MODELLING DIRECT AND INDIRECT EFFECTS ON LIFE EXPECTANCY

A large body of economics literature on Covid-19 has been developed since the outbreak began. Much of it focuses on modelling the number of Covid-19 cases (Acemoglu et al. 2020; Manski and Molinari 2020), the number of deaths (Homburg 2020; Gibson, forthcoming) or the years of life lost (Decerf et al. 2020). Much less attention has been paid to the impacts on life expectancy. We argue that considering these impacts has at least three benefits:

1. Life expectancy, as a measure of the average lifespan that can be expected, is intuitively understood by most people.

2. Life expectancy can be used to summarise the impact on everyone in Indonesia and is naturally weighted in the sense that the people who will suffer the longest from the effects of the Covid-19 economic crisis (the young) contribute more to the average value of life expectancy.

3. Life expectancy is affected by both health shocks and income shocks, so it naturally allows analysis of trade-offs and provides a way to factor in both the direct effects of Covid-19 deaths and the indirect effects that operate through lower future incomes.

In contrast, approaches that focus on either economic measures or health measures leave an unanswered question of how to compare dollars and lives. While such comparisons are a sensible part of risk analysis—and are typically made using estimates of the value of a statistical life—they may be politicised. Indeed, politicians in several countries have claimed that if responding to Covid-19 is a contest between protecting public health and helping the economy, the only choice is to protect public health, as one (supposedly) cannot put a value on human life. ¹  While such claims are likely to be electorally popular because they appeal to the emotions of the economically naive, they undermine the basis for a rational discussion about appropriate responses to risk. Conducting the analysis in terms of life expectancy helps us to sidestep the risk of an unproductive debate about health versus the economy.

A basic tool for expressing the impact of Covid-19 on life expectancy is the life table, or mortality table, which shows for each age, x, the probability of death before age x + 1. Given that everyone eventually dies, the probability of death for all in a starting cohort of 100,000 members eventually reaches 1. We can denote the period this occurs as tD. Based on age-specific death rates, we can then calculate l(x), which is the number of people from the cohort who survive to age x. One can then work backwards, starting from age tD− 1 to sum the number of people expected to be alive at each year of age from x to tD. When these sums are added together, they give the total number of years yet to be lived by the surviving members of the cohort, as of age x. This sum is typically denoted as T(x). The ratio of T(x) to l(x) gives the average life expectancy, or the remaining years expected before death, for someone from the original cohort, at age x.

To use a life table with empirical data, we appear to need to wait for all members of a cohort to die so that we can calculate the age-specific death rates. In other words, cohort studies cannot cover the more recently born, unless they use projections for age-specific mortality rates. Instead of using these projections, which can be highly unreliable and are available for only a few countries, many statistics agencies produce period life tables. These tables summarise the mortality experiences of people of all ages, drawn from different birth-year cohorts, at particular points in time.

For Indonesia, we use period life tables based on data from the 2010 census conducted by BPS (2015). The census collected data separately for males and females and used five-year age bands, excluding ages zero and one, for which single-year bands were used. Across the various cohorts observed at the time of the census, the male life expectancy at birth was 67.2 years and the female life expectancy was 72.6 years. These expectancies provide our baseline estimates, which can be ‘shocked’ by adding data on Covid-19-related deaths (which will bring forward the expected time of death).

At the time of writing, data on Covid-19-related deaths where the ages were known were available up until 19 July 2020. These deaths total 3,919 and make up about 95% of all recorded Covid-19 deaths in Indonesia for this period. People aged 60 or older accounted for almost half of the deaths and people aged 46–59 for about 40% of the deaths. Figure 1 shows that the deaths are clearly skewed towards older Indonesians. However, the skew is less pronounced than in some other countries. For example, in the United Kingdom, people aged 65 or over accounted for 88% of Covid-19-related deaths, people aged 50–64 for 10% and people younger than 50 for only 2% (Leach and Swann 2020). ² For simplicity, we assume that males and females die at equal rates. We then express the implied number of deaths in terms of the number per 100,000 in order to incorporate this new risk (from Covid-19) into the life tables, which are based on a starting cohort of 100,000 people. We report only the impacts on male life expectancy, because these provide an upper bound in proportionate terms, given that male life expectancy is lower than female life expectancy.

Direct and Indirect Effects of Covid-19 On Life Expectancy and Poverty in Indonesia

All authors

John Gibson & Susan Olivia

https://doi.org/10.1080/00074918.2020.1847244

Published online:

04 December 2020

FIGURE 1 Pattern of Covid-19-related Deaths in Indonesia by Age Group (as of 19 July 2020)

Source: Satgas Covid-19 (2020).

Display full size

Using the age pattern of Covid-19 deaths shown in figure 1, we generate a shock to the 2010 period life tables (distributed according to the age pattern in figure 1). This shock causes an imperceptible fall in life expectancy. To find impacts that are perceptible, we consider scenarios in which Indonesia’s Covid-19 death toll is 10 times higher than the actual death toll, which would mean almost 40,000 deaths; 20 times higher (80,000 deaths); 30 times higher (120,000 deaths); and 40 times higher. Under the most extreme scenario, almost 160,000 people in Indonesia would die from Covid-19, more than in the United States (as of late July). This would translate to a death rate of more than 700 deaths per million people. This death rate would fall between the rates in Belgium and the United Kingdom, which had the second and third-highest rates, respectively, (behind San Marino) at the time of writing. ³ In other words, our modelling includes a scenario where Indonesia may be thought of as having among the worst Covid-19 outcomes in the world.

If Indonesia had 10 times the number of Covid-19 deaths it has, life expectancy would fall by 0.001 years (or 0.4 days) (table 1). If it had 20 times more Covid-19 deaths, one day of life expectancy would be lost. Even under the worst scenario—a death toll higher by 40 times, giving Indonesia the most Covid-19 deaths and third-highest death rate in the world—life expectancy falls by just 0.006 years, or 2.2 days per person. In proportionate terms, this is a reduction of 0.009% (in terms of male life expectancy; the decline in life expectancy for females is even smaller, given their because hospital beds are full of (or being reserved for) patients with Covid-19. Docherty et al. (2020) found that the number of deaths from all causes between February and early May in England and Wales was about 11% above usual for people younger than 45 but 65% above usual for 65–74 year olds and 82% above usual for those aged 75 or older. higher initial life expectancy). Even if Covid-19 were to kill more young people, such as those aged 30, which was the global median age of death during the 1918 influenza pandemic, life expectancy would fall by just four days, based on a death toll 40 times higher than actually experienced. It appears that the media attention on Covid-19 has not been proportional to this moderate threat to life expectancy, compared with the inattention paid to some other risks—such as various sources of infant mortality—that do far more to lower life expectancy.

In contrast to the minute direct effects of the Covid-19 death toll on life expectancy—even given a toll that is 40 times higher than experienced—the indirect effects through a reduction in income are likely to be far larger. The income elasticity of life expectancy is quite high for Indonesia, compared with neighbouring countries. The results in table 2 are based on World Bank data on real GDP per capita (in 2010 dollars) and life expectancy, from 1960 to 2018 for all ASEAN countries. Over this period, life expectancy rose rapidly, by an average of 3.6 years per decade for Indonesia. This was the second-highest rate of increase among the ASEAN countries (behind the rate of 4.3 years per decade for Cambodia). The rise in life expectancy in Indonesia is sensitive to real income growth, with an elasticity of 0.21 (and a standard error of 0.01), which is the highest elasticity for any ASEAN country. In other words, for every 10% rise in real per capita income, life expectancy rises by 2.1%. At the current levels of longevity, this magnitude of percentage increase is equivalent to a rise in life expectancy of 1.4 years for males and 1.5 years for females. ⁴

The cointegration of real income and life expectancy is a long-run relationship. Short-term deviations from the trend growth in income will not necessarily cause short-term deviations from the trend growth in life expectancy. Thus, life expectancy may not immediately fall because of the effects of a negative economic shock, as forecast by various bodies. (In June, the World Bank forecast that Indonesia’s real GDP in 2020 would be the same as in 2019, while in September, the ADB forecast that it would be 1% lower than in 2019.) However, the long-run effects of lower output due to foregone activity during the Covid-19 economic crisis will likely decrease future gains in life expectancy, based on the relationship in table 2.

In figure 2, we present two time paths for real per capita GDP in Indonesia through to 2050. The first of these projections appeared in a PricewaterhouseCoopers report in 2017 (Hawksworth, Clarry and Audino 2017). It assumes annual population growth of 0.6% per annum and real per capita GDP growth of 3.1% per annum. Under this baseline scenario, Indonesia would reach a per capita GDP level of $11,500 (in 2010 dollars) by 2050. The second time path incorporates the Covid-19 shock, based on a forecast by McKibbin and Fernando (2020). They forecast that real GDP will fall by about 4% in 2020, recover partly in 2021, and then slowly return to the pre-Covid-19 level by 2025. The trend growth rate will be re-established from 2026 onwards. ⁵

Direct and Indirect Effects of Covid-19 On Life Expectancy and Poverty in Indonesia

All authors

John Gibson & Susan Olivia

https://doi.org/10.1080/00074918.2020.1847244

Published online:

04 December 2020

FIGURE 2 The Trajectory of Real Per Capita GDP (2010 dollars) with and without the Covid-19 Shock

Note: The baseline trajectory assumes annual population growth of 0.6% per annum and real per capita GDP growth of 3.1% per annum.

Display full size

The area between the two series in figure 2 reflects the long-term value of the output foregone because of the Covid-19 shock. Before reporting the monetary value of this area, we note again that at least three factors contribute to this shock: the economic shock from social distancing measures (PSBBs) mandated by national and local governments to reduce the spread of Covid-19; other economic shocks, such as lower export demand; and the rational response of individuals who naturally reduce their public movement and gathering to avoid infection, even if they are not forced to take this precaution. We calculate the net present value of the two time paths through to 2050 (using a discount rate of 10%). ⁶ The present value of the flows of real per capita income is 13% lower than it would have been without the Covid-19 shock.

With an income elasticity of life expectancy of 0.21, this 13% reduction in real income over the long run translates into 2.6% less life expectancy. If we use current life expectancy as a base, this percentage decline is equivalent to life expectancy that is 1.7 years lower than would otherwise have been the case. This indirect effect is far larger than the direct effect of the current Covid-19 death toll on life expectancy (table 1). Even if the Covid-19 death toll were 40 times worse in Indonesia than experienced, the indirect effect on life expectancy would be almost 300 times larger than the direct effect. ⁷ Even with different assumptions—such as a less severe shock, a faster recovery or lower discount rates—we see clearly that the indirect effects of the Covid-19 shock on life expectancy are going to be far larger than the direct effects.

These indirect effects operate through lower incomes, which affect health and life expectancy in multiple ways. We do not attempt to identify and quantify all of the ways in which these indirect effects may eventuate. However, even at this early stage, evidence is emerging of some of the more important indirect effects. For example, child immunisation services have been reduced at local health centres such as posyandu (local integrated health centres) and puskesmas (primary health clinics) because of a reorientation of these centres towards Covid-19 cases (Faizal 2020). A failure to immunise children at the right time (or at all)—combined with the health burden of other childhood diseases, such as malaria, that may now go untreated—could potentially have very large negative effects. This is especially because rates of child stunting and wasting were higher in Indonesia than in other ASEAN countries even before Covid-19 (See Anne Booth, Raden Purnagunawan and Elan Satriawan’s August 2019 Survey in BIES). The burden of childhood disease can combine with poor nutrition to stunt the growth of children, leading to shorter height as adults (Alderman, Hoddinott and Kinsey 2006). These adults, in turn, will tend to have lower lifetime incomes. Schultz (2002) found that each extra centimetre of adult height increased wages by almost 2% in Ghana and Brazil, while Indonesian men have 2.5% higher annual earnings for each additional centimetre of height. Taller Indonesian men also report being happier (Bargain and Zeidan 2017). Thus, the economic shock shown in figure 2 may eventually lead to lower productivity through the effects on child health and nutrition.

Another possible pathway is through limits on food supply, or increases in food cost, as Covid-19 spreads from urban areas to the countryside (Yusuf et al. 2020). Most farmers in Indonesia are aged 45 or older, which puts them at higher risk of getting sick from Covid-19. ⁸ If these farmers reduce their activity either because they become ill or because they are taking precautions, then food production may suffer. Moreover, the leading centres of rice production are West Java, Central Java and East Java, which recently experienced a surge in cases as the pandemic shifted from the epicentre in Jakarta towards other parts of Java. Consequently, food security is threatened by Covid-19, which could lead to reduced child health and lower human capital in future. ⁹

The indirect effects on adults also work through various pathways. For example, if clinics focus on treating or preparing for Covid-19 patients, the detection and treatment of diseases such as hypertension and diabetes may be delayed, leading to earlier deaths. This reality is highlighted in an analysis by the UK government’s Scientific Advisory Group for Emergencies (SAGE). It found that about 6,000 excess deaths occurred up to 1 May 2020, compared with 25,000 direct deaths from Covid-19, because of reductions in accident and emergency admissions. The study noted that these reductions were likely due to patients’ reluctance to attend hospital because of fear of exposure to Covid-19, and to changes in hospital procedures to make space for potential Covid-19 cases. Even more people were indirectly affected by Covid-19 in care homes. About 10,000 excess deaths of people in care homes resulted from their not wanting or not being allowed to transfer to hospital when they required hospital care, or from their being discharged from hospital early to free up hospital beds for Covid-19 cases. Also expected are a further 12,500 excess deaths from delays to elective care during the lockdown period (SAGE 2020). When deaths from these three indirect pathways are combined, the number of indirectly caused deaths slightly exceeds the number of directly caused deaths, even without considering pathways that involve economic impacts, such as higher unemployment and lower incomes. ¹⁰

MODELLING IMPACTS ON POVERTY

In order to mitigate reductions in life expectancy, we first need to detail the likely effects of Covid-19 on poverty. This is because people at the greatest risk of poverty are the most likely to have their life expectancy indirectly affected by Covid-19. The pathways through which this can occur include the lifetime and intergenerational impacts of impaired health, education, nutrition and employment. Over the past two decades, evidence on epidemics and pandemics caused by illnesses such as H1NI, MERS and SARS has shown that the resulting economic contractions were accompanied by rising income inequality, which occurred partly because the employment prospects for people with low education levels were badly affected, while employment of the more highly educated was less affected (Furceri et al. 2020).

Likewise, the absence of normal in-class education will have worse effects on low-income students than high-income students. Dorn et al. (2020) estimate that US students will lose over six months of learning overall if their schools do not return to typical in-class learning by January 2021. However, the learning loss will be twice as high (12.4 months) for low-income students, who have far less access to high-quality remote learning (Dorn et al. 2020). Similar patterns of learning disadvantage are likely to occur in other countries. In Indonesia, the gap in online learning may exacerbate the gap in education quality between major cities and smaller communities (Cahya 2020).

Many forecasts of the effects of Covid-19 on poverty have already been made, both globally and domestically. The global forecasts show a broad range of estimates, from 49 million people being pushed into extreme poverty (Mahler et al. 2020), to up to 550 million (Sumner, Hoy and Ortiz-Juarez 2020). ¹¹ All of these estimates reveal that the Covid-19 shock is likely to cause the first increase in global poverty since 1997, wiping out the gains made over the past decade. For Indonesia, simulations by Asep Suryahadi, Ridho Al Izzati and Daniel Suryadarma show that the poverty rate could increase by between one percentage point (if the economy grows at a rate of 4.2% in 2020) and 7.4 percentage points (if the economy contracts by 3.5% in 2020). ¹² These simulations entail increases in the number of poor Indonesians of between 1.3 million and 19.7 million. ¹³ To put these figures into context, we note that the poverty rate in 2019 was 9.2%, with 25 million people living below the poverty line. Therefore, the most pessimistic scenario of economic growth entails poverty almost doubling. Such a large increase would set Indonesia back to the poverty rate seen in the early 2000s (World Bank 2020a).

A key difficulty in forecasting the increase in poverty due to the Covid-19 shock comes from a lack of recent data. Poverty estimates in Indonesia are based on household expenditure data available only with a considerable lag. Suryahadi, Al Izzati and Suryadarma tried to work around this lag by looking at how the shock of fuel price rises in 2005–06 affected various points of the household expenditure distribution. They then used these distributional effects to model the distribution of the predicted decline in average per capita household expenditure between 2019 and 2020. While this approach is able to predict aggregate increases in poverty, it cannot identify the areas of Indonesia at the greatest risk of increased poverty. Therefore, this approach cannot guide the geographical targeting of social assistance.

The alternative approach used in our study exploits non-traditional data generated in near real-time. This allows us to estimate where the declines in economic activity have been most severe in Indonesia. One type of data used elsewhere to track the impact of the Covid-19 shock is point of sale (POS) data. This has been used in Denmark (Andersen et al. 2020), Spain (Carvalho et al. 2020) and the United States (Baker et al. 2020). An advantage of such data is that they allow sectoral and regional variations in household expenditure to be tracked at high frequency. However, a problem with getting such data in developing countries is that retail payment systems are not as advanced as in developed countries; in developing countries, some cash transactions through traditional market channels will not be included in the data, and obtaining data from card-based payment platforms can be difficult.

A key study that informs our research design was authored by researchers at the central bank of Mexico. These researchers had access to all Mexico’s POS data, covering about 10 million transactions per day (Campos-Vazquez and Esquivel 2020). In contrast, POS studies in other countries have been limited to working with data from a particular bank or company. The POS transactions in Mexico during the Covid-19 shock are related to state-level changes in public mobility, estimated using mobile device location data from Google (2020). In Mexico, the elasticity between POS expenditure and mobility is almost one, which is far higher than in richer countries. A plausible reason for the high elasticity is that online shopping is much less developed in Mexico than in richer countries, so consumers must travel to retailers to buy goods. The implication of this is that in developing and middle-income countries, such as Indonesia, the near real-time mobility data from Google should serve as a good proxy for changes in expenditure. Given that poverty is measured using household expenditure data in Indonesia, these mobility data provide a way to simulate changes in poverty.

Specifically, we combine Google mobility reports, as a proxy for changes in economic activity, with pre-crisis household expenditure data from the 2018 National Socio-Economic Survey (Susenas). The mobility reports use the location history of mobile devices that have the Google Maps application. Using data from this application, Google generates its Covid-19 Community Mobility Reports. These reports use aggregated, anonymised data showing changes in the frequency and length of visits to various places compared with a baseline. The reports are available by province and for categories such as workplaces, retail and recreational venues, grocery and pharmacy stores, parks, transit stations and residential areas. Similar data have been used in developing countries, including nine countries in Africa and Latin America, to estimate the impacts of Covid-19 lockdowns on poverty (Bargain and Aminjonov 2020).

The mobility data show that people began to spend more time in residential locations and less time in (or make fewer visits to) workplaces, parks, retail and recreational venues, and especially transit stations, from mid-March (figure 3). At the nadir, in late April, mobility at transit stations had fallen by 60%, compared with a pre-Covid-19 baseline based on the five weeks from 3 January to 6 February 2020. The declines in mobility at workplaces, parks, and retail and recreational venues were generally in the 30%–40% range from late March until early June, with mobility at grocery and pharmacy stores declining less drastically. Notably, the decline in mobility at places outside the home and the increase in time spent in residences began before the health emergency was declared at the end of March and before the large-scale social restrictions began, starting in Jakarta on 10 April. A similar pattern was seen elsewhere, with fear of infection leading to precautionary declines in mobility that either preceded or dwarfed those attributable to restrictions imposed by governments (Goolsbee and Syverson 2020).

Direct and Indirect Effects of Covid-19 On Life Expectancy and Poverty in Indonesia

All authors

John Gibson & Susan Olivia

https://doi.org/10.1080/00074918.2020.1847244

Published online:

04 December 2020

FIGURE 3 Changes in Mobility in Indonesia, based on Google Mobility Reports (seven-day moving average)

Source: Google (2020).

Display full size

The national-level mobility changes in figure 3 disguise considerable regional variation. In figure 4, we show mobility changes for the workplaces category for selected provinces. We focus on workplaces as being the most relevant to economic activity. The figure shows an early decline in mobility in Bali in February, a recovery in mid-March and a sharp decline to more than 50% below the baseline by late March. Bali generally had the deepest decline in workplace-related mobility of all provinces, except from mid-April to early June, when Jakarta had bigger declines. All provinces had at least 20% declines from early April, with some volatility in late May, then rising mobility until mid-June. By this time, mobility in East Nusa Tenggara was almost back to the baseline level, whereas mobility in Bali and Jakarta was still 30%–40% below the baseline. The variations by province partly correspond to the timing of social restrictions implemented in some provinces.

Direct and Indirect Effects of Covid-19 On Life Expectancy and Poverty in Indonesia

All authors

John Gibson & Susan Olivia

https://doi.org/10.1080/00074918.2020.1847244

Published online:

04 December 2020

FIGURE 4 Changes in Workplace-Related Mobility in Selected Provinces (seven-day moving average)

Source: Google (2020).

Display full size

The median declines for Indonesia’s 34 provinces are shown in figure 5. We use the median rather than the mean, because it is a more robust measure, especially because of the volatility in the time series due to public holidays, such as Lebaran, that affected mobility. The provinces with the biggest falls in mobility were Bali (36%), Jakarta (33%) and Yogyakarta (31%). The smallest falls in mobility were in East Nusa Tenggara (9%), and West Sulawesi and North Kalimantan (both 12%). In general, people in poorer provinces, such as Papua and East Nusa Tenggara, reduced their mobility for work less than people in richer provinces, such as Bali and Jakarta. A similar pattern in other countries was noted by Bargain and Aminjonov (2020). They found that poorer people were more likely to travel outside their homes and continue to work, because staying at home would likely entail a potentially devastating loss of income.

Direct and Indirect Effects of Covid-19 On Life Expectancy and Poverty in Indonesia

All authors

John Gibson & Susan Olivia

https://doi.org/10.1080/00074918.2020.1847244

Published online:

04 December 2020

FIGURE 5 Median Declines in Workplace-Related Mobility by Province (15 February to 25 July 2020)

Source: Google (2020).

Display full size

The median changes in mobility by province shown in figure 5 provide the basis for estimating the expenditure shock by province. Noting that provinces are Indonesia’s first subnational level, evidence from a similar level in the subnational hierarchy (states in Mexico) from Campos-Vazquez and Esquivel (2020) shows that the elasticity between expenditure and mobility is almost one. Validating this elasticity estimate using data from Indonesia, and also examining if the elasticity depends on how aggregated the spatial units are, is an interesting task for future research. In order to go from expenditure change predictions to poverty estimates, we account for the share of consumption that is purchased (using province-level Susenas data for 2018). We do this because the mobility shocks shown in figure 5 should only have affected purchases; other parts of consumption (such as home production) should have been unaffected. Following Sumner, Hoy and Ortiz-Juarez (2020), we operationalise the reduction in real consumption by raising the value of the provincial poverty line (obtained from BPS 2018). For example, for a per capita real consumption contraction of x%, the poverty line z is adjusted upward as z/(1 − x).

Table 3 estimates the impact of projected falls in the real value of consumption, as derived from mobility changes, on the rate of poverty and the number of poor. We use the poverty rate calculated from the 2018 Susenas as our baseline. As the table shows, poverty in Indonesia had a strong spatial dimension even before Covid-19 began to affect households. In general, the poverty rates in the eastern provinces were much higher than those in the western or central areas. For Indonesia as a whole, the baseline poverty rate in 2018 before Covid-19 was 10%, which indicates that 26.2 million people were living below the poverty line.

The decline in real consumption during the Covid-19 shock results in an increase in the national poverty rate of about nine percentage points. This suggests an increase of about 23 million poor people. This result is in line with the findings of Suryahadi, Al Izzati and Suryadarma in the August BIES. They found that a –3.5% economic growth rate for 2020 would increase the poverty rate by 7.2 percentage points. Table 3 shows that the increase in poverty occurs in all provinces but with considerable regional variation; poverty increases by three percentage points in Papua and East Nusa Tenggara but by 16 percentage points in Yogyakarta and 13 percentage points in Bali. As mentioned, the economic cost of the Covid-19 economic shock has varied significantly across sectors and regions. While some sectors, such as tourism and transportation, have been severely hit, others, such as telecommunications, have been affected much less. This is reflected in table 3, which shows that Indonesian provinces that are highly dependent on tourism (such as Bali and Yogyakarta) are among the worst affected areas.

In addition to highlighting the spatial dimension of poverty changes during Covid-19, our results support the broadening of social assistance programs to include non-regular beneficiaries. As shown in figure 6, the increases in poverty caused by the Covid-19 shock are much higher in provinces that have lower initial poverty. In other words, provinces with little existing poverty are likely to experience the greatest increases in poverty. This means that the social protection system needs to be expanded in places where people have not widely relied upon it previously.

Direct and Indirect Effects of Covid-19 On Life Expectancy and Poverty in Indonesia

All authors

John Gibson & Susan Olivia

https://doi.org/10.1080/00074918.2020.1847244

Published online:

04 December 2020

FIGURE 6 Relationship between Provinces’ Initial Poverty Rates and Their Projected Poverty Rates due to Covid-19

Source: Google (2020).

Display full size

Concerning, then, is that a lack of reliable data has caused delays in the distribution of social assistance to eligible recipients, even though the government has taken steps to cushion the impact of Covid-19 by allocating Rp 210 trillion to expand social safety nets to cover the existing poor and new poor. The government had disbursed just 34% of the social assistance budget as of 29 June 2020 (Rahman 2020), and many needy people were being missed owing to targeting problems. One weakness that may slow the distribution of the funds is that the most recent updates of the national unified database used by the Ministry of Social Affairs to target social assistance beneficiaries was done in 2015. This increases the risk of ineffective targeting, either by giving benefits to the non-poor or by missing the new poor, which means that fiscal spending on poverty alleviation could be inefficient.

One implication of the results in figure 6, and of the government’s difficulties in disbursing social assistance, is that traditional administrative data may not be adequate for responding quickly to new demands during a pandemic. In other words, the official statistics may not by sufficiently up to date to be useful. Higher-frequency, alternative data are needed to complement the official statistics. Now may be an opportune time for the government to evaluate the need to update and invest in near real-time data on the poor to improve targeting. This might include working with data held by private companies, which have unprecedented capacity to measure economic activity at a granular level very rapidly. If such data could be harnessed, they may help policymakers make better decisions. ¹⁴

CONCLUSIONS

The Covid-19 shock has led to an unprecedented set of economic and health policy responses in Indonesia. To avoid unnecessary harm to quantity and quality of life, such responses must consider both the direct and indirect effects of the shock.

Our analysis shows that the indirect effects of Covid-19 on life expectancy— operating through lower future income—are likely to exceed the direct effects of Covid-19 deaths by at least five orders of magnitude. Therefore, any interventions to reduce the risk of Covid-19 must be precisely targeted and mindful of the indirect effects. Otherwise, actions to prevent deaths from Covid-19 may end up doing more damage than good, by indirectly reducing life expectancy through lower future incomes.

We also find that the effects of Covid-19 on poverty are spatially heterogeneous in Indonesia and that the increase in poverty is higher in provinces with lower initial poverty rates. This finding suggests a need to broaden social assistance programs to cover not only the existing poor but also the new poor. While the government has taken steps to mitigate the effects of Covid-19 on the poor, a lack of reliable data has resulted in ineffective targeting, as well as delays in the distribution of social assistance to the eligible. The implication is that now may be an opportune time for government to evaluate the need to update and invest in innovative data collection methods on the poor to improve targeting.