent increase in transfer enrollment at WCU large but plausible.

Despite this increase in transfer access, there does not appear to have been widespread increases in racial diversity on the Promise campuses. With the exception of the growth of enrollment of Hispanic students at UNC Pembroke, we did not consistently observe overall increases in enrollment in White, Black, or Hispanic students at Elizabeth City or Western Carolina. This ultimately puts our work in line with Monaghan (2025), who notes that free college programs encourage enrollment at the colleges that offer them, but these changes do not ultimately appear to reduce race-based enrollment disparities. Part of this result may stem from the characteristics of the Promise institutions and the local populations they serve. Neither Elizabeth City nor Western Carolina serves parts of the state with demographics that are likely to shift the overall composition of their student bodies. Pembroke, in contrast, is located in a part of the state with some of the highest proportions of Hispanic residents. The increase in Hispanic students at UNC Pembroke, combined with our finding that transfer-student enrollment increased, does suggest connections to research in the Oklahoma context, which found that the Tulsa Achieves program, which offered free tuition to a community college, made it more likely that Hispanic students who enrolled in the community college ultimately transferred to a four-year institution (Bell & Gándara, 2021). Others have suggested that this population is particularly sensitive to concerns about financing college (e.g., Boatman et al., 2017), so this phenomenon is worth further investigation.

As the data become available, future research will need to evaluate the degree-completion rates of these transfer students, given that the lower-income students who the Promise policy likely incentivized to transfer often complete their degrees at lower rates than higher-income students (Jenkins & Fink, 2016). Even though Promise may have cleared financial barriers for community college students seeking a bachelor’s degree, the bureaucratic and academic barriers to transfer, such as unclear pathways and credit loss, may remain (Jabbar et al., 2022; Jenkins & Fink, 2015).

Finally, we did not find that the NC Promise was associated with changes in the rates at which students remained enrolled in Promise institutions. On the one hand, policymakers may find this result disappointing because the lower cost of attendance at Promise institutions should have made it easier for students to afford to remain enrolled, suggesting that retention rates should have increased. On the other hand, although we did not see substantial changes in enrollment among nontransfer students or by racial/ethnic categories, it is possible that enrollment patterns shifted among categories of students we were not able to study, such as by student income or level of academic preparation for college. These students may have had a lower likelihood of remaining enrolled in college, so the overall stability of retention rates at the Promise institutions suggests that NC Promise may have helped these students succeed at rates comparable to students who enrolled prior to the start of NC Promise.

Conclusion

The results of this study provide early evidence of the NC Promise’s contribution toward the state’s access and attainment goals. This study comes at a crucial time for NC policymakers, as the policy has received national media attention (e.g., Brown, 2022) and the legislature has announced an expansion of the policy to Fayetteville State University. Furthermore, this study will be one of the first to quantify the impact of programs that reduce sticker prices at public four-year colleges. Our findings here represent just the start of understanding the effects of this program. Ultimately, it will be important to understand whether NC Promise helped students complete college degrees. Unfortunately, the short amount of time NC Promise has been in place and the COVID-19 disruption make it difficult to study these outcomes at this time. The one effect of promise that is clear, however, is that it appears to open a new path to a four-year degree for community college students. And that itself is promising.

Supplemental Material

sj-docx-1-ero-10.1177_23328584261443496 – Supplemental material for Do Students Respond to Sticker-Price Reductions? Evidence From the North Carolina Promise

Supplemental material, sj-docx-1-ero-10.1177_23328584261443496 for Do Students Respond to Sticker-Price Reductions? Evidence From the North Carolina Promise by Daniel Klasik, William Zahran, Rachel Worsham and Matthew G. Springer in AERA Open

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The John M. Belk Endowment provided generous support for this work.

ORCID iDs

Daniel Klasik

William Zahran

Rachel Worsham

Matthew G. Springer

Notes

Authors

DANIEL KLASIK is an associate professor in the School of Education at the University of North Carolina at Chapel Hill. His research focuses on how policy can help students make choices about whether and where to go to college.

WILLIAM ZAHRAN is a Postdoctoral Research Fellow at the University of Texas at Austin. His research examines how state-level policies shape postsecondary education affordability and student success.

RACHEL WORSHAM is a senior research manager at the Center for Education Policy Research at Harvard University. Her research focuses on how public policy shapes college access and student success.

MATTHEW G. SPRINGER is a cofounder and managing partner at Basis Policy Research, with more than two decades of experience leading large-scale research and evaluation initiatives aimed at improving educational effectiveness across K–12 and postsecondary systems through evidence-based reform.

References

Abadie

(2021). Using synthetic controls: Feasibility, data requirements, and methodological aspects. Journal of Economic Literature, 59(2), 391–425. https://doi.org/10.1257/jel.20191450

Abadie

Diamond

Hainmueller

(2010). Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. Journal of the American Statistical Association, 105(490), 493–505. https://doi.org/10.1198/jasa.2009.ap08746

Abadie

Diamond

Hainmueller

(2015). Comparative politics and the synthetic control method. American Journal of Political Science, 59(2), 495–510. https://doi.org/10.1111/ajps.12116

Abadie

Gardeazabal

(2003). The economic costs of conflict: A case study of the Basque Country. American Economic Review, 93(1), 113–132. https://doi.org/10.1257/000282803321455188

Acton

(2021). Effects of reduced community college tuition on college choices and degree completion. Education Finance and Policy, 16(3), 388–417. https://doi.org/10.1162/edfp_a_00313

Aina

Baici

Casalone

Pastore

(2022). The determinants of university dropout: A review of the socio-economic literature. Socio-Economic Planning Sciences, 79, 101102. https://doi.org/10.1016/j.seps.2021.101102

Bailey

Dynarski

(2011). Gains and gaps: Changing inequality in U.S. college entry and completion (NBER Working Paper No. 17633). https://doi.org/10.3386/w17633

Barnes

Harris

D. N.

Nguyen

(2022). What effect did the New Orleans school reforms have on youth crime? Policy brief. Education Research Alliance for New Orleans.

Bartik

T. J.

Hershbein

Lachowska

(2021). The effects of the Kalamazoo Promise Scholarship on college enrollment and completion. Journal of Human Resources, 56(1), 269–310. https://doi.org/10.3368/jhr.56.1.0416-7824R4

10.

Bassok

Miller

L. C.

Galdo

(2016). The effects of universal state pre-kindergarten on the child care sector: The case of Florida’s voluntary pre-kindergarten program. Economics of Education Review, 53, 87–98. https://doi.org/10.1016/j.econedurev.2016.05.004

11.

Becker

G. S.

(1962). Investment in human capital: A theoretical analysis. Journal of Political Economy, 70(5), 9–49. https://doi.org/10.1086/258724

12.

Becker

G. S.

(1993). Human capital: A theoretical and empirical analysis, with special reference to education (3rd ed). University of Chicago Press.

13.

Bell

Gándara

(2021). Can free community college close racial disparities in postsecondary attainment? How Tulsa achieves affects racially minoritized student outcomes. American Educational Research Journal, 58(6), 1142–1177. https://doi.org/10.3102/00028312211003501

14.

Bettinger

(Ed.). (2004). How financial aid affects persistence. In College choices: The economics of where to go, when to go, and how to pay for it (pp. 207–238). University of Chicago Press.

15.

Bifulco

Rubenstein

Sohn

(2017). Using synthetic controls to evaluate the effect of unique interventions: The case of Say Yes to Education. Evaluation Review, 41(6), 593–619. https://doi.org/10.1177/0193841X17742233

16.

Boatman

Evans

B. J.

Soliz

(2017). Understanding loan aversion in education: Evidence from high school seniors, community college students, and adults. AERA Open, 3(1), 2332858416683649. https://doi.org/10.1177/2332858416683649

17.

Brown

(2022). Model or fluke: North Carolina’s radical college-affordability program saved one floundering university and helped two others grow. Will it last? The Chronicle of Higher Education.

18.

Carruthers

C. K.

Fox

W. F.

(2016). Aid for all: College coaching, financial aid, and post-secondary persistence in Tennessee. Economics of Education Review, 51, 97–112. https://doi.org/10.1016/j.econedurev.2015.06.001

19.

Carruthers

C. K.

Welch

J. G.

(2019). Not whether, but where? Pell grants and college choices. Journal of Public Economics, 172, 1–19. https://doi.org/10.1016/j.jpubeco.2018.11.006

20.

Cooper

D. M.

Nguyen

Karandjeff

Purnell

Kretz

Rodriguez-Kiino

Chaplot

Nguyen

(2020). Students speak their truth about transfer: What they need to get through the gate. RP Group.

21.

Cornwell

Mustard

D. B.

Sridhar

D. J.

(2006). The enrollment effects of merit-based financial aid: Evidence from Georgia’s HOPE program. Journal of Labor Economics, 24(4), 761–786. https://doi.org/10.1086/506485

22.

Crisp

Horn

C. L.

Kuczynski

Zhou

Cook

(2019). Describing and differentiating four-year broad access institutions: An empirical typology. The Review of Higher Education, 42(4), 1373–1400. https://doi.org/10.1353/rhe.2019.0069

23.

Denning

J. T.

(2017). College on the cheap: Consequences of community college tuition reductions. American Economic Journal: Economic Policy, 9(2), 155–188. https://doi.org/10.1257/pol.20150374

24.

Dynarski

Libassi

C. J.

Michelmore

Owen

(2021). Closing the gap: The effect of reducing complexity and uncertainty in college pricing on the choices of low-income students. American Economic Review, 111(6), 1721–1756. https://doi.org/10.1257/aer.20200451

25.

Dynarski

Page

L. C.

Scott-Clayton

(2022). College costs, financial aid, and student decisions (NBER Working Paper No. 30275).

26.

Dynarski

Scott-Clayton

(2013). Financial aid policy: Lessons from research. The Future of Children, 23(1), 67–91. https://doi.org/10.1353/foc.2013.0002

27.

Dynarski

(2003). Does aid matter? Measuring the effect of student aid on college attendance and completion. American Economic Review, 93(1), 279–288. https://doi.org/10.1257/000282803321455287

28.

Engberg

M. E.

(2012). Pervasive inequality in the stratification of four-year college destinations. Equity & Excellence in Education, 45(4), 575–595. https://doi.org/10.1080/10665684.2012.717486

29.

Fitzpatrick

M. D.

(2008). Starting school at four: The effect of universal pre-kindergarten on children’s academic achievement. The BE Journal of Economic Analysis & Policy, 8(1), 1897–1897. https://doi.org/10.2202/1935-1682.1897

30.

Gonzalez

Bozick

Daugherty

Scherer

Singh

Suarez

Ryan

Schweig

(2014). New Haven Promise: An early look at college preparation, access, and enrollment of New Haven Public School Students (2010-2013). RAND Corporation. https://doi.org/10.7249/RB9811.1

31.

González Canché

M. S

. (2018). Geographical network analysis and spatial econometrics as tools to enhance our understanding of student migration patterns and benefits in the U.S. higher education network. The Review of Higher Education, 41(2), 169–216. https://doi.org/10.1353/rhe.2018.0001

32.

Grodsky

Jones

M. T.

(2007). Real and imagined barriers to college entry: Perceptions of cost. Social Science Research, 36(2), 745–766. https://doi.org/10.1016/j.ssresearch.2006.05.001

33.

Holzman

Klasik

Baker

(2020). Gaps in the college application gauntlet. Research in Higher Education, 61(7), 795–822. https://doi.org/10.1007/s11162-019-09566-8

34.

Jabbar

Schudde

Garza

McKinnon-Crowley

(2022). Bridges or barriers? How interactions between individuals and institutions condition community college transfer. The Journal of Higher Education, 93(3), 375–398. https://doi.org/10.1080/00221546.2021.1953339

35.

Jenkins

Fink

(2015). What we know about transfer. Columbia University, Teachers College, Community College Research Center.

36.

Jenkins

Fink

(2016). Tracking transfer: New measures of institutional and state effectiveness in helping community college students attain bachelor’s degrees. Columbia University, Teachers College, Community College Research Center.

37.

Kane

T. J.

(2007). Evaluating the impact of the DC tuition assistance grant program. Journal of Human Resources, 42(3), 555–582. https://doi.org/10.3368/jhr.xlii.3.555

38.

Klasik

(2013). The ACT of enrollment: The college enrollment effects of state-required college entrance exam testing. Educational Researcher, 42(3), 151–160. https://doi.org/10.3102/0013189X12474065

39.

Klasik

Holzman

Jones

A. J.

(2025). The many paths to college enrollment: Re-conceptualizing the transition to college (EdWorkingPaper No. 25-1269). Annenberg Institute at Brown University. https://doi.org/10.26300/54c6-2316

40.

Levine

P. B.

(2014). Transparency in college costs (Brookings Institution Economic Studies Working Paper).

41.

Levine

P. B.

Russell

L. C.

(2023). Do college applicants respond to changes in sticker prices even when they don’t matter? Education Finance and Policy, 18(3), 365–394. https://doi.org/10.1162/edfp_a_00372

42.

McClelland

Gault

(2017). The synthetic control method as a tool to understand state policy. The Urban Institute.

43.

Millea

Wills

Elder

Molina

(2018). What matters in college student success? Determinants of college retention and graduation rates. Education, 138(4), 309–322.

44.

Monaghan

D. B.

(2025). How powerful are promises? A comprehensive review of the causal mechanisms and outcomes of ‘free college’programs in the United States. Policy Reviews in Higher Education, 9(1), 47–80. https://doi.org/10.1080/23322969.2024.2436923

45.

Noble

Flynn

N. T.

Lee

J. D.

Hilton

(2007). Predicting successful college experiences: Evidence from a first-year retention program. Journal of College Student Retention: Research, Theory & Practice, 9(1), 39–60. https://doi.org/10.2190/6841-42JX-X170-8177

46.

Odle

T. K.

(2022). Free to spend? Institutional autonomy and expenditures on executive compensation, faculty salaries, and research activities. Research in Higher Education, 63(1), 1–32. https://doi.org/10.1007/s11162-021-09642-y

47.

Odle

T. K.

Delaney

J. A.

(2022). You are admitted! Early evidence on enrollment from Idaho’s direct admissions system. Research in Higher Education, 63(6), 899–932. https://doi.org/10.1007/s11162-022-09675-x

48.

Page

L. C.

Scott-Clayton

(2016). Improving college access in the United States: Barriers and policy responses. Economics of Education Review, 51, 4–22. https://doi.org/10.1016/j.econedurev.2016.02.009

49.

Perna

L. W.

(2006). Studying college access and choice: A proposed conceptual model. In Smart

J. C.

(Ed.), Higher education: Handbook of theory and research (Vol. 21, pp. 99–157). Kluwer Academic Publishers. https://doi.org/10.1007/1-4020-4512-3_3

50.

Rubin

P. G.

Canché

M. S. G.

(2019). Test-flexible admissions policies and student enrollment demographics: Examining a public research university. The Review of Higher Education, 42(4), 1337–1371. https://doi.org/10.1353/rhe.2019.0068

51.

Rubin

R. B.

(2011). The Pell and the poor: A regression-discontinuity analysis of on-time college enrollment. Research in Higher Education, 52(7), 675–692. https://doi.org/10.1007/s11162-011-9215-6

52.

Seltzer

(2016, June 5). Black colleges dropped from bill they opposed. Inside Higher Ed. https://www.insidehighered.com/news/2016/06/06/hbcus-cut-north-carolina-500-tuition-bill

53.

Skinner

B. T.

(2019). Choosing college in the 2000s: An updated analysis using the conditional logistic choice model. Research in Higher Education, 60(2), 153–183. https://doi.org/10.1007/s11162-018-9507-1

54.

Thaler

(1980). Toward a positive theory of consumer choice. Journal of Economic Behavior & Organization, 1(1), 39–60. https://doi.org/10.1016/0167-2681(80)90051-7

55.

Undergraduate Profile Classification. (n.d.). Carnegie Classification of Institutions of Higher Education. https://carnegieclassifications.acenet.edu/carnegie-classification/classification-methodology/undergraduate-profile-classification/

56.

Villemain

(2021, November 14). NC Promise is expanding. The Assembly NC. http://www.theassemblync.com/education/higher-education/nc-promise-imminent-expansion/

57.

Ward

J. D.

Corral

(2022). Resetting prices: Estimating the effect of tuition reset policies on institutional finances and enrollment. Research in Higher Education, 64, 862–892. https://doi.org/10.1007/s11162-022-09723-6

58.

Zahran

Klasik

(2025). Tuition reduction and student loan debt: Evidence from the North Carolina Promise. Research in Higher Education, 66(6), 32. https://doi.org/10.1007/s11162-025-09853-7

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.48 MB