Abraham Asfaw

Economist at Amazon, Econ PhD, former assistant professor.

[Abraham Asfaw @abrishAbebe]

@SuraYilma @addisstandard Ethiopian business don’t have the capacity to withstand a short term shock let alone the one that is happening right now. What will happen is that the future business owners in the country will be non-Ethiopians.

[Abraham Asfaw @abrishAbebe]

I was in Addis in December and this is my observation on the changes that is happening in the country. If this trend continues, Ethiopia owned business in Ethiopia will become a minority.

[Addis Standard @addisstandard]

#Opinion: Cost of Modernizing Addis: When progress pushes small businesses out
https://t.co/Sfwwu4MpxA

In recent years, #Ethiopia is undergoing rapid transformation, with digitalization expanding across daily life and #Addis_Abeba experiencing one of its most ambitious urban overhauls. However, Abraham Abebe Asfaw argues in this opinion piece that such ambition risks overwhelming the small and medium enterprises (#SMEs) that employ roughly 70% of the urban population.

He explains that tens of thousands of businesses are cancelling their licenses. This issue extends beyond tax compliance, as the combined pressures of taxation, informal payments, and displacement have eliminated the “margin that allowed firms to survive.”

Abraham emphasizes that “ignoring this distress is costly” and cautions that “development without guardrails risks becoming extractive rather than enabling.”

[Abraham Asfaw @abrishAbebe]

@nemozen @alemayehuGeda Additionally for some reason, our PM believes that investing on beautifying cities across the country is more important than the troubles the country find itself due to his reform. You can imagine how these unproductive investment will erode the supply of USD for importers.

[Abraham Asfaw @abrishAbebe]

Good time to revisit this in light of what IMF said about the pressure from the parallel market on birr. @nemozen it has now been a year and things are definitely not going the way the simple demand and supply model predicts. :-) also tagging @alemayehuGeda

[Addis Standard @addisstandard]

#Commentary: #Ethiopia's Currency Market: A fragile convergence and the path to stability

In this commentary, Abraham Asfaw (PhD), an economist at #Amazon Inc., analyzes Ethiopia's floating exchange rate decision for the birr. He claims the convergence between official and parallel market rates is fragile without government intervention involving parallel market actors.

Abraham observes that official market intermediaries have adjusted their rates, but banks are still unable to meet the high demand for foreign currency. He argues that this forces importers and travelers to use the parallel market, keeping rates high. He suggests a balanced supply and demand is key for stability in the official market.

Abraham attributes the divergence to structural challenges, stating that banks operate in an oligopolistic environment and fail to meet demand. He also notes the parallel market's competitiveness, especially among remitters, complicating efforts to shift flows into official channels.

He recommends two strategies: offering competitive rates to remitters and lowering barriers for smaller intermediaries in the official market. Abraham believes these measures could reduce reliance on the parallel market and stabilize the birr.

https://t.co/PwY8Eam18V

[Abraham Asfaw @abrishAbebe]

@nemozen @alemayehuGeda *to reduce the pressure due to currency market imperfections.

[Abraham Asfaw @abrishAbebe]

@nemozen @alemayehuGeda I will find the interactions we had here. But how did you believe convergence would be achieved despite the imperfections you had in mind at the time ? I just want to understand the logic.

[Abraham Asfaw @abrishAbebe]

@nemozen @alemayehuGeda Well at least at the time, your prediction didn’t take the currency market imperfections. Actually , I wrote this piece as a response to the debate we had here. The rate didn’t converge. We had a price matching game and the banks gave up on that game after a certain point.

[Abraham Asfaw @abrishAbebe]

Z cost of getting excluded from AGOA is now 0 for ethiopia, thanks to z tariff zat z US impose across z globe. Companies zat r located in high tariff countries will restart searching 4 cheaper labor. Is Ethiopia ready?Or is zis going to be another lost opportunity?

[Abraham Asfaw @abrishAbebe]

The real threat to American prosperity . https://t.co/bwAXo8X0VH

[Daron Acemoglu @DAcemogluMIT]

Some questions triggered by the release of DeepSeek R1 on January 20. These are formulated as questions, because I do not know the answers and it may well be that most of these answers are only things we can find out over time.

First, perhaps the most important question is this: does DeepSeek’s success mean that the US tech industry was approaching the problem the wrong way?
US AI investment is massive. Goldman Sachs estimates that the tech sector is set to spend $1 trillion: https://t.co/t9NiJfg1ZL

For a long time, a number of commentators (including myself) have questioned the direction of AI investment and development in the US tech industry. To the best of my understanding, all of the leading companies are following essentially the same playbook (with the small difference that Meta is partially open source). These companies are unwilling to consider different approaches than foundation models pre-trained as next word predictors on massive data sets, and, for the most part, anything other than diffusion models and chatbots aimed at performing human tasks.

While DeepSeek is not reinventing the wheel and is broadly within the same agenda, it appears to have relied much more heavily on reinforcement learning and mixture-of-experts methods and refined chain-of-thought reasoning very effectively.

As widely reported, it has also done so at a fraction of the cost of the models of leading companies, about $5.5 million, as compared to sums running into hundreds of millions of dollars for the leading models.

One interpretation therefore is that the US industry was blind to alternative, cheaper and more promising approaches. By the way, this type of “groupthink”, combined with hype, is what Simon Johnson and I predicted in Power and Progress, written before the generative AI saga began: https://t.co/NtJ8Ae5sAT

So, put differently, the first key question triggered by this episode is: are there other even more important things to which the US industry is blind to? Could developing these models in a more “pro-human direction” be one of these things that is promising but completely and collectively ignored by the industry?

Second, is this episode proof that China has leapfrogged or is at the cusp of leapfrogging the United States? If so, does this mean that innovation under authoritarian, top-down institutions (or what James Robinson and I have called “extractive institutions), can equal or exceed more bottom-up innovation?

My bias here is to think that innovation is hampered by top-down control, as James Robinson and I have argued in Why Nations Fail: https://t.co/eOODPRgA9s, and also in The Narrow Corridor, https://t.co/QewtMlMsdq

But I admit there is this possibility now. We will just have to see.

Nevertheless, I would like to point out that DeepSeek is building on years of advances in the US (and some in Europe). More importantly, all of the methods used by DeepSeek have been developed in the United States, some of which, such as mixture-of-experts models and reinforcement learning have been developed in academic research decades ago; and some like transformer models and chain-of-thought reasoning have been introduced and used in leading tech firms. Nevertheless, DeepSeek has combined them differently and very effectively. It remains to be seen whether Chinese firms and academia can really take the next step of coming up with game-changing techniques, products and approaches.

Moreover, DeepSeek appears rather different from other Chinese AI firms, which often produce products and technologies for the government or with government funding. See, for example, https://t.co/by1RyLx7dS https://t.co/m427L48RRZ. In some sense, the company may have been “under the radar”.
Now that it is no longer so, will its creativity and dynamism continue?

My interpretation is therefore that what we have witnessed is far from conclusive evidence that the Chinese model can outperform innovation in more open societies.

Third, does this mean that the US approach of export controls and other methods to hold back Chinese AI research has already failed?

I believe the answer to this is still unclear as well. DeepSeek trained their leading models, including V3 and R1, on older, less powerful chips. But they may need the best chips for the next advances and for scaling up.

My interpretation here is that a complete zero-sum approach to China was unworkable and a mistake. Such an approach only makes sense if you believe (a) we are heading towards AGI; and (b) whoever gets to artificial general intelligence (AGI) first will have a huge geopolitical advantage. Neither of these two assumptions may be warranted (and more on AGI below). If these assumptions aren’t right, there are many areas in which the US and China can collaborate. For example, if innovation in one country enables models that increase human productivity or help us regulate energy better, they would be beneficial to both countries, especially if they are widely diffused and used.

Finally, is DeepSeek taking us a step closer to imminent AGI?

The company’s aspiration (like their American cousins) is AGI. Models that are cheaper to train and effectively using reinforcement learning could be game changers.
But ultimately, as noted above, these are known methods and making the training of these models cheaper is not miraculously going to get us to AGI in the next few years. Whether near-term AGI is an achievable goal remains an open question (and whether it’s a desirable goal is even more questionable).

[Abraham Asfaw @abrishAbebe]

@nemozen @tirusewt @alemayehuGeda Ok I can see what you are saying here. How competitive is the bitcoin mining business? This is to say that would the non-stranded energy price in Ethiopia be high enough to force them move to stranded locations?

[Abraham Asfaw @abrishAbebe]

@nemozen @tirusewt @alemayehuGeda why would they care selling it to factories who can employ the poor and households. That is why I mentioned a story on the relationship between oil revenue and tax enforcement capacity.

[Abraham Asfaw @abrishAbebe]

@nemozen @tirusewt @alemayehuGeda How would this help the Ethiopian poor? That is my question. U could say it doesn’t hurt them. Then I would say may be for a short time& it can hurt them in the long term if it changes the behavior of those who distribute electricity. If they can collect enough from
Mining,

[Abraham Asfaw @abrishAbebe]

@nemozen @tirusewt @alemayehuGeda That is to say at a point of time what you are doing is Pareto improving. It enhances welfare without hurting others. But it can change the behavior of the government and hurt the poor.

[Abraham Asfaw @abrishAbebe]

@nemozen @tirusewt @alemayehuGeda If I were them, I would develop a metric to measure how inclusive or shareable an idea is before letting one put in the ground. Even if the idea is Pareto improving in a static sense ( at a point of time). It can hv a long term -ve effects in a way that hurt z poor(exclude them).

[Abraham Asfaw @abrishAbebe]

@nemozen @tirusewt @alemayehuGeda I will watch it. But z priority of the country is completely inverted upside down almost in all dimensions. They put billions to beautify the city mostly at z expense of z welfare of z majority. If u ask them how this will help the 80%, they can’t give you compelling explanation.