A method for mapping high resolution Gross Value Added and its changes between 2012 and 2021 in Gauteng

Introduction

GCRO's March and April 2024 Maps of the Month focused on aspects of the spatial economy. In March, we looked at formal employment by municipality across South Africa, and in April we looked at manufacturing jobs in Gauteng. This month we continue the economic theme by mapping modelled Gross Value Added (GVA) at a ward level over time. The map uses an innovative method of modelling economic estimates from satellite imagery of nighttime lights.

Subnational and regional level economic data is important for understanding nuanced spatial trends that can inform planning and development strategies. It is not only important to understand the changes in economic activity but also where these changes occur. However, data at a fine scale, such as at ward level, is not easily available or accessible in South Africa. Traditionally, economic data such as Gross Domestic Product is made available at a national or provincial level (Chen and Nordhaus, 2011; Doll et al., 2006; Ghosh et al., 2010), and because national and provincial figures cannot be disaggregated it is difficult to align to datasets at finer spatial resolutions.

This Map of the Month uses modelled economic data to understand the change in economic activity at ward level across Gauteng between 2012 and 2021. The analysis presented here is an extension of work that has been undertaken in collaboration with the City of Johannesburg's Group Strategy, Policy, Coordination and Relations Unit to model economic activity in Johannesburg at fine spatial scales (Naidoo et al., 2023; Naidoo et al. 2024). See that work, published as a GCRO Rapid Research Paper, here. The analysis used accessible datasets to provide a measure of economic activity at the finest possible scale and to project this measure for years where data was not available. During that study, it was determined that Gross Value Added (GVA)¹ was the best available economic data that could be mapped (CSIR (Council for Scientific and Industrial Research), 2018). GVA is a well-known economic metric that measures the amount of goods and services produced across all economic sectors (Kenton, 2022). However, a limitation of this GVA data was that it was only available until 2016, and only at the level of CSIR's mesozone layer. This sparked the need to develop a method that could use the 2016 GVA data to predict GVA at more granular levels for subsequent years.

A novel approach to understanding the spread of economic activity has been to correlate available economic data with nighttime lights data from satellite imagery (Bhandari and Roychowdhury, 2011; Bickenbach et al., 2016; Ghosh et al., 2010; Gibson et al., 2021; Gibson and Boe-Gibson, 2021). Nighttime lights imagery measures artificial light emissions that highlight human activity across the world, such as the constant brightness of urban centres and areas of industrial activity (Levin et al., 2020, Maree and Naidoo, 2019, Naidoo and Maree, 2020). This provides a foundation for the development of a model to quantify economic activity using non-traditional datasets. In our analysis, GVA data was used along with VIIRS nighttime lights imagery and LandScan population data in a Random Forest model environment to predict average annualised GVA (Naidoo et al., 2023; Naidoo et al. 2024). The result was the production of an annual GVA value at a square kilometre scale for the period between 2012 and 2021. This data was aggregated at a ward level for comparative analysis over time. A number of other socioeconomic data sets, for example, demographic data, are available at a ward level making this a useful scale for analysis and planning.

Mapping GVA at a ward level

Figure 1 maps percentage change in modelled GVA values between 2012 and 2021. Wards shaded in brown show areas of decline and wards in blue/green show areas of growth during this time period.

Figure 1: Percentage change of GVA per ward between 2012 and 2021.

All municipalities have areas of growth and decline. Many of the areas of decline are in central business areas or areas traditionally associated with economic activity. For example, Roodepoort, Randburg, Kempton Park, Pretoria Central, Vanderbijlpark, and the Johannesburg CBD show declining GVA between 2012 and 2021. However, Sandton, Midrand, Mamelodi, eastern portions of Tembisa, Krugersdorp, and Springs have shown an increase in GVA. Additionally, some areas of mining, industry, and manufacturing have shown a decline. This includes areas like Carletonville (N), Boksburg (M), and Pretoria West (B). Areas that include industrial and mining activities saw larger proportions of economic loss compared to commercial areas that showed moderate negative change such as Roodepoort (F) and Randburg (G).

One of the advantages of using nighttime lights imagery and population data to model economic activity is that it captures both formal and informal economic activity (the April 2024 Map of the Month was generated using anonymised tax data and therefore does not reflect informal economic activity). In Figure 3 many areas with the highest proportion of growth are in areas where there is the expansion of new settlements and where the informal economy dominates, for example in parts of Soshanguve (A), Mamelodi (D), and Soweto (K). However, these areas do not show uniform growth and do show differences in GVA change. There have also been positive change around areas of more formalised growth, such as Midrand (E), Fourways and Sandton (H), which has developed as commercial zones during this period. Though this map emphasises the positive and negative changes over time, it also shows areas of minimal change. For instance, the Johannesburg CBD (L) and Pretoria Central (C) are examples of areas where there has not been much change during this time.

Change in economic activity over time

Between 2012 and 2021 Gauteng as a whole saw a compound annual growth rate in modelled GVA of 1.47%. Against this overall average, Figure 2 compares the compound average growth rate for each municipality. Five municipalities have a compound average change greater than the Gauteng average, with Lesedi having the largest positive change over time of 2.7%. Of the metropolitan municipalities, the City of Johannesburg is the only metro that has a compound average that is less than the Gauteng average. Two municipalities, Emfuleni and Rand West City have negative compound averages indicating a decline in economic activity. The Emfuleni negative change is a notable exception and indicates a worrying trend of economic decline. It is important to note that these figures indicate percentage change and not the absolute value of an economy. So while Lesedi shows the highest compound average change it is off a much smaller economic base compared to the City of Johannesburg.

Figure 2: Average annual compound change per municipality between 2012 and 2021

Another way to look at the change over time is to chart the annual average GVA per capita in each local municipality. Figure 3 shows that Midvaal’s GVA per capita is much higher than the other municipalities largely due to its smaller population size and presence of manufacturing areas around Meyerton. In contrast, the municipalities with larger populations, such as the City of Johannesburg, have a GVA per capita closer to the Gauteng average. It is interesting to note that Johannesburg’s GVA per capita between 2012 and 2016 was lower than the Gauteng average, and since 2016 has maintained a value slightly higher than the provincial average. This is in contrast to Rand West City, which had a GVA per capita that was much higher than the Gauteng average in 2012, but saw a gradual decline to below the provincial average. Although most municipalities tend to have a GVA per capita that is above the Gauteng average, Merafong City and Emfuleni are consistently lower than the Gauteng average over time and Emfuleni's GVA per capita has been declining over time.

Figure 3: GVA per capita for Gauteng municipalities between 2012 and 2021

Conclusion

Fine-scale economic data offers valuable insights into nuanced spatial economic trends. This is essential in order to craft targeted economic development strategies. Despite its importance, accessing fine-scale spatial economic data remains a challenge due to the traditional availability of economic metrics predominantly at national or provincial levels. The analysis presented in the Map of the Month provides one way to model economic data at the ward level and sheds light on significant trends in economic activity over time. It highlights areas of both positive and negative growth, showcasing the dynamic nature of economic development within Gauteng.

Limitations

It is important to note that the analysis is based on modelled data and is an approximation of the actual values, and therefore the data does include outliers. The most recent GVA data at a mesozone level was only available until 2016. The modelling exercise would benefit from more current economics data for model calibration, but this is not currently available. Although this model provides a predicted measure, it is still useful in the identification of spatial and economic trends over time rather than presenting an absolute economic value.

The issue of scale adds to the complexity of data synthesis as the various data sources had different scales, and needed to be aggregated to a single similar ward level. This required the use of tools that could aggregate, for instance, satellite imagery with a spatial resolution of one square kilometre to ward-level polygons and mesozones. This was done in QGIS using the Zonal Statistics tool, where the sum of GVA and the sum of the population were extracted from satellite imagery and aggregated into ward polygons for each year between 2012 and 2016. Further details on the method developed here are available in this Rapid Research Paper done on request from and in partnership with the City of Johannesburg.

This project was an experiment in trying to understand where economic activity happens in space. The data can show trends in economic growth or decline but is not able to quantify the impacts of economic shocks or what would have caused them. It is clear that between 2019 and 2021 there was a noticeable decline across the province. The most likely causes of this are the COVID-19 lockdown and the start of increasing levels of loadshedding. It is well known that loadshedding has had a severe impact on the economy, but it is also a limitation in our analysis because it impacts measures of average light intensity detectable by satellite. It is very difficult to ascertain the relative contributions to changing light intensity of loadshedding and actual economic decline. Hence, due to the significant increase in loadshedding in 2022 and 2023 the time scale for this analysis was limited to 2021.

Footnote

¹ GVA is an economic metric that measures the total value of all goods and services at the price at which these are produced in an economy. It is a component of GDP, with GDP being the sum of GVA for all economic sectors, adjusted for taxes and subsidies.

References

Bhandari, L., Roychowdhury, K., 2011. Night Lights and Economic Activity in India: A study using DMSP-OLS night time images. APAN Proceedings 32, 218. https://doi.org/10.7125/APAN.32.24

Bickenbach, F., Bode, E., Nunnenkamp, P., Söder, M., 2016. Night lights and regional GDP. Rev World Econ 152, 425–447. https://doi.org/10.1007/s10290-016-0246-0

Chen, X., Nordhaus, W.D., 2011. Using luminosity data as a proxy for economic statistics. Proc. Natl. Acad. Sci. U.S.A. 108, 8589–8594. https://doi.org/10.1073/pnas.1017031108

CSIR (Council for Scientific and Industrial Research), 2018. stepSA Socio-Economic Indicators and Mesozones [WWW Document]. MesoZone 2018v1 Dataset. URL http://stepsa.org/socio_econ.html#Geospatial (accessed 11.16.22).

Doll, C.N.H., Muller, J.-P., Morley, J.G., 2006. Mapping regional economic activity from night-time light satellite imagery. Ecological Economics 57, 75–92. https://doi.org/10.1016/j.ecolecon.2005.03.007

Ghosh, T., Powell, R.L., Elvidge, C.D., Baugh, K.E., Sutton, P.C., Anderson, S., 2010. Shedding Light on the Global Distribution of Economic Activity. TOGEOGJ 3, 147–160. https://doi.org/10.2174/1874923201003010147

Gibson, J., Boe-Gibson, G., 2021. Nighttime Lights and County-Level Economic Activity in the United States: 2001 to 2019. Remote Sensing 13, 2741. https://doi.org/10.3390/rs13142741

Gibson, J., Olivia, S., Boe-Gibson, G., Li, C., 2021. Which night lights data should we use in economics, and where? Journal of Development Economics 149, 102602. https://doi.org/10.1016/j.jdeveco.2020.102602

Kenton, W., 2022. Gross Value Added (GVA): Explanation, Formula, Example [WWW Document]. Investopedia. URL https://www.investopedia.com/terms/g/gross-value-added.asp (accessed 11.3.22).

Levin, N., Kyba, C.C.M., Zhang, Q., Sánchez de Miguel, A., Román, M.O., Li, X., Portnov, B.A., Molthan, A.L., Jechow, A., Miller, S.D., Wang, Z., Shrestha, R.M., Elvidge, C.D., 2020. Remote sensing of night lights: A review and an outlook for the future. Remote Sensing of Environment 237, 111443. https://doi.org/10.1016/j.rse.2019.111443

Maree, G., Naidoo, Y., 2019. Gauteng at night: light intensity at 2am. GCRO Map of the Month. September. https://doi.org/10.36634/WQCU4974

Naidoo, Y., Maree, G., 2020. Gauteng going dark. GCRO Map of the Month. July. https://doi.org/10.36634/YJTG3210

Naidoo, Y., Maree, G., Naidoo, L., 2023. Estimating change in the spatial economy of the city of Johannesburg using nighttime lights imagery and population data in a Random Forest modelling environment, in: Chrysoulakis, N., Erbertseder, T., Zhang, Y. (Eds.), Remote Sensing Technologies and Applications in Urban Environments VIII. Presented at the Remote Sensing Technologies and Applications in Urban Environments VIII, SPIE, Amsterdam, Netherlands, p. 4. https://doi.org/10.1117/12.2684320

Naidoo Y., Maree, G., Naidoo, L. and Götz, G. (2024). Using night lights satellite imagery to estimate spatial economic change in Johannesburg between 2011 and 2021. GCRO Rapid Research Paper. Gauteng City-Region Observatory (GCRO). Johannesburg. https://doi.org/10.36634/JBDZ2536

Sims, K., Reith, A., Bright, E., McKee, J., Rose, A., 2022. LandScan Global 2021. https://doi.org/10.48690/1527702

Vermote, E., Franch, B., Claverie, M., 2016. VIIRS/NPP Surface Reflectance 8-Day L3 Global 500m SIN Grid V001. https://doi.org/10.5067/VIIRS/VNP09H1.001

Comments and input: Christian Hamann, Richard Ballard, Graeme Götz

Map design: Jennifer Murray

Suggested citation: Naidoo, Y., Maree, G. and Naidoo, L. (2024). A method for mapping high resolution Gross Value Added and its changes between 2012 and 2021 in Gauteng. Map of the Month. Gauteng City-Region Observatory. May 2024. https://doi.org/10.36634/ENXB6242.