Making policy decisions under uncertainty

Co-authored with Sir Paul Collier and Victoria Delbridge

The COVID-19 pandemic has exposed policymakers to high levels of uncertainty: governments simply do not have enough information to know what to do. The result is changing policy decisions mid-course: Britain reversed its strategy of building herd immunity when a new estimate put the likely death toll from this strategy at 250,000 people; America’s Centres for Disease Control initially discouraged the wearing of face masks by citizens but has since reversed its course. Policy changes like these have risked policymakers’ credibility, as well as their citizens’ compliance.

The uncertainty is far worse for developing country policymakers. Developed countries have more resources to spare if their decisions do not lead to desired outcomes. Trust in governments tends to be higher, and data is also better. While policymakers in developed countries have benefited from sophisticated analysis based on data shared by large firms, the best bet in developing countries are small surveys, or projections which typically use mortality data from elsewhere and assumptions to provide local estimates. Even core health data, such as the rate of infections, might be significantly understated: across Africa, only 685 tests have been carried out per million people, while Italy has conducted nearly 37,000 per million. If policymakers in developed countries are making decisions in fog, their counterparts in developing countries are doing so in the dark.

For policymakers at the city level, decision making is even harder. While the spread of the pandemic is deeply spatial by nature, local governments often lack the authority to act to contain the virus. For cities that are led by parties in opposition to those at the national level, authority is prone to re-centralisation. When Delhi’s local government attempted to reserve hospital beds exclusively for its residents, the decision was quickly overruled by the central government-appointed governor. Data sharing is also a challenge: even more tricky than sharing between various national government departments is sharing with sub-national tiers.

Making policy decisions under this level of uncertainty requires four key steps:

Distinguish between questions that can be answered without further information and those which cannot.

Some policies make sense no matter what new information comes in — what economist Stefan Dercon calls no-regret policies. As Michael Callen and Edward Glaeser also suggest, encouraging hygiene is one such policy: across sub-Saharan Africa, only a quarter of people have access to basic handwashing facilities including soap and water. These investments are important regardless of the pandemic, but more so given the increased importance of handwashing to limit the spread of the COVID-19 virus.

Other policies involve significant trade-offs — costs and benefits — which are currently poorly measured. For these, the first policy response is to gather information that would enable a decision to be taken, as far as possible. Since capacity is limited, all such efforts should be driven by the specific policy choices that need to be made. For example, through combining easy-to-access administrative datasets with a pre-existing socio-economic survey, policymakers in Pakistan have been able to target cash transfers to the poorest. Cape Town uses live data from funeral homes to quickly identify areas with excessive deaths in the city. Where possible, technology can also help: Medellín asks residents who need financial support to register their vulnerability, as well as their symptoms, online — nearly 90 percent of the residents have.

Some questions cannot be answered at all at present: for example, how long will the virus continue to spread? When will it end? How many people will lose their lives? — all of these can only be known ex-post. Until then, while we may be able to guess, ‘we don’t know’ is the only right answer.

Uncertainty requires experimentation: it is learning from doing.

When ‘we don’t know’ is the only right answer, it is unlikely that our first attempt at solving a problem will be correct. In this case, the most appropriate response is to use the best information available and quickly learn from what is working and what isn’t.

This requires experimentation — varying activities, locations, and time periods. As the costs and benefits of lockdowns is unclear, our understanding of lockdowns can benefit from such localised experimentation. On the one hand, China’s decision to lockdown the city of Wuhan may have reduced the infection rate in other Chinese cities by about65%. On the other hand, emerging evidence shows that lockdowns in developing countries lead to significant income losses and reduced food consumption. In these cases, targeted lockdowns that impose restrictions based on local risk profiles might work better.

With experimentation, we need to rapidly monitor, evaluate, and adapt. When Madrid shut down schools and offices and asked people to stay home, many people instead congregated at parks. Only after seeing this did the city adapt and closed its parks. The city of Durban has daily “war room” meetings to undertake and evaluate decisions in the city.

Models can help, but they are no more reliable than the numbers fed into them. The epidemiological model being used by the American government is repeatedly wrong.  — Vox has reported that the actual death numbers fell outside the range it predicted 70 percent of the time. While other models have been more accurate, their importance lies in understanding what they can do: provide a range of outcomes likely under certain conditions., their importance lies in understanding what they can do: provide a range of outcomes likely under certain conditions. As more data is collected and fed into models and assumptions are tweaked based on real-world evidence, models become better. Until we have that data, they are best used with caution.

Communicate with citizens so that they can make good choices.

Policymakers need to communicate two big messages:

  • There are serious risks associated with COVID-19, and that by their own actions, people can reduce these risks to themselves and to others.
  • Policymakers do not yet have enough evidence, and so may well need to revise policies as they learn more: people should stay alert to further advice on what to do.

The authorities must balance the need to reassure — which is very important to avoid panic — with the need to avoid losing credibility as events unfold in unexpected ways. We know from past shocks, such as from the Ebola epidemic in West Africa, that misinformation can have a disastrous impact on containment efforts. We are seeing some of this play out right now: across cities in Pakistan, rumours and misinformation have fuelled incidentsof patients’ families attacking hospital staff.

More worryingly, some countries have outright stopped sharing information with people: Tanzania stopped sharing data on the spread of COVID-19 six weeks ago and has just announced that the country is free of any cases, even though border testing of Tanzanian truck drivers by Uganda indicates otherwise. Countries which have clear and consistent information are doing better. Vietnam stands out: the government communicates protection measures via text messages, and posters across Vietnamese cities highlight the seriousness of the virus — so far, Vietnam has prevented any large-scale community outbreak and the death toll is at zero.

The messenger is as important as the message.

As far as possible, decisions need to be taken close to the communities and people who understand their specific constraints. In South Sudan, researchers note that localised priorities are competing with the threat from coronavirus, and the messaging needs to be tailored to account for these other challenges. In China, devolvement went even further, with some cities actively recruiting their residents into containment measures. Guangzhou alone hired as many as 80,000 of its residents to conduct community patrols to ensure compliance.

Working with local community leaders can also help — as Freetown Mayor Aki-Sawyerr put it, “it is not only the message, but also the messenger”. By working with local allies, policymakers can leverage greater trust in their decisions. During the Ebola epidemic in Liberia, door-to-door community outreach with community leaders was incredibly effective in increasing the uptake of preventative measures. Not working with community leaders can undermine containment efforts: in the Punjab province of Pakistan, nearly half of all urban mosques surveyed still conducted congregational Friday prayers, despite government asking them not to. But when the surveyors provided them with information on the pandemic, around 8 out of 10 mosque leaders said they would consider postponing congregational prayers.

Making policy decisions under this level of uncertainty is not easy. The best we can do is to restructure the way we think about it: recognise the limitation of the available evidence and experiment with different policy decisions, actively learn from the outcome, devolve decisions to local governments who best understand the context, and clearly communicate the information to the people.

Published by the International Growth Center on 23rd June, 2020.

Categorized as Featured

Devolve more power to cities: they will need it more than ever

Co-authored with Astrid Haas.

COVID-19 pandemic started in a city and spread through them. The core benefit of cities — as an enabler of proximity — is also their biggest liability. This is exacerbated when infrastructure and services are scarce. Across sub-Saharan Africa, only a quarter of people have access to necessary handwashing facilities, including soap and water. More than a third of Indian urbanites are packed into a single room or a house without a roof.

Yet some cities were better prepared to respond to this pandemic than even their own national governments. In Colombia, where city leaders have sufficient power to make decisions, many acted quickly and have been able to tailor policy decisions to the situation at hand. Medellín’s mayor began preparing for the virus in late January, far before the national government. Bogotá’s mayor created an 84 km emergency bike network to help essential workers get around safely.

A defining difference in how well cities are responding to COVID-19 is whether city governments actually have the power to respond. Many cities, particularly in developing countries, don’t because national governments have often been reluctant to devolve power to them. This may be for a number of reasons such as politics (cities can be opposition strongholds) — or perhaps due to economics (they are important sources of tax revenue). Yet cities are complex systems, and they need context-specific policies to work effectively.

This reluctance to devolve power affects how well cities can be managed: given the complexity of urban systems, cities need context-specific policies to work effectively. The pandemic is a case in point: the initial response across the globe was to impose blanket national-level lockdowns. However, there is increasing evidence that they are inefficient and economically very costly. Instead, localised containment measures that can be implemented and lifted by responsive authorities are proving to be more flexible, efficient and effective.

But localised containment measures require local leaders who to have adequate control. In Britain, the ongoing stand-off between the national government and the mayor of Greater Manchester is a stark example of how the central government’s desire to impose local containment measures can fail without local leaders willing to lend it legitimacy. As our LSE colleague, Tony Travers, said: “There is a sense that people in Greater Manchester are having things done to them by a distant government.” This might have been avoided if the Mayor was given more control over the city’s pandemic response and resource-raising powers.

For many countries, the power over cities may have been devolved on paper, but very rarely in practice. Some of this is historical: many cities in developing countries inherited centralised urban governance structures and laws from their colonial pasts. These are ill-suited to their current needs. For example, Britain’s Town and Country Planning Acts still cast a long shadow over many African and South Asian cities in the form of stringent height regulations and large, pre-defined plot sizes for land. This encourages sprawling, low-density cities where people struggle to travel between their homes and workplaces.

Reluctance to devolve power has also at times resulted in incomplete and inconsistent outcomes. Cities are often only given partial control over urban policies or have their administrative jurisdiction siloed. In Karachi, Pakistan, the mayor barely has jurisdiction over a third of the city. The rest is split between provincial and federal governments, while three different political parties control each tier.

National governments can and should change this. Across history and across the globe, cities have given individuals the quickest route from poverty to prosperity: in Africa, one can earn as much as 23% more in cities than in rural areas. Cities provide opportunities that attract people and create ‘agglomeration’ economies that cluster people and businesses together, which is attractive to the private sector. The outcomes of this proximity can be innovation, economic growth, and development.

But not all cities lead to this outcome: it is only well-managed cities. This requires giving cities control over key areas that shape the path of urbanisation. At a city-level, all these policy areas, such has housing, transport, and planning, complement each other and require context-specific approaches. For example, new transport investments need to consider where people live and work to be most effective.

Devolving control over policy areas also needs to be coupled with increasing the ability of city leaders to raise their own revenues. This is not only important to allow them to deliver local services and infrastructure effectively, it also reduces the burden on national treasuries. Perhaps most importantly, it increases the accountability of city leadership to their citizens.

In 1960, no low-income country had more than one-third of its population living in its cities: now nearly 40% of them do. This trend is projected to continue well into this century, with Africa and South Asia urbanising rapidly. Furthermore, as more people are pushed into poverty due to the pandemic, we are likely to see more rapid urbanisation rates on these continents, as people move towards cities looking for opportunities.

Cities need to not only meet the expectations of the current citizenry, they also need to plan for the people to come. As shocks — health, climate, and others — are likely to increase, city leaders need the power to prepare and respond. Luckily for Bogotá, their mayor has this control and is now planning to add 280 additional kilometers of bike lanes, paving the way for a more resilient city. Other cities in rapidly urbanising countries need more of this local control as well.

Editor’s note: This article first appeared on LSE blog.

Categorized as Featured

The case for taxing landlords in Pakistan

A replica of the Statue of Liberty stands on a hill overlooking the construction of new homes in Bahria Town on the outskirts of Islamabad, Pakistan March 16, 2016. To match Interview PAKISTAN-PROPERTY/TYCOON REUTERS/Caren Firouz

BUY land for they aren’t making it anymore, Mark Twain once advised. Our politicians certainly seem to agree — some have over a dozen parcels of land to their name, according to recent asset declarations published by the Election Commission.

None of this is a surprise. Pakistan’s rich (like the rest of us) like to buy land because it is a tangible and easily definable asset, instead of company stocks or government bonds — you can see land, touch it, build upon it. It gives you social status and helps you keep it. It also tends to go up in value by a lot: on average, Lahore’s residential plot increased its market value by 85 per cent between 2013 and 2018.

Better yet, once you own land, you don’t have to do anything for its value to go up. The majority of the value of properties comes from its land. Once the owner of a piece of land, one benefits from other people’s work: people who run businesses; those whose taxes allow governments to invest in improving infrastructure. All of these people work, and they drive up the value of land around them. Those who own the land benefit without having to lift a finger.

Hence, the fair thing to do is to tax people who own land, so that the landlords pay for the infrastructure they benefit from. It is also the smart thing to do so the government can use this tax revenue and invest in better infrastructure that drives up the value of land and paves the way for more property tax revenue in the long run. The best way to do this is by enforcing an annual tax on property that owners linked to the land’s market value, so as the land value goes up, a portion of it is actively captured to pay for public services.

As most fair and smart things go, we don’t do this well right now. All of Punjab, with its 100 million-plus population, collects less urban property taxes than the city of Chennai, which is home to about 10m people.

This is because our current system allows landlords to pay little regular taxes on their properties. This is by design: the tax method used underestimates property values. An owner of a one-kanal (approximately 500 square metres) house in a posh area in Lahore or Karachi rarely has to worry about an annual tax bill above a few thousand rupees.

By not taxing land enough, we have designed a system in which people buy land to increase their wealth rather than buy land for housing. This is why parts of our cities are left vacant, as landlords wait for their plots to increase in value. Those who aren’t lucky enough to own land have to move to the city’s outskirts to afford a house.

How do we change this? If the government is willing, there is enough technical expertise to design a fairer tax system that actively targets the richest landlords. But landlords are a powerful lobby, especially if they’re the ones who dominate our legislatures. Any such reforms will depend on the reformer’s ability to balance winners (the people at large) and the losers (the landlords).

An excellent way to do this is to link any such expansion of taxation with local service delivery led by independent urban governments. This would help create a clear coalition of beneficiaries from the expansion of property taxes in every city. But currently, we don’t have empowered local governments: Punjab’s decentralisation bill would pave the way for such governments in the province, but elections are repeatedly delayed.

For such local governments to properly tax wealth kept in the form of land, we also need to change the special provisions given to parts of the city — such as the cantonment areas and Defence Housing Authorities where the highest-value land is located and they collect property taxes independently of local and provincial governments. As residents of these areas benefit from the city’s larger economic system, it is only fair that they contribute towards city-wide public investment.

Such reforms are possible. Consider: Freetown’s mayor revamped her city’s property tax system earlier this year. Her message is simple: we need to pay for public services, and it is only fair that the richest landlords pay their fair share. The taxes on the top 20pc of the landlords will triple, while taxes on the poorest will be halved. The city expects its property tax revenue to increase five-fold.

Until we do the same, our legislators-cum-landlords will continue to benefit from unearned profits while accumulating larger landholdings. Aspiring landlords will buy plots in shady housing societies. Our cities will continue to struggle to pay for essential services. As for those who want to buy land to build a house — they will have to wait.

Published in Dawn, November 24th, 2020

Image: Dawn.

Categorized as Featured

What prevents developing countries from taxing more?

Published: International Growth Centre, 12th Aug 2020.

Virtually every country in the world taxes their populations. But, some do so more successfully than others. Most developing countries raise tax revenue equivalent to between 10% to 20% of their GDPs, some even less. Rich countries raise much more, on average 34%. If Pakistan distributed all its annual tax revenue equally among its 200 million citizens each would get less than 1% of the country’s legal minimum wage. The result of this is considerable differences between the abilities of governments to fund public services, explaining why India spends $62 per citizen on healthcare, while Germany spends over $4,000.[1]

Why do developing countries find it so hard to tax? There is no one answer. For starters, developing countries are information-scarce environments – governments typically do not know who owes how much. This is why when developing countries do tax, they focus on taxing consumption and trade instead of labour, as the former are easier to target.[1]

In most developed countries, this information is largely created by third-party reporting of income. Firms report how much they pay their employees, making it harder for people to cheat. In most developing countries, large informal sectors make this nearly impossible to achieve. Even in China, where formal employment is more widespread, only 28 million people pay direct income taxes, as of 2015. In Pakistan, it is about one million people.

Increasing income taxation is hard without widespread formal employment. In a 2019 working paper, Anders Jensen, of Harvard University, uses data from the United States to show that as people moved away from self-employment to wage labour, this was associated with increased income taxation usage. In a similar vein, research from Denmark shows that tax evasion is significantly higher among people who self-report their income than those whose income is reported by employers.

Some developing countries have made some progress on this front by targeting the small number of people who work in formal firms. This is largely done using withholding taxes, in which businesses deduct the tax before giving paychecks to their employees. But doing so risks pushing the administrative burden on formal private firms — a bad precedent to set, especially when so few exist.

The expansion of value-added taxation (VAT), under which the government levy tax on each firm across a supply chain, is also in part motivated by the desire to create more information. To calculate the value, each firm needs to subtract the difference between its output and its input. By doing so, they create an information trail that can verify the tax liabilities of other firms in that chain. 8 out of 10 countries in sub-Saharan Africa levy the VAT today.

It is ingenious, on paper. In practice, it has led to mixed results. In Chile, third-party reporting due to the VAT has strongly disincentivised tax evasion. While researchers in Uganda have found widespread discrepancies in amounts reported by sellers and buyers, despite the paper trail created by the VAT.

The secret might be how much enforcement capacity a given state has. Firms may eventually realise that the state does not do anything if they do not accurately report their profits. After all, information is worth only if someone acts upon it. But, capacity is a whole other beast: you need the capacity to raise revenue, and revenue to raise capacity.

One way to address this problem is to motivate tax bureaucrats to perform better. When researchers in Pakistan provided property tax collectors with performance pay schemes, revenue jumped by as much as 46%. But, incentives can be tricky to tweak. Some of them can also be costly: in Ghana, one out of five revenue collectors earn a monthly salary that is greater than the revenues they collect, giving them bonuses might not help nor be feasible for the government coffers.

Given all these limitations, developing countries may need to think about a different set of tax policies, such as those we find unwise for developed countries. Economists Adnan Khan and Henrik Kleven term such policies as ‘third-best’.

One policy they consider is taxing full firm revenue instead of profit. Doing so is inefficient because it does not allow firms to deduct the costs of input and may distort a firm’s production decisions. But some inefficiency maybe tolerable if it raises enough revenue. Researchers estimate that in Pakistan’s case this could raise as much as 74% more tax revenue.

There are other factors: less technocrat, more political. An important one is the prevalence of double tax treaties between countries. Signed to prevent taxing the same income twice, they have undermined the tax capacity of several developing countries. Firms can base their operations in one country, usually that has a treaty with another country and very low taxes, evading taxes in the latter because they pay in the former.

In a 2018 paper, IMF Economists Sebastian Beer and Jan Loeprick have used the fact 11 African countries have such treaties with Mauritius, where firms are effectively taxed at a rate of just 3%. They find that these treaties have not only not led to increased foreign investment in countries that sign these treaties with Mauritius but also lead to significant revenue losses. Zambia has just torn up its tax treaty with Mauritius.

The cross broader implications do not end here. The elite in many developing countries evade taxes by moving their wealth abroad, supported by a network of financial and legal firms. African governments collectively lose about $15 billion a year because of offshore wealth, calculations by economist Gabriel Zucman show. It is hard to tax when much of the tax base is signed away, or hidden in Swiss bank accounts.

Published under the title “What is stopping developing countries from taxing more?” on as part of IGC’s Ideas Matter campaign to celebrate the launch of the Little Book of Growth.

[1] This moves the tax burden to lower-income groups. However, in new research, Bachas et al. (2020) argue that consumption taxes might be more progressive in countries with large informal sectors, because people in the bottom decile tend to rely more on informal transactions which are not taxed.

[1] Data by World Bank’s WDI database. PPP dollars.

Categorized as Featured