Escalating schedules of incentives increase physical activity with no differences between deposit and no‐deposit groups: A systematic replication

Abstract

Physical inactivity has increasingly affected public health in the United States during the COVID‐19 pandemic as it is associated with chronic diseases such as arthritis, cancer, and heart disease. Contingency management has been shown to increase physical activity. Therefore, the present study sought to evaluate the effects of an escalating schedule of monetary reinforcement with a reset contingency on physical activity, as compared between 2 counterbalanced groups in which a monetary deposit of $25 was either required (deposit group) or not (no‐deposit group). Twenty‐five adults wore Fitbit accelerometers to monitor step counts. An ABA reversal design was used; in the 2 baseline phases, no programmed contingencies were in place for step counts. During intervention, step goals were set using a modified 70^th percentile schedule with a 7‐day window: Reaching the first goal would result in $0.25, and incentives increased by $0.25 for each subsequent day in which the goal was met. Failure to reach a goal resulted in a reset of the monetary incentive value to $0.25. Ten out of 12 participants from the deposit group were determined to be responders to intervention, whereas 8 out of 13 participants from the no‐deposit group were determined to be responders to intervention. Overall, there were no significant differences between the groups' step counts. However, the deposit group's intervention was cheaper to implement, which suggests that deposit contracts are a viable modification for physical activity interventions.

Physical inactivity, or sedentary behavior, is any activity that expends less than or equal to 1.5 metabolic equivalents in a sitting or reclining posture (Sedentary Behaviour Research Network, 2016). Approximately one‐quarter of the population in the United States is physically inactive and does not engage in activity beyond the baseline activities of daily living (Centers for Disease Control and Prevention [CDC], 2022). Americans' inactivity increases as they age, with 25.4% of adults aged 50 to 64 years and 35.3% of adults aged 75 years and older classified as inactive (Watson et al., 2016). From 2003 to 2016, self‐reported sedentary behavior in the United States increased from 5.5 to 6.4 hr per day (Yang et al., 2019); during the COVID‐19 pandemic, Americans reduced their physical activity by more than 20% and increased their sitting time by almost 40% (Meyer et al., 2020). Notably, physical activity is one of the most important modifiable factors to prevent cardiovascular disease (Lippi & Sanchis‐Gomar, 2020; Mok et al., 2019), cancer, and all‐cause mortality (Mok et al., 2019). Physical inactivity also increases healthcare costs, with physically inactive adults aged 50 and older costing the United States $860 billion a year (CDC, 2021). The risk of cardiovascular disease is 50% higher for adults who sit for 8 hr per day than for those who sit for less than 4 hr per day (Ekelund et al., 2016). Reductions in sedentary behavior are related to reduced risk of cardiovascular disease and all‐cause mortality (Lavie et al., 2019).

The CDC physical activity guidelines (CDC, 2018) recommend that for important health benefits, adults engage in at least 150 min of moderate‐intensity activity every week with muscle‐strengthening activities on two or more days per week, or 75 min of vigorous‐intensity activity every week with muscle‐strengthening activities on two or more days per week. Moderate physical activity is any activity that increases metabolic activity to between 3.0‐5.9 metabolic equivalents (e.g., walking fast), whereas vigorous physical activity is an activity that increases metabolic activity to more than 6.0 metabolic equivalents (e.g., running; CDC, 2018). More time spent doing moderate‐intensity activity will lead to increased benefits (CDC, 2018). One behavioral method that has successfully increased individuals' physical activity levels to meet these guidelines is contingency management (e.g., Batchelder & Washington, 2021; Krebs & Nyein, 2021; Kurti & Dallery, 2013; Stedman‐Falls & Dallery, 2020; Washington et al., 2014; Washington et al., 2016).

Contingency management uses behavioral techniques to modify a variety of health behaviors in addition to physical activity, such as abstinence from drug use (e.g., Higgins et al., 1991; Roll & Higgins, 2000; Roll et al., 2006; Romanowich & Lamb, 2015) and medication adherence (e.g., Raiff et al., 2016). A target health behavior is objectively monitored and measured, and reinforcers are delivered contingent on the target behavior and withheld when the target behavior is not emitted (Meredith et al., 2014). Percentile schedules of reinforcement can be used to incrementally set goals (Galbicka, 1994); in other words, they are designed to gradually shape behavior toward a terminal goal, such as moderate to vigorous levels of physical activity (e.g., Washington et al., 2014). Similarly, escalating schedules have been effective in contingency management studies treating substance use disorder (e.g., Higgins et al., 1991; Roll & Higgins, 2000; Roll et al., 2006; Romanowich & Lamb, 2015) and increasing physical activity (e.g., Batchelder & Washington, 2021). In escalating schedules, the magnitude of reinforcement gradually increases as goals are consecutively met. In many cases, a reset contingency is used in which failure to meet goals or emit a target behavior results in resetting the reinforcer magnitude to its original value.

In a comparison of five different schedules of reinforcement in contingency management‐based treatment of methamphetamine use, Roll et al. (2006) found that an escalating schedule of reinforcement with a reset contingency initiated abstinence the quickest and protected against relapse. Although some studies have mixed results (e.g., Regnier et al., 2022), several drug abstinence studies have supported the notion that escalating schedules with reset contingencies are more effective than other schedules (e.g., escalating schedules without resets; fixed schedules; see Roll & Higgins, 2000; Roll et al., 1996; Romanowich & Lamb, 2015).

Although contingency management is effective for changing health behaviors, it is also costly. One method to reduce the cost is to use a deposit contract, in which the participant deposits their own money to earn back as reinforcement for meeting the contingency related to the target behavior. Deposit contracts have been effective for weight loss (e.g., John et al., 2011), increasing physical activity (e.g., Krebs & Nyein, 2021; Stedman‐Falls & Dallery, 2020; Washington et al., 2016), and smoking abstinence (e.g., Dallery et al., 2008; Halpern et al., 2015; Jarvis & Dallery, 2017).

Washington et al. (2016) used an ABA reversal design to target the physical activity of 25 underactive adults (i.e., less than 10,000 daily steps in baseline) via a fixed schedule of reinforcement. After the initial baseline, participants were randomized to a no‐deposit or deposit group. During the intervention, daily step count goals (based on a 70^th percentile schedule) were set, and participants could earn a fixed amount of $1.50 per day for reaching goals. A bonus of $2.65 was given for reaching goals on three consecutive days. Participants could earn a total of $50 (with the deposit group participants contributing $25 of this total). Both groups had higher step counts during the intervention, and the improvement was not significantly different between groups. Therefore, both deposit and no‐deposit incentives were effective at increasing step counts during the intervention.

Similarly, Dallery et al. (2008) compared the effects of deposit and no‐deposit interventions on the cessation of cigarette smoking between groups using escalating schedules of reinforcement with a reset contingency. The daily goal for breath carbon monoxide was set at less than or equal to four parts per million. Both groups could earn a total of $78.50; the deposit group's earnings were comprised of $50 of their own money and $28.50 of researcher money and the no‐deposit group's earnings comprised $78.50 of researcher money. The incentive was initially set to $0.50 and increased by $0.10 for each day they met the target goal. A $3.00 bonus was earned for three consecutive days in which they met the target goal. Following any reset, three consecutive days of meeting the target goal resulted in a return to the highest magnitude earned before the reset. The deposit and no‐deposit groups both showed improvements in abstinence (65% vs. 63% abstinence during treatment, respectively), and the improvements were not statistically different.

The results from Washington et al. (2016) and Dallery et al. (2008) suggest that individuals who deposit money in contingency management studies seem to perform at least as well as those who do not deposit money. Romanowich and Lamb (2015) have suggested that depositing one's own money might increase the efficacy of contingency management due to loss aversion (i.e., the tendency to avoid losses rather than to acquire equivalent gains; Kahneman & Tversky, 1984). In this vein, Halpern et al. (2015) found that many participants in a smoking cessation clinical trial were reluctant to deposit $150 despite having the opportunity to earn their money back (plus an additional $650) for quitting smoking. Participants who agreed to the deposit contract and those who did not were included in the analyses, but those that agreed to the deposit contract were coded as having accepted the intervention. After accounting for the acceptance of the $150 deposit, the deposit intervention was more efficacious than the no‐deposit intervention. Specifically, 52% of participants in the deposit group that accepted the deposit remained abstinent through 6 months, whereas only 17% of those in the no‐deposit group remained abstinent through 6 months. Thus, Halpern et al. proposed that for those willing to deposit money, deposit‐based interventions can leverage participants' loss aversion to produce more meaningful behavior changes.

In the present study, we used an escalating schedule of incentives with a reset contingency similar to Dallery et al. (2008), however, the present study did not include bonuses or returns to the previous highest magnitudes for consecutive emissions of target behaviors. As a result, this study is a systematic replication and extension of Dallery et al. in that the efficacy of an escalating reinforcement schedule with a reset contingency was evaluated between deposit and no‐deposit groups. Further, this study serves as a systematic replication of Washington et al. (2016) wherein a percentile schedule was used to increase physical activity; in the current study, we employed an escalating schedule rather than a fixed schedule.

Method

Participants

Forty‐three participants were recruited by flyers posted around a college campus in the southeastern United States. Exclusion criteria included having asthma, being pregnant, any reason that would preclude participants from partaking in physical activity, being under the age of 18 years old or over the age of 64, and walking greater than 9,000 steps daily during baseline. In the first baseline phase, 18 participants were excluded because of ineligibility. Thus, 25 individuals were eligible to participate in the study and they were assigned to the deposit or no‐deposit groups via ABBA counterbalancing. A total of 12 participants were assigned to the deposit group and 13 participants were assigned to the no‐deposit group (see Table 1 for participant demographics).

Table 1.

Participant Demographics and Additional Intake Data

Variable	Deposit		No‐Deposit		p‐value
Participant Demographics	n	%	n	%
Gender
Male	3	25.00	3	23.08	0.99
Female	9	75.00	10	76.92
Additional Intake Data	M	SD	M	SD
Age	20.5	2.9	24.8	10.4	0.18
Days/Week Exercise	1.9	1.5	2.6	1.4	0.24
Minutes/Day Sedentary	367.5	152.5	465.5	483.1	0.51

Note. There were no significant differences at baseline across groups according to Chi‐Square and t‐tests.

Materials

The Fitbit Charge HR and Fitbit Zip accelerometers measured steps taken. Surveys on exercise habits and fitness were administered via Survey Monkey at both intake and exit. At exit, a treatment acceptability measure assessed whether participants (on a scale of 1‐10) believed the intervention, the Fitbit device, goals, incentives, and deposits (if applicable) were effective and convenient. In addition, the exit survey assessed any technology concerns (e.g., privacy issues; texting physical activity data).

Participants used their personal cell phones to text step counts to researchers before midnight each night. Researchers then entered daily step counts into Microsoft Excel, from which data were graphed and the next day's goal was calculated. The Fitbit website automatically synced with the participant's Fitbit device and researchers used this to verify step counts. All statistical analyses were carried out using R Statistical Software Version 3.3.2.

General Procedures

An ABA reversal design was used. When participants emailed the researcher with interest in the study, the researcher determined initial eligibility based on their answers to the Physical Activity Readiness Questionnaire (PAR‐Q; Hafen & Hoeger, 1994). If eligible, they were scheduled for an intake session where they completed a questionnaire about their current level of fitness. Eligible participants were given a Fitbit and charger for the duration of their participation in the study. Participants were also scheduled for weekly meetings with a researcher; and during these meetings, they would visit the lab to sync (i.e., verify) their Fitbit data with the Fitbit website, and receive money earned.

During baseline, participants were instructed to (a) wear the Fitbit, (b) walk as they typically did, and (c) report step counts before the end of each day. Baseline was at least 7 days and was extended, if necessary, until the stability criterion was reached (i.e., no indications of an increasing trend within the last 5 days; Sidman, 1960). After the final day of baseline, participants whose daily step counts averaged above 9,000 steps per day during baseline drew from a prize ticket jar with a 1 in 50 chance of winning a $25 Amazon gift card, and then were discharged from the study. At this visit, participants were assigned to either (a) the deposit group or (b) the no‐deposit group using ABBA counterbalancing. All participants were instructed to bring $25 in cash in the event that they were assigned to the deposit group.

The intervention lasted approximately 21 days (with some variation due to weekends). During this phase, participants continued wearing the Fitbit. The initial goal was the participants' fifth highest step count of the last 7 days of baseline, according to a modified 70^th percentile schedule. Once a participant reached a goal, future goals did not decrease. A goal remained unchanged until the participant met that goal, and then step counts continued to increase. For example, if participants' fifth highest step count was 4,995, this would be their goal for the following day; if they walked 5,000 steps that day, their goal would then increase for the following day based on the readjusted 7‐day lookback window. Therefore, participants could get a new goal every day.

The reinforcement schedule consisted of a daily increase of $0.25 for each consecutive goal met. Failure to meet any given goal resulted in no payment for the day and a reset of the next day's incentive to the original value of $0.25. In the deposit group, participants could earn their $25 back plus an additional $32.75 (for a total of $57.75). In the no‐deposit group, participants could earn up to $57.75. After participant‐texted step counts were corroborated by researchers through syncing the Fitbits to the website, the researchers informed the participant (a) whether they had met their goal or not, (b) about their reward when the goal was met, and (c) what the goal and potential reward was for the following day. Participants were given one “free day” per week (three free days in total) in which they could request to have a break from the intervention (requests were required at least one day in advance). Regardless of free days, each participant had approximately 21 days of active intervention.

There was a return to the baseline phase following the intervention; during baseline, participants continued to wear the Fitbit and text step counts to researchers before the end of the day. No programmed goals or contingencies for walking were in place. At the conclusion of the study, participants were asked to come in for a final visit to return their Fitbit and complete the Treatment Acceptability Questionnaire.

Data Analysis

The primary outcome measure was steps per day. Visual inspection of time‐series graphs was used to determine changes in step counts between the three phases. Descriptive statistics were calculated for step count in each condition and for each group. To calculate the effect size for each participant across the first baseline and intervention phase, a Tau‐U was conducted via the SingleCaseES package. Tau‐U is a calculation of the nonoverlap of phases with a correction for baseline trends; values generally range from ‐1 to 1 with values closer to 1 indicating less overlap between the phases; a small effect is .59 or less, medium effect is .6 to .85, and large effect is .85 to 1. Post‐hoc power analyses for the mixed methods linear regression were calculated via the SIMR package in R, with 100 simulations and a level of α = .05. Although the present study was underpowered to detect between‐subject effects, statistical analyses were included in accordance with prior research (Dallery et al., 2008; Washington et al., 2016). Power to detect between‐subject effects was 4% for the present sample; to have 80% power, we would need a sample size of 374 participants. A mixed methods linear regression was conducted via the lme4 package to determine if there were differences between the two groups and their step counts in the three different phases, with a significance level of α = .05. Welch's independent‐sample t‐tests were conducted to determine whether the percentage of goals met, percentage of change from baseline, money earned, and treatment acceptability differed across deposit and no‐deposit groups.

Results

Deposit Group

Figure 1 shows daily step counts across the course of the study. Many participants in the deposit group were responsive to the intervention; that is, step counts increased systematically in the intervention phase as they met their daily step count goals generated by the percentile schedules, relative to baseline conditions. Further, step count variability of within‐participant data was generally lower during intervention compared to baseline phases (see Figure 1). Table 2 includes (a) means and standard deviations of individual and group step counts for each phase, (b) the percentage of change in step counts during intervention compared to the first baseline phase, (c) Tau‐U effect size estimates, (d) the percentage of goals met, and (e) the total amount/highest amount of money earned. As summarized in Table 2, participants in the deposit group met their goals on most days (M = 74.6%); furthermore, the mean change in step count from the first baseline to intervention was 48.6%. In general, during the second baseline phase, step counts decreased to levels comparable to those of the first baseline.

Figure 1.

Deposit Group Step Count

Note. Step count per day for each participant in the deposit group in all phases of the study, ranked from highest to lowest percentage change from baseline 1 during intervention. Each open data point indicates one day's step count, each dotted black line indicates the goal for each day, and the dashed gray line indicates the 10,000 steps/day recommendation. Breaks in participants' data points in the intervention phase indicate that a “free day” was taken or that there were technical problems with the Fitbit.

Table 2.

Mean Steps Per Day

Participant	Baseline 1	Intervention	Baseline 2	Change in Intervention	Tau‐U*	Goals Met	Total Earned / Highest Amt
	M (SD)	M (SD)	M (SD)	(%)		(%)	($)
Deposit Group
622	5.9 (2.5)	10.4 (0.8)	4.0 (1.6)	77.3	0.79	90.0	17.50 / 2.25
605	6.6 (3.1)	11.6 (3.4)	5.9 (3.3)	74.4	0.82	100.0	57.75 / 5.25
618	8.6 (3.0)	15.0 (1.3)	8.1 (3.8)	73.1	0.98	95.0	26.50 / 3.00
633	7.8 (2.3)	13.1 (1.7)	9.0 (4.1)	68.9	0.92	81.0	22.25 / 2.75
612	7.0 (2.0)	11.3 (3.4)	8.8 (4.5)	59.8	0.69	66.7	23.00 / 3.25
604	4.3 (2.1)	6.7 (2.8)	4.0 (3.1)	54.9	0.50	81.0	14.50 / 2.00
632	6.6 (3.4)	10.2 (2.0)	9.3 (3.0)	53.1	0.70	53.8	26.50 / 3.50
629	5.5 (3.3)	8.3 (1.3)	7.0 (2.4)	50.2	0.60	76.2	13.25 / 1.75
627	6.3 (2.5)	8.9 (1.2)	6.8 (2.2)	41.5	0.66	95.0	28.00 / 3.25
607	5.8 (3.3)	6.9 (2.1)	3.7 (2.3)	18.9	0.32	71.4	10.00 / 2.00
608	5.9 (3.5)	6.9 (2.4)	6.5 (3.1)	17.8	0.39	52.4	5.25 / 1.00
620	6.6 (3.7)	6.1 (2.7)	8.7 (4.6)	‐6.4	0.10	33.3	1.75 / 0.75
Mean	6.4 (1.1)	9.6 (2.8)	6.8 (2.1)	48.6	0.62	74.6	20.52 / 2.56
No‐deposit Group
634	3.6 (1.7)	10.9 (3.3)	9.2 (3.2)	206.9	0.91	95.4	40.75 / 4.25
643	8.3 (3.7)	16.2 (1.2)	11.5 (2.3)	94.7	0.91	100.0	57.75 / 5.25
644	7.5 (3.3)	13.3 (3.9)	12.3 (7.1)	78.3	0.86	100.0	57.75 / 5.25
602	6.2 (6.4)	10.8 (6.6)	7.8 (4.5)	73.9	0.90	90.5	19.00 / 2.25
621	8.0 (3.3)	13.8 (2.7)	3.8 (2.5)	72.0	0.84	85.7	16.75 / 2.00
601	4.5 (3.4)	6.6 (3.2)	2.6 (1.1)	47.0	0.20	28.6	3.00 / 1.00
611	5.0 (2.3)	6.8 (1.2)	4.4 (1.8)	37.2	0.79	100.0	57.75 / 5.25
609	8.8 (3.6)	12.0 (3.8)	10.8 (1.5)	36.3	0.42	45.0	3.50 / 0.75
635	5.6 (2.9)	7.4 (4.0)	9.1 (3.2)	32.5	0.85	95.2	35.75 / 3.75
614	7.0 (2.5)	8.6 (3.1)	5.6 (1.7)	23.1	0.38	36.8	2.75 / 0.75
628	6.2 (2.7)	6.9 (2.9)	5.4 (2.9)	11.1	0.20	38.1	4.00 / 1.00
610	6.9 (1.5)	7.3 (2.5)	8.7 (2.0)	5.4	0.05	38.1	4.25 / 1.00
631	6.0 (3.1)	4.6 (2.2)	5.3 (2.6)	‐22.7	‐0.14	23.8	2.00 / 0.75
Mean	6.4 (1.5)	9.6 (3.5)	7.4 (3.1)	53.5	0.55	67.5	23.46 / 2.56

Note. Steps per day listed in thousands (e.g., 4.3 = 4,300 steps).

^*Tau‐U is an effect size measure that controls for baseline trend and measures nonoverlap between baseline and intervention phases; values closer to 1 indicate lesser overlap between the phases.

Some participants' physical activity levels increased to a larger extent than others. For example, Participant 605 met 100% of their goals and had a 74.4% change in step counts compared to the first baseline phase. Similarly, Participant 622 had a 77.3% increase in step counts. Overall, the effect size of the intervention for Participants 605 and 622 was moderate with Tau‐U values of 0.82 and 0.79, respectively. Additionally, both participants consistently recorded over 10,000 steps in the second half of the intervention phase.

Participant 629's data best reflect the average deposit group performance with a mean increase of 50.2% in step counts during the intervention. The effect size for this participant also reflected the mean of the group with a moderate effect (Tau‐U = 0.60). Likewise, their step count goals gradually increased, and they often met them (76.2%). In general, step counts for Participant 629 were often very close to the goal for any given day.

A few participants did not respond according to the walking contingencies: For example, step counts for Participants 607 and 608 increased during the intervention by only 18.9% and 17.8%, respectively. Tau‐U values of nonoverlap from baseline to intervention reflected this marginal increase, with a small effect (Tau‐U = 0.32 and 0.39, respectively). Lastly, step counts for Participant 620 remained variable throughout the entirety of the study and decreased from baseline to intervention (‐6.4%). This participant also had the lowest percentage of goals met in the group (33.3%).

No‐Deposit Group

Figure 2 displays daily step counts across the course of the study. Participants in the no‐deposit group responded comparably to those in the deposit group. Table 2 shows that mean percentage change from baseline was 53.5%, and also that participants in this group met their goals often (M = 67.5%). However, the mean percentage change and mean percentage of goals met for this group were more variable across individual participants, as compared to the deposit group. Like the deposit group, within‐participant step counts were generally less variable during intervention, and step counts decreased to initial baseline levels when incentives were removed.

Figure 2.

No‐Deposit Group Step Count

Note. Step count per day for each participant in the no‐deposit group in all phases (A‐B‐A) of the study, ranked from highest to lowest percentage change from baseline 1 to intervention. Each open data point indicates one day's step count, each dotted black line indicates the goal for each day, and the dashed gray line indicates the 10,000 steps/day recommendation. Breaks in participants' data points in the intervention phase indicate that a “free day” was taken.

As can be seen in both Figure 2 and Table 2, Participant 634's step counts increased the most from baseline in the no‐deposit group. Specifically, step counts for Participant 634 gradually increased throughout the course of intervention (though this increase was steeper at the start of intervention), resulting in a change from baseline of 206.9%. This participant also demonstrated a large effect size (Tau‐U = 0.91). The percentage of change from baseline was also generally high for Participants 643 (94.7%) and 644 (78.3%), who were both behind only Participant 634 in this regard. Additionally, Participants 643, 644, and 611 met 100% of their goals.

Although the data in the no‐deposit group were variable, Participant 621 best represented a medium‐to‐high performer in the no‐deposit group, in that they met 85.7% of their goals, and their Tau‐U demonstrated a medium‐to‐large effect size of 0.84. This participant had 3 days during the intervention when a goal was not met, and they maintained a steady step count after relatively high step counts during the first 6 days of intervention.

In contrast, although Participant 601's step counts increased by 47% during intervention, their goals were no longer met after day 9. This response pattern resulted in a small effect size of Tau‐U = 0.20. Overall, this participant had a low percentage of goals met throughout the intervention (28.6%), which was lower only for Participant 631 (described below).

Participant 631 was the lowest performing participant in the no‐deposit group. Baseline responding of this participant was highly variable; additionally, step counts were lower during intervention (‐22.7% of baseline). Tau‐U values of nonoverlap from baseline to intervention also reflected a negative effect (Tau‐U = ‐0.14; i.e., lower performance in intervention than baseline). This participant met 23.8% of their goals during the intervention, the majority of which were met during the first 3 days of intervention.

Group Comparisons

The percentage of goals met (Table 2) did not differ between the no‐deposit group (M = 67.5%) and the deposit group (M = 74.6%), t(64.91) = 1.20, p = .23. Percentage change in steps during intervention was also not significantly different compared to the first baseline across the deposit (M = 48.6%) and the no‐deposit groups (M = 53.5%), t(54.5) = ‐0.50, p = .62.

Figure 3 shows mean step counts for each group across the three phases. As shown, there was little difference in step counts between groups during each of the three phases. Effect sizes for the deposit and no‐deposit groups were similar, with mean Tau‐U values of 0.62 and 0.55, respectively. This indicates a small‐to‐medium effect of the intervention overall for both groups.

Figure 3.

Mean Step Counts for Deposit and No‐Deposit Groups across Phases

Note. Mean daily step counts for the deposit group (light gray bars) and no‐deposit group (dark gray bars) across each phase. Individual data points represent the mean step counts of individual participants in each group.

According to the mixed model linear regression, there was a significant effect of the intervention phase in both groups, B = 3185.61, t = 2.01, p < .05. Pairwise t‐test post‐hoc contrasts found that step counts in both groups during intervention (M = 9626.7, SD = 3096) were significantly higher than in the first baseline (M = 6421.3, SD = 1322.6; t = 3.45, p < .001) and second baseline (M = 7142.5, SD = 2621.3; t = 3.55, p < .001). There was no difference between the two baselines in both groups (t = 0.56, p = .39). Similarly, there was no significant difference across phases between the deposit group (M = 7411.1, SD = 2403.8) and the no‐deposit group (M = 7828.3, SD = 3073.4, B = 13.05, t = 0.01, p = .99).

Overall, this intervention cost researchers a total of $251.25 for 25 participants. The mean cost per participant was $10.05 for both groups. For each participant in the deposit group, participants earned $20.52, on average, but researchers gained $4.48 per participant because eight participants did not earn their deposit back. For the no‐deposit group, the mean amount earned per participant (and the mean cost to researchers) was $23.46. The amount earned was not significantly different between groups, t(64.29) = ‐0.68, p = .50.

Treatment Acceptability

Treatment acceptability was generally high for both groups. The overall intervention was rated with a mean of 7.0 (SD = 1.7) out of 10, and there was no significant difference in ratings between the deposit and no‐deposit groups (t(20) = 0.20, p = .84). Similarly, the use of the Fitbit was rated with a mean of 8.0 (SD = 1.9) out of 10, and there was no significant difference in the rating of using the Fitbit between the deposit and no‐deposit groups (t(23) = 0.18, p = .86). Goals and incentives were rated with means of 7.4 (SD = 1.8) and 8.7 (SD = 1.9), respectively, and there were no significant differences for the deposit and no‐deposit groups on ratings of goals (t(23) = 1.58, p = .13) or incentives (t(23) = 0.88, p = .39). Concerns about technology were minimal with a mean rating of 0.4 (SD = 0.6) out of 10, and there was no difference between the deposit and no‐deposit groups on concerns about using technology (t(23) = ‐0.55, p = .59). Finally, the deposit group rated the deposit as neutral, but not highly acceptable, with a mean of 5.0 (SD = 2.2).

Discussion

Overall, in the comparison of outcomes for the deposit versus no‐deposit group, treatment was successful for most participants, regardless of group. Eight out of 12 participants from the deposit group (622, 605, 618, 633, 612, 632, 629, 627) had a treatment effect characterized as Tau‐U greater than or equal to 0.6. Seven out of 13 participants in the no‐deposit group (634, 643, 644, 602, 621, 611, 635) also had a treatment effect. The remaining participants were categorized as nonresponsive to intervention. Eight participants either did not respond to the contingency at all (604, 620, 631) or had high levels of variability that made it difficult to determine the effect of the intervention (601, 607, 608, 609, 628). Finally, three participants in the no‐deposit group (610, 614, 628) initially met the treatment goals, but eventually missed goals and contacted the reset contingency multiple times. Following this, all three of these participants' step counts reduced to near baseline levels. It is possible that the reset contingency decreased the value of incentives such that they were no longer reinforcing. To ensure reengagement, future studies may consider returning incentive values to the highest amount previously achieved after a goal is met for a consecutive number of days, as others have done (e.g., Batchelder & Washington, 2021; Dallery et al., 2008; Raiff et al., 2016).

During the intervention phase, step counts increased relative to baseline for both groups, and there were no significant differences between groups in terms of the number of steps taken. This replicated the findings of Washington et al. (2016), in which there was a significant increase in steps taken during intervention relative to baseline, but the number of steps taken did not differ across the deposit and no‐deposit groups. Similarly, Dallery et al. (2008) found no difference in smoking reduction between deposit and no‐deposit groups. However, it should be noted that the current study was underpowered to detect between‐subject effects, with a 4% likelihood of detecting a significant effect across groups. Therefore, future studies should be conducted between larger groups of participants to definitively determine whether deposit contracts impact the effectiveness of contingency management interventions.

The current study demonstrated that it was feasible to reduce the cost of delivering a contingency management intervention for physical activity. Participants in the deposit group earned, on average, $20.52 throughout the study ($4.48 less than the deposit), resulting in a daily net gain of $0.21 for the researchers. Although money was not gained from the deposit contract group in the Washington et al. (2016) study, daily costs remained minimal ($0.48). In comparison, participants in the no‐deposit group in the present study earned, on average, $23.46 throughout the study, with costs averaging approximately $1.80 per day. This is similar to the average daily cost ($1.81) of the group that did not deposit money in the study by Washington et al. Further, daily costs were comparable to groups that had not deposited money in studies that sought to increase physical activity by prize draws (Washington et al., 2014; $1.80) and internet‐based incentives (Kurti & Dallery, 2013; $1.02–$2.44). However, it is important to note that direct comparisons of deposit and no‐deposit interventions were not made in those two studies.

In general, findings regarding deposit‐based interventions (as compared to no‐deposit interventions) are encouraging in light of the criticism that contingency management is expensive to implement. However, given that neither form of intervention has been shown to produce generalized behavior change (i.e., behavior returns to baseline levels when incentives are removed), there are still criticisms that have yet to be addressed (e.g., the view that contingency management is not a sustainable intervention). Nonetheless, interventions that feature deposits from participants or invested parties (e.g., insurance companies) may increase both the cost‐effectiveness and social validity of contingency management.

Overall, the treatment, goals, and Fitbit were rated as highly acceptable by participants in both groups. Additionally, the deposit group rated the deposit of their own money as neutral, but not highly preferred. These findings are somewhat similar to those of Halpern et al. (2015), who found that participants were more willing to participate in no‐deposit‐based interventions than in deposit‐based interventions. However, the deposit programs were more effective when the data of participants who were unwilling to participate in the deposit groups were removed. We attempted to control for this selection bias in the present study by including the possibility of assignment to the deposit group in the recruitment materials and telling participants that everyone had an equal chance of being in the deposit group. However, it is possible that some participants who were enrolled would not have participated, as occurred in the Halpern et al. study, if they had been assigned to the deposit group.

The difference in deposit amounts between Halpern et al. (2015; $150) and the current study ($25) is worth nothing, as this may have been a factor in participants' willingness to participate. Likewise, to further protect from study attrition, researchers may allow participants to determine their own deposit amounts (e.g., Jarvis & Dallery, 2017; Krebs & Nyein, 2021). Doing so may also result in increases in treatment enrollment in deposit contract interventions given that they may be more affordable across economic groups.

There were several logistical limitations of the present study. First, participants had to come to the lab once a week to sync their Fitbit and receive their earned money. Second, in the interim, researchers relied on participants' self‐reported step counts when adjusting goals. Researchers in future studies could allow participants to synchronize their Fitbits themselves each night and may also consider allowing participants access to their own graphs when doing so (e.g., Batchelder & Washington, 2021; Nastasi et al., 2022).

Another limitation of this study is the use of activity trackers to measure daily step counts. In some instances, the Fitbit malfunctioned (individual Fitbits malfunctioned a total of five times over the course of the study). It is also possible that participants were not wearing the Fitbit when step counts were being recorded (e.g., someone else was wearing it). Additionally, participants may have removed the Fitbit early in the day to keep their goals from increasing too rapidly. In follow‐up research, it would be beneficial to determine a reliable way to ensure that each participant is wearing their Fitbit and for the correct duration.

An additional limitation is that step count goals and incentives were confounded and were never arranged independently of each other; thus, it is difficult to interpret the effects of the intervention as being due to either the goals or incentives. The present study also had a short duration of the intervention (i.e., 3 weeks); this should be extended in the future to determine if it is associated with any changes in health outcomes related to physical activity including the maximum volume of oxygen used during exercise (i.e., VO₂), blood pressure, and glycemic control, which were not measured in the present study due to the short duration.

In summary, the present study offers a systematic replication of prior deposit contract comparisons (e.g., Dallery et al., 2008; Washington et al., 2016). Like these studies, there were no detected differences in efficacy between the deposit and no‐deposit groups. The present study expands the literature by demonstrating that escalating schedules are effective in conjunction with deposit contracts to improve physical activity. Similarly, our study addresses a common criticism of contingency management by being exceptionally cost‐effective for the deposit group, resulting in a net monetary gain for researchers.