Comment

Comment by Janosch Ondarczek on December 20, 2013 at 4:13am

Hi Evan, Thanks for your follow up question.

The kerosene price during the time of the KIHBS survey was not part of the questionnaire, so I had to infer it from the data: The resulting average kerosene price (indirectly) reported in our dataset is approximately 63 KSH/liter (or 0.73 USD/litre at today's exchange rate). Unfortunately, I do not have data on long-term price developments, but I believe Arne addressed this question in his previous post of February 11th.

Best regards and merry Christmas.

Comment by Evan Mills on December 6, 2013 at 12:46pm

Hi Janosh,Can you remind us what the prevailing kerosene prices were for the period of analysis and how those compare to longer-term prices?  As you no doubt know, prices in some areas have fluctuated by a factor of 4 in recent years, in an amplified manifestation of the swings in world oil prices.

Comment by Janosch Ondarczek on December 6, 2013 at 12:30pm

Hello everyone,

Thanks for the lively interest in our short report that Arne posted earlier this year, as well as your various comments on the issue of lighting costs in Kenya. Here are some responses to your comments and questions.

Firstly, as the original paper that started the whole debate has now been published in Energy Economics, I would like to draw your attention to it: Jann Lay, Janosch Ondraczek, Jana Stoever, Renewables in the energy transition: Evidence on solar home systems and lighting fuel choice in Kenya, Energy Economics, Volume 40, November 2013, Pages 350-359, ISSN 0140-9883, http://dx.doi.org/10.1016/j.eneco.2013.07.024.

In the final version of the paper we also clarify that our data do not suggest major differences in the expenditure patterns of rural and urban households: "It is worth mentioning that, notwithstanding marked differences in the characteristics of rural and urban households, we do not separate our estimation by these two groups. When estimating the two samples individually, we find that the covariates exert similar effects in the respective sub-samples. As the corresponding sample size reductions would result in less precise estimates of coefficients, especially for not so common fuel choices,we instead choose to include a dummy variable for urban areas and conduct the regression using the whole sample." (cf. page 355, right hand column).

Secondly, I have revisited our data to see whether we find a significantly higher proportion of income going to lighting among the poorest households than our data (which is aggregated in quartiles), and tables A.2. and A.4. in our Lighting Africa note, suggest at first glance. What we find in the case of kerosene is a more or less continuous decline from first decile to last decile, much as some of you expected: Whereas the median expenditure on kerosene in the 1st decile (i.e. the poorest households) is 3.47% of annual income, it is only 0.88% for the 10th decile (i.e. the richest households), see Table 1 below. What we do not observe, contrary to some of the other studies cited by some before, is an initial increase in the percentage of income allocated to kerosene, and then a decline.

Table 1: Median expenditure on kerosene by decile, absolute, and as a percentage of annual income:

Income decile	Absolute Monthly Expenditure (KSh)	Percent of Annual Income Spent on Fuel
1	70	3.47
2	100	2.79
3	120	2.41
4	150	2.41
5	150	2.18
6	168	2.02
7	200	1.97
8	240	1.73
9	250	1.41
10	280	0.88
Total	156	2.01

We do observe a similar effect for household spending on electricity; i.e. the share of income allocated to electricity tends to be highest among the poorest households, but our sample is very restricted in the bottom five deciles (N below 100 for each of the five deciles). Hence, I would not put too much emphasis on the curve suggested by the next table (Table 2), which shows the income share to go up between deciles 4 and 5. On the other hand, the increase seen between deciles 9 and 10 is probably statistically significant, as N > 300 in both cases.

Table 2: Median expenditure on electricity by decile, absolute, and as a percentage of annual income:

Income decile	Absolute Monthly Expenditure (KSh)	Percent of Annual Income Spent on Fuel
1	120	4.68
2	100	2.74
3	100	2.22
4	100	1.82
5	200	2.61
6	200	2.28
7	200	1.94
8	200	1.72
9	300	1.61
10	850	1.99
Total	332	1.93

Lastly, Evan Mills rightly identified an editing error in Table A.5., where we not only mixed up cooking and lighting, but also omitted heating altogether. Sorry! The correct table is shown below, and we will try to re-post a corrected version of the Lighting Africa note soon.

Table A.5. *corrected*: Primary use of collected and purchased firewood, in percent
	Collected firewood		Purchase firewood
Purpose of use	Frequency	Percent	Frequency	Percent
boiling	109	1.5	64	3.4
heating	61	0.8	18	0.9
cooking	7,280	96.9	1,817	95.0
lighting	63	0.8	10	0.5
other	3	0.0	3	0.2
Number of observations	7,516	100.0	1,912	100.0

- Best

Janosch

Comment by Evan Mills on October 26, 2013 at 10:24pm

Here is another data point (IMF study), going back to 2006 (before the big energy price spikes of 2008).

Ghana: lowest quintile spent 5.9% of income on kerosene.  The number dropped to 1.6% for the highest income quintile.  That said, the values were lower for the other countries in the study (Bolivia, Jordan, Mali, and Sri Lanka), although fell with rising income in each case.

Comment by Betsy Teutsch on October 7, 2013 at 5:14pm

Thanks Darin.  I am writing a book on tools to help women work their out of poverty.  Solar lanterns - many of them, including DLight - will be featured.  It is not an academic study so I'm not citing sources or footnoting but I do want to use defensible data.  The argument for solar lanterns is very compelling even if the figure is just 1.5% improvement because of the improved quality of light, health benefits and eco benefits.  

It is a big difference - percent of budget vs. income.  The more affluent you are, the smaller percentage you spend on lighting, though presumably the cost of lighting is stable.  What is useful to the lay reader is equivalent amounts.  So getting rid of kero expenses yields x amount of, say, milk, vegetables or eggs, or medicine or school fees.  Simon Berry at ColaLife pegs the cost of their Oral Rehydration course of treatment packs as "the price of 5 bananas."  That helps people visualize savings/expenses.  Please keep my posted on your findings.  betsy at betsyteutsch.com

Comment by Darin Kingston on October 7, 2013 at 4:35pm

Hi Betsy, 

The figure you note from our Customer Benefits page comes from Lighting Africa's Solar Lighting for the BOP report (see page 16; technically this should say "budget" not income). There is no citation because this page is intended for distribution partners and a general audience for whom a more qualitative discussion of "benefits" may resonate better than one of "impact" per se. 

Thus, this is not a figure that we use on our Impact page or within our broader impact measurement framework, as we agree the data is not yet robust/tight enough to be used meaningfully within impact measurement initiatives. 

Nonetheless the Customer Benefits page is due for a review and we will certainly keep your comments in mind. And thanks to Evan for pointing us to this revived discussion!

Regards, 
Darin

Comment by Betsy Teutsch on October 2, 2013 at 7:05pm

Thanks, Evan.  What is exciting about solar lamps is the health improvements, and the ability to reallocate income to more productive uses, and of course the global eco-benefits from eliminating solar.  The reallocation amount makes a big difference - no one will get out of poverty by freeing up 1.5% of their income.  The other benefits still accrue, of course, but the storyline is different.  I can see why DLight would go with the highest defensible number, though I notice they don't site sources.

Comment by Evan Mills on September 29, 2013 at 6:26pm

Betsy - We would all benefit from more precise analysis and language to clear up this great ambiguity that has pervaded the global discussion on this important topic.  Note that the d.light page that you referenced says "may spend ...."

Some of the examples here focus on highly aggregated results, where a value for the poorest quarter or fifth of the population is given.  Other examples may point, rather, to the very worst cases.  Neither really tells us what the large group at the real bottom of the pyramid (still a subjective concept) spend on lighting fuels.  I will ping some of our d.light members to see if we can better qualify the number on their page.  Thanks for sharing it with us.

Most of the information I've seen suggests that lighting fuel spend as a % of income increases steadily as income falls. Personally, I think standardizing on something like the lowest decile of income would be particularly informative.

Comment by Betsy Teutsch on September 28, 2013 at 12:08pm

d.light says studies show 15-25%.  That is wildly different! http://www.dlightdesign.com/impact-dashboard/customer-benefits/

Comment by Evan Mills on September 24, 2013 at 12:05am

Here's another illustration (Djibouti this time - see chart on page 4).

http://pdf.usaid.gov/pdf_docs/PNACY452.pdf

Note how the share of kero in household expenditures declines as income rises.  Resolution is low (just quartiles).  Effect probably continues into the lower income segments.

• ‹ Previous
• 1
• 2
• 3
• 4
• Next ›
• Page