A common feature of most low-income countries is a government that collects little tax revenue and provides few public goods. The state’s capacity and development literature argues that the reason for the inability to collect taxes efficiently is why low-income countries are as poor as they are (Besley and Persson 2009). This study suggests that the path to economic growth for low-income countries may begin by investing in the government’s ability to raise tax revenue (Albers et al. 2020), to provide productivity-enhancing public goods. Several theories argue that tax administration technology investment is at the heart of increasing government size, including due to improved collection process efficiency (Brennan and Buchanan 1980, Baker and Mulligan 2003, Cohen 2021). Fan et al. (2018) provides a recent study in China on computerized VAT.
In this context, our new research (Dzansi et al. 2022) provides descriptive and experimental evidence on the role of technology in improving government tax efficiency. The setting is the local government of Ghana, which oversees the collection of property taxes but actually collects very little (Ghana Government 2014). The technology in question consists of a geospatial database of features embedded in an electronic tablet with GPS capability. The adoption of similar technologies in developing countries has increased significantly in the last decade (Fish and Prichard 2017).
To measure the potential of technology to overcome power constraints, we first conducted a census of local governments. We conducted the census in the fall of 2017 in collaboration with several national ministries and 216 local governments of the country. The purpose of the census was to gather information on each relevant level of the local tax collection system of each local government. Respondents were interviewed in three main sets: local government officials, locally elected members of parliament, and citizens. In addition to the survey data, we have digitized and adapted administrative records to measure all sources of local tax collection and all types of government spending in each district.
The census data highlights how poor tax collection infrastructure – including limited street naming and property addresses – shapes collection practices in most areas. Almost all bills are delivered to taxpayers and collectors usually go with individual taxpayers more than once before collecting (if they collect at all). Most taxes are paid in cash and are paid directly to revenue collectors. Not surprisingly, government officials have cited tax evasion by tax collectors as a significant limitation of their revenue. In this challenging context, 17% of local governments have chosen to adopt technology to assist in tax collection in the form of revenue management software and electronic database of property. We see that this minority of local governments who have decided to invest in technology have significantly better results at every stage of the tax collection process. In particular, they provide more bills, collect more revenue, and lower non-payment rates than non-technological governments.
The rigorous empirical link between the use of technology in local government cross-sections and the results of tax collection naturally invites questions about the rationale. To address this issue, we have partnered with a large Ghanaian municipal government and a private technology company to randomize its use of technology within the government’s jurisdiction. Specifically, we’ve shuffled the use of a new revenue collection software and property geospatial database at the revenue collector level. In the trial, both the treatment and the control collector were given a stack of about 135 bills of about 135 bills in a randomly assigned area and were tasked with collecting as much revenue as possible within six weeks. The treatment group was given an electronic tablet that used geospatial data to facilitate family location (see Figure 1). Otherwise, the two groups of collectors and their designated areas were similar in terms of observation.
Comments: This graph shows the average amount of property tax collected by the revenue collector, separately on the day of intervention and treatment appointment.
Revenue collectors using the new technology paid 27% more than control collectors at the end of the study. We see this result as a reflection of the mechanical convenience that technology allows taxpayers to identify more efficiently in an environment with a shorter property address. The series shows a concave pattern during the rising bill distributed in both groups, as collectors insist on follow-up with families who have already served a bill to collect money from them from bill distribution. Revenue collection was 103% higher among collectors assigned to the technology group, on average, implying a much greater impact on revenue collection than bill distribution. Furthermore, we see that the effect of treatment on the collection increases over time, leading to an increasing average effect on the amount collected per bill distributed through the experiment (Figure 2).
Comments: These images show the navigation support provided by the tablet, which is used by revenue collectors in the treatment group.
We explore several possible hypotheses as to why the impact of treatment on the collection is so greater than the effect of treatment on bill delivery. A simple story is that families have a different attitude towards paying when inspecting by collectors shown with technology than ‘sustainability’ collectors without technology. Yet families in the treatment and control areas have been surveyed just after the test report for statistically similar integrity and the ability to enforce tax-paying powers among local government officials. A second hypothesis is that the technology helps reduce leaks – for example, in the form of family payments but removed by revenue collectors before they reach the local government treasury. However, a variety of family survey questions about the predominance of bribery indicate that the activity of bribery in treatment is less – not less – moderately higher than in the control area.
We argue that the most promising process is that the technology allows collectors to learn about – and focus on the times when they are in short supply – households that are more likely to make tax payments. Using a survey of collectors’ behaviors and strategies, we show that treatment collectors have a better understanding of individual family payments over time and that families who are more able to pay, are more aware of tax payments, and more satisfied are collecting from them. Reports with more focus on. Local public products (Figure 3). Importantly, none of these family characteristics were known to collectors at the beginning of the study period. This means that technology allows collectors to learn about family features through repeated visits (or long visits) that are otherwise difficult to observe. In line with this notion, we document that families with higher liquidity and higher incomes – which previously disappeared from collectors – are more likely to be targeted by the treatment group than the control group. We formally transform this differential learning mechanism into a simple dynamic Bakerian time use model where visionary revenue collectors maximize incremental revenue collection for each time period.
Comments: This graph shows how many collectors know where families are located who are more able to pay property taxes, separately through survey waves (baseline, midline, endline) and treatment assignments. The gray bar measures the difference in knowledge between treatment and control
Advanced learning through technology has important distributive effects. Increased information on household income collected by medical collectors, and subsequently targeting high-income families makes the local tax system more progressive. In particular, technology increases tax payments as part of tax arrears in the top quarter of income-wealth distribution, but keeps tax payments unchanged in the bottom quarter. However, enhanced information seems to be a double-edged sword, as technology also increases the incidence of bribery, the effects of which are concentrated in the lower quarters. Additional analyzes suggest that treatment collectors also learn about families involved in bribery and target later. This is consistent with our preferred interpretation of how technology facilitates learning about which families are more likely to make tax payments.
Our experimental results on technology investments shed light on the promise and disadvantages of using technology to build tax power, and society’s aspirations must balance the positive and progressive tax effects against the effects of recurring bribes. Our results suggest that the positive effects of technology on tax results are only partly due to the presence of electronic devices embedded with geospatial data. The technology allows collectors to overcome learning limitations (in this case, arising from navigation challenges) which limits their ability to generate information about taxpayers’ tendency to pay. Our results therefore relate to papers that show how pre-existing data sources from third parties can be used to improve collection (Cleven et al. 2011, Pomeranz 2015, Naritomy 2019, Balan et al. 2022). Most previous studies place third party information at the center of the government’s informational power (Gordon and Li 2009, Kleven et al. 2016); Our work shows how, in settings where such sources of information do not exist, the state can still strengthen its informational capacity by generating direct information about taxpayers’ payments.
Albers, T, M Jerven and M Suesse (2020), “On the Development of Fiscal Capacity: New Insights from African Data”, Vox.EU.org, 22 November.
Balan, P, A Bergeron, G Tourek and JL Weigel (2022), “Local Elite As State Capacity: How City Chiefs Use Local Information to Increase Tax Compliance in DR Congo”, American Economic Review 112 (3): 762-97.
Baker, G. and Mulligan (2003), “Deadweight Costs and Government Size”, Journal of Law and Economics 46 (2): 293-340.
Basel, T. and T. Parsons (2009), “Sources of State Power: Property Rights, Taxation, and Politics”, American Economic Review 99 (4): 1218-44.
Brennan, G. and J. Buchanan (1980), Ability to taxCambridge University Press.
Cowen, T (2021), “Does technology drive government growth?”, In Articles on government growth, Springer.
Dzansi, J, A Jensen, D Lagakos and H Telli (2022), “Technology and Local State Power: Evidence from Ghana”, NBER Working Paper 29923.
Fan, H. Y. Liu, N. Qian and J. Wen (2018), “The Dynamic Impact of Computerized VAT Invoices on Chinese Manufacturing Companies”, VoxEU.org, July 29.
Gordon, R. O. W. Lee (2009), “Tax Structures in Developing Countries: Many Puzzles and a Possible Explanation”, Journal of Public Economics 93 (7-8): 855-866.
Government of Ghana (2014), “Internally Generated Revenue Strategies and Guidelines: Maximizing Internally Generated Revenue Possibilities for Providing Improved Local Level Services”, Ghana, Ministry of Finance.
Fish, P. and W. Pritchard (2017), “Strengthening the IT System for Property Tax Reform”, African Property Tax Initiative Brief.
Kleven, HJ, MB Knudsen, CT Kreiner, S Pedersen and E Saez (2011), “Unintentional or unable to cheat? Evidence from tax audit in Denmark “, Econometrics 79 (3): 651-692.
Naritomi, J (2019), “Consumers as Tax Auditors,” American Economic Review 109 (9): 3031-72.
Pomeranz, D (2015), “No Taxes Without Information: Barriers to Value Added Taxes and Self-Introduction”, American Economic Review 105 (8): 2539-69.