2023 Digital Divide Index (DDI)

The digital divide is the number one threat to community economic development in the 21st century. Public policy 101 argues that a problem needs to be defined before exploring potential solutions.

The Digital Divide Index or DDI ranges in value from 0 to 100, where 100 indicates the highest digital divide. It is composed of two scores, also ranging from 0 to 100: the infrastructure/adoption (INFA) score and the socioeconomic (SE) score.

The DDI uses the Speedtest® by Ookla® Global Fixed and Mobile Network Performance Maps available from   https://registry.opendata.aws/speedtest-global-performance.   Ookla trademarks used under license and reprinted with permission.

The INFA score groups four variables related to broadband infrastructure and adoption: (1) percent of homes without a computing device (desktops, laptops, smartphones, tablets, etc.); (2) percent of homes with no internet access (have no internet subscription, including cellular data plans or dial-up); and (3) average download and (4) average upload speeds in Megabits per second (Mbps) weighted by number of speed tests based Ookla Speedtest® open dataset. Please note that 2017 & 2018 data uses median advertised speeds due to data availability. 

The SE score groups five variables known to impact technology adoption: (1) percent population ages 65 and over; (2) percent population 25 and over with less than high school; (3) individual poverty rate; (4) percent of noninstitutionalized civilian population with a disability: and (5) a digital inequality or internet income ratio measure (IIR). In other words, these variables indirectly measure adoption since they are potential predictors of lagging technology adoption or reinforcing existing inequalities that also affect adoption.

These two scores are combined to calculate the overall DDI score. If a particular county or census tract has a higher INFA score versus a SE score, efforts should be made to improve broadband infrastructure. If on the other hand, a particular geography has a higher SE score versus an INFA score, efforts should be made to increase digital literacy and exposure to the technology’s benefits.

The DDI measures primarily physical access/adoption and socioeconomic characteristics that may limit motivation, skills, and usage. Due to data limitations, it was designed as a descriptive and pragmatic tool and is not intended to be comprehensive. Rather it should help initiate important discussions among community leaders and residents.

Legend Each variable is mapped into three equal quantile ranges and are measured as Low, Moderate, or High.

Dear Digital Divide Index (DDI) user: We are proud to add this new section called DDI trends. This section provides a state-level break down of DDI scores and underlying variables for 2019, 2020, 2021, 2022, and 2023. Unfortunately, previous years are not compatible with the data used for those DDI. 

Each state profile showcases annual numbers and also provides the difference between 2022 (post COVID) and 2019 (pre COVID). The intent is for stakeholders, residents, and leaders to quickly see which variables moved in what direction and which scores increased (bad) or decreased (good). 

Please select your state from the map below and download the PDF for your state. As always, reach out if you have any questions. Also, for a cost-recovery fee (except for Indiana), these are available in the  Rural Indiana Stats website  , we can also generate county-level DDI trend profiles. 

Data for the digital divide index (DDI) was obtained from the 5-year American Community Survey (ACS) and Ookla Speedtest® open dataset. Since Ookla Speedtest® open dataset data is available in “quads” on a quarterly basis, these were average weighted by number of tests to obtain annual variables and geocoded into Census tracts and counties. This data was then matched to the ACS Census tract and county data.

The DDI consists of two components. The first one is the infrastructure/adoption component (INFA) that includes average download (DNS) and upload (UPS) speeds, percent of homes without internet access or not subscribing (NIA), and percent of homes with no computing devices (NCD).

Within the INFA component, more weight was given to broadband access (0.35) and computing devices (0.35) than to download (0.15) and upload (0.15) speeds. Although speeds are becoming more important, access and adoption precede speed and, thus, deserve more weight. Likewise, the download/upload speed range is more susceptible to extreme outliers, so assigning less weight helps minimize the impact of extreme outliers. Important to clarify is that multiple weight combinations were utilized and the results did not change drastically.

The second component is the socioeconomic (SE) component that includes the percent of the population age 65 and over (AGE65), percentage 25 and over with less than a high school degree (LTHS), individual poverty rate (POV), percent of the noninstitutionalized civilian population with any disability (DIS), and a digital inequality or internet income ratio measure (IIR). This measure is the ratio between the share of homes making less than $35,000 per year without internet and the share of homes making $75,000 or more per year without internet access. A higher ratio indicates greater internet access inequality between wealthier and lower-income homes. Equal weight was given to all indicators within this component.

Keep in mind that these socioeconomic variables indirectly measure adoption since they can be considered as potential predictors of lagging technology adoption or of inequities. If a particular county scores high on these variables but low on broadband infrastructure, it may be better to focus on digital literacy and promoting the personal benefits of the technology.

Because these variables have different units and normal distributions, z-scores were calculated for each variable and geography. Z-scores standardize the data and indicate where a particular observation fall compared to the mean and standard deviation of the sample. Please note that these scores were calculated by looking at the geographic units (Census tracts, counties) and comparing them with their peers. For this reason, scores are not comparable across different geography tiers (Census tract versus counties versus states).

Since the DDI was designed to show a larger digital divide as the score increases, careful attention was paid to the signs in equations 1 and 2. The rationale behind the infrastructure/adoption (INFA) score (equation 1) was: as the z-scores of the percent of population without fixed 100/20 (NBBND), no internet access (NIA), and no computing devices (NCD) increases (+), the divide increases; while the z-scores of the median download (DNS) and upload (UPS) speeds increase, the digital divide decreases (-).

A similar rationale was used to calculate the socioeconomic score (SE) in equation 2: as the z-scores of the percent population ages 65 and over (AGE65) increases (+), so does the potential lag in technology adoption; same as the z-scores of individual poverty rate (POV) increases (+), percent population 25 and over without a high school degree (LTHS) increases (+), and percent non institutionalized population with any disability (DIS) increases (+), so does the digital divide.

• Equation 1: INFA = NIA*0.35 + NCD*0.35 – DNS*0.15 – UPS*0.15

• Equation 2: SE = AGE65 + POV + LTHS + DIS

Notice however that the SE components are given equal weight while INFA components are not. This may result in more variance in the SE score compared to the INFA score. This in turn gives SE more influence on the DDI score compared to the INFA score. For this reason, z-scores of the INFA and SE scores were calculated and then added up to calculate the final DDI score giving both components equal influence as shown in equation 3. 

• Equation 3: DDI = INFA + SE

All scores were normalized to fall between a 0 to 100 range where the higher the number, the higher the digital divide.

Worth mentioning is that two important variables are lacking in the DDI: broadband cost and how the technology is being used. Without a doubt, these two variables would strengthen the DDI, but, unfortunately, nationwide data is not available.

On the other hand, access to cellular wireless was not included because most of the benefits of digital applications are undermined by mobile devices and limited data plans. It is much harder to complete a job application or complete a homework assignment using a smartphone that also has limited data. As broadband applications become more sophisticated and require more data, limited data plans undermine usage and can become very expensive.

DDI citation information

Gallardo, R. (2025). Digital Divide Index. Purdue Center for Regional Development. Retrieved from Digital Divide Index (DDI):  http://pcrd.purdue.edu/ddi 

Researchers: dataset is available. Please see contact information below if interested in the dataset.

• Roberto Gallardo, PhD
•  robertog@purdue.edu